Hey BZ, I think one main reason a lot of your videos are misunderstood is that they they are so long and a lot just gets lost in the sauce lol. If you made a clear summary for people who are new to tweaking, non-native english speakers, or who may not even have the aptitude for this stuff, I bet many more people would understand. Edit to say: Source... I have experience in teaching. :P
yeah , he dosn't use any chart or tables or a simple notepad that are good for non-native English speakers videos are too long to understand and there is too much testing this and that . it's like streaming type of content . dude it's youtube , just give me the juice first then we can sit and watch rest of the video talking about ac/dc/bbc . his voice and delivery is so good that i can site there and listen for hours in this video he get started 37k with intel default setting . our goal is to undevolt cpu with CEP turn on . we end up with 39k and we didn't saw how it is with temp or cores frequency or voltage . idk what i just watch tbh . we already know that you can undervolt a little with CEP on but it doesn't let you get a good temp or low voltage . that is why peoples turn it off .
Not that I completely disagree but I'd say his content is probably not aimed for that target audience. It is understandable and testing real-time all the possibilities is a great teaching method but: - you have to do your homework (read/watch other source material - Intel spec, for instance) - you have to pay attention, like really pay attention and try to anticipate the outcome of the next 'text' Then he explains why the outcome was this or that on each scenario. Really I find it to be great content in this format. Not found elsewhere.
I agree with @@DeliriumT - if the goal is to learn, like to gain understanding to the point where you can adapt it to different situations or even come up with some of this yourself, this format is great. At least, short of giving people exercises that make them follow along. It feels a lot like how university courses have lab practical or assignment discussion periods. But if the goal is more about the bottom line, like "job training", there are plenty of mainstream outlets that can digest things into a tutorial or instructions to follow. Sorry I don't know the terms in didactics, but this doesn't seem far off from their best practices.
I watch them at x1.5-1.75 speed to not get too distracted while watching them... but overall I don't really see this changing, that seems to be just his style and it's unlikely to change... ever
Based on Intel documentation, CEP is the mechanism by which Fast-V Mode operates. It exists so CPUs can be safely used with low quality motherboards and power supplies that would crap themselves when trying to power a 14900K under heavy load. I think people with higher-end motherboards and power supplies would be fine with CEP disabled, more safety features is never a bad thing.
In today’s world you need a PhD in electronic engineering to understand and safely operate a piece of silicon that doesn’t even work properly by default 😂
Another reason why you go with AMD, just plop it into the motherboard, update BIOS and enjoy your new PC... None of this rocket scientist crap is needed to just play some damned video games
@@lessdatesmoreonmyplates1457 I remember when that was being said the other way around when AMD CPUs were exploding. Always some corporate fan ready to pounce on an issue to say "Should had bought from this company that doesn't pay me to sell for them, but I do it anyway!". Maybe you do get paid, in which case, fair play. In many cases though, a lot of these people aren't getting paid for their brand tribalism, and it's just pathetic that they hang around their corporate biases competition videos, forums, subreddits just to say "This is why you buy AMD!". They only do it when the other side is down, when their own side is down, they try to lie to everyone that it's actually up. I've seen Intel fans do it too when earlier in the year, you could just turn on a Intel PC and it worked while AMD was having the issues.
@@lessdatesmoreonmyplates1457 Agreed. I built my PC only 6 months ago and I'm thinking of cutting my losses and getting a 7800X3D just so I don't have to deal with this BS anymore.
Thank you so much for making this and all the other videos as of late!! It has been a INVALUABLE source for me as a novice bios tinkerer. This kind of information and testing is so scarce, not even the board manufactures have a site you can go to that breaks down and explains how a mother board works. You have filled this gap of non information that really shouldn't exist. Its so weird that there's not comprehensively direct information available when using someone else's product that is designed for you to tinker with it. How exactly are we suppose to learn this stuff on our own without studying to becoming an electrical engineer?!
I've been out of the overclocking game for a long time now - I'm still rocking a Core i7 2600K, and at 4.3 GHz on an air cooler! - so I'm completely lost with all the new configuration options. Maybe I'll get back into it whenever I get around to upgrading my desktop haha. But I'll still chill and listen to Buildzoid talk. Maybe I'll be able to piece together how some of the new silicon is configured.
Thank you so much for your efforts! Been devouring your videos lately cause I was concerned about my i9-14900k which I bought in May 2024. I'd love to see you doing some AC\DC Loadline testing on motherboards with the OnSemi (NPC81530) controller, cause I have a Gigabyte Z790 UD which uses it. Keep up the great work :)
For stressing arbitrary combinations of test program + frequency + power, you can limit peak clock frequency with the CPU frequency scaling controls, then set a CPU quota (cgroups in Linux, Battle Programmer Shriase in Windows) to PWM the test program such that it doesn't hit power or thermal limits.
Another thing I see a lot of people incorrectly suggesting has to be turned off in order to undervolt, is "Undervolt Protection". All that setting does is establish a floor for the voltage at POST, it doesn't prevent you from setting a lower voltage at boot in the BIOS
No, you do not only mostly affect high load voltages with AC LL, run Geekbench 6 and compare VID (or vcore, whatever) on light workloads, you'll get significant difference in VID between 1.1 and 0.4 mOhm. You can already see the influence just in the bios, there's tons of voltage pumped regardless of load, since AC LL is a flat offset calculated from AC and probably ICCMAX. The same value is being added to VID all the time. Adaptive 1.435V + 1.1 mOhm LLC + 0.65 mOhm AC LL = up to 1.48V in GB6, 1.435V + 0.06 mOhm LLC + 0.1 mOhm AC LL = 1.41V in GB6. Also, MSI gives the per ratio voltage offset, but it's completely broken too, so while bIntel gives that opportunity - motherboard manufacturer has to be not idiotic too. Also2 - at the time of writing this comment I didn't get to 26-ish minute yet, where you've said just what I wrote :P. Which is also contradicting what you said at the 12 minute mark, it's influence is pretty much linear, although by dropping just AC LL you actually hit high load voltage stability limit way sooner, than low load voltage stability.
I have enough another issue, if I keep the 1.0/1.0 AC/DC (recommended LLC4 setting with latest ASUS BIOS, then try to undervolt with the SVID (requires a ton of negative offset to get the same voltage as before), it ends up tripping up the 307A ICCMAX on my 13700K. Meanwhile if I use my normal undervolt mode of 0.4/1.1 with LLC3 with SVID offset of -0.05 but with CEP disabled, it doesn't trip the ICCMAX and downclock. The only way I can somewhat get CEP to work is use LLC5 with 0.4/0.73 AC/DC with SVID offset of -0.11 with CEP enabled. I still end up with about a 200 point lower C23 score, I even tried creeping up the AC LL to .5 and .6 with more SVID offset to compensate, still noticed the loss of roughly 200 points if CEP was enabled vs disabled. I would gain those 200 points back if I left it locked at 0.73/0.73 on AC/DC with roughly a -0.19 offset, but the voltage under load wasn't constant, it drooped heavily in the beginning and raised up over 0.02v by the end of the test. With the other settings, it stayed constant the entire test.
@@PolskiJaszczomb Could be that too, it's under IA: Electrical Design Point/Other and PL4 is listed there too. It wasn't maxing out PL1 or PL2 though. I'm convinced the stock out of the box settings is trying to kill my CPU. Gave it vcore's in the mid 1.4's and tried to run cinebench23 at 1.32v before immediately throttling. It's trying to run it at full load at the exact VID listed in the VID table with no vdroop at all! I'm not sure if this is on Intel or this is just an ASUS screw up. Really considering going back to my old BIOS I was on with just a voltage limit since it hasn't given me any issues for 2 1/2 years.
I did his settings with the LL to high/55/55/-100v offset, and my 13900k went from running above 1.5v to running up to 1.35v max for what seems to be the same if not better performance. So idk, he says you don't need to do it, but it seems like it really helps. So I'm even more confused now. My clocks are the same, if not better since I seem to hit 5.8Ghz more often, but maybe that's placebo. Temps are waaaay better, 20 C difference on cores that were hitting 70 C on startup alone. Now my highest core is currently 53 C. Also I'm running a Gigabyte Z790 Aorus Elite AX gen 1.
I think one of the big takeaways from this whole debacle is that Intel and board partner documentation is an absolute joke. They are shipping products and features with no clear or comprehensive explanation for how they work and leaving it up to the consumer to figure it out through trial and error. It's actually insane that we have to rely on what are the equivalent of unverified community forum posts for information on how the commercial products of these multi-billion dollar corporations are supposed to be operated instead of having a detailed manual we can refer to.
I agree the video is a tad long, but I need to offer confirmation that your recommendations are spot-on. I have the Z690 AORUS Elite AX with F30 BIOS (0x12B Microcode) and a brand new i9-14900K (never subjected to faulty default voltage settings, BTW). With your settings: Adaptive Vcore, Legacy mode and -100mv offset, Loadline Calibration = High, VR Voltage Limit=1400, and IA AD/DC Loadline=55, I consistently get CBR23 score of 39500-39700. And highest vcore never exceeds 1.34v under all core load. And I'm still seeing 2 Perf-cores hit 6 GHz periodically and multi CPU ALL-core loads at 5.7 GHz on the 8 Pcores. I suppose I could tweak at the margins, but I really see no need to as these are rock stable under gaming loads, and CPU voltages drop to .77-.78 v under idle or low loads. Thanks for your explanation of this most-confusing topic.
for my system; 13600kf>msi z790 gaming pro wifi, setting AC/DC loadline at 100, llc8 and setting cpu vcore to adaptive+offset with a negative offset seems to be giving the best results. i dont have fancy tools to test exact voltages, but in hwinfo and cpu-z my vcore and vid are the same (+/- 3mv) and i havent had any instability issues. excited for you to test the msi bios and hear your thoughts. thanks for the great content!
Honestly, i dont see the point in using LLC8, for me anything above 4 is just unnecessary heat i tried two ways: 1. using LLC6 with undervolting (no CEP) just surfing an internet browser causes the CPU to go all the way up to 80+ degrees 2. using LLC3 no undervolting (no CEP) doing same silly test, will max at 65+- degrees more LLC = more heat = less performance in real life things sure there is still the silicon lottery, but for most CPUs, i would stick to default settings with the minimal LLC that is still stable (LLC3 in my case) also i would suggest using ac and dc load line to be exactly same as your VRM load line, 1.1 for Asus
@@CASIOMAN20 For MSI, the higher the level, the higher the Vdroop. So level 8 (which is almost the same as Auto, but only for Auto voltage (resp Auto voltage +Offset)) would give the lowest Vcore and temps.
If you set LLC to "auto" (1.1 mOhms), don't set an high enough Voltage Offset and then limit the VRM to 1.4 V, you will likely see lower boosting P-Cores. Just because it needs a higher VID and if that doesn't fit under the 1.4 V ceiling, the next lower boost stage is used that fits beneath that ceiling. I've tested this on an Gigabyte Board lately. Lower LLCs (and matched AC and DC Loadlines) obviously can fix that - depends on the quality of the CPU.
@@notexactlyrocketscience and yet he wonders why people don't get what he says - he needs less fly-by-seat videos and to actually plan it out and less of the "well.... [ramble on some disclaimer that doesn't really matter to the main point]" moments.
Legacy offset is entirely ignored on a 13600K AND 13600KF with the same motherboard and settings in my experience. Even tried CEP disabled. DVID has been the only way for me to undervolt on a 13600K/F on both the Z690 Aorus Master and the Z790 Aorus Master X. Oddly though, HWInfo detects the offset correctly when it is applied, but the VID and VRVOUT do not change.
I am pretty sure there are weird quirks going on here from motherboard to motherboard. Or same board, different CPU. I've experienced some weird stuff as well. It's supposed to work in a certain way but it really doesn't.
Did you by any chance switch the profile from Intel Base Line to something else? I found by doing so you introduce a lot of weird stuff. Also there is a feature called undervolt protection which I did not have to Disable for my i9 but for an i3 I owned I had to with led to a bunch of problems! TLDR; What are you current BIOS settings?
@@interceptor001 Yes, I've tried with CEP on and off, yes I've turned intel's default settings off. I've mimicked BZ's exact settings and it's not respected. I've tried on default and with custom settings. I believe it is a 13600K issue. I feel the need to reiterate that this behavior was present 2 years ago on the Z690 Master with a launch batch 13600KF, and is still present.
what does xtu use? I thought it worked in there with undervolting, but not sure. And could you do a video on a asus z790 bios? It's so different than others, I don't know where I find things, because they have other names, are other places and things like that. I just don't know how to apply your showed settings 😅
I'm tired of all this. I installed the new microcode. You recommend enabling CEP, does disabling it remove the 1.55v hard limit for VID voltage included in the new microcode? In the meantime I disabled CEP and set an IA VR limit of 1430mV (below that number my cpu clocks decrease) in addition to the hard limit which I hope is active... (I have a 13700k, an Asus Z790 mobo)
I was able to drop voltages by -0.175 on my 14700kf with CEP enabled and using your loadline settings. Also I was able to increase clocks by 100 MHz, so they are 5.6 GHz all core, 5.7 GHz 1-2 cores. In HWINFO, the VID doesn't go above 1.199 V and VR VOUT 1.190 V. R23 ~37000 with 253w power limit. Also put 1.4 V limit just in case. Have to use my PC more to be 100% sure about stability but seems to be doing good.
you are missing an R in undeRvolt in the title, as always great video, whoever doesn't appreciate this type of content or the duration can go for eye candy content with 0 value in plenty other channels and thats fine
MSI seems to have this on lock. No brainer with their boards. Lock the cores at 5.6 or 5.7 and set CPU Lite Load to mode 7 - 12. Lower is better as long as it doesn't crash. Too low and it will crash just from the CPU not getting enough voltage. You will see performance loss as you work your way down from Mode 12 once you get to low. When you hit that performance loss go back up one mode and enjoy your cooler, more power efficient, fast @$$ CPU! Don't let 2 cores boost to 6Ghz it over volts ever single core. Even the E Cores.
The thing is that MSI Lite Load is exactly what BZ explains as the second wrong way to undervolt. It lowers AC_LL and DC_LL. You can see what it does if you set your "normal" LiteLoad, safe & reboot and then change LiteLoad mode to "advanced". You can honor CEP with the advanced mode, if you bring AC_LL, DC_LL and LLC together manually. You can find a table online that shows you which LLC equals to which AC_LL/DC_LL.
@@bauerns5er MSI's current solution to that is to set the Lite Load (and so AC/DC) stupid high under Intel defaults. So you can drop it a long way from its default value and still not hit the CEP tolerance limit.
hour long videos for something you can do in 10 minutes on an MSI board with Lite Load. Lock P cores to 5.5. Lock E cores to 4.1. Set a negative 0.055 voltage offset. -2 AVX ratio. BCLK 100mhz lock enabled. Speed Shift enabled. Intel Turbo Boost Max 3 disabled. Enhanced Turbo disabled. CPU Current Limit 307a. Power Limits 250w. GT Current Limit 30. CPU Lite Load 9, or 10 if you want to be conservative. CEP will probably never activate anyway so you might as well leave it on. This has got me between 38k & 39k in Cinebench R23 runs, voltage will be below 1.4v (I've done runs where the max was 1.309v) and you won't go near throttling with decent cooling and a contact frame (360mm aio, max core temp 93c on my hot core #5, rest below 90c, e cores below 80c), and CPU power doesn't go above 230w. Using Z790 Carbon Wifi board, 2x32gb ram and 14900k with the new bios installed.
CEP was introduced due to silent corruption since 12th Gen I assume. I udervolted a few builds without CEP and they eventually corrupted the windows installation. This has never happened with 11th Gen or older, they would just freeze if voltage wasn't enough. Update: The root cause was SA CEP, ALWAYS ENABLE SA CEP even if SA voltage is left on auto
On my gigabyte board, intel default settings force something called "Dynamic Overclocking Undervolt Protection" which you can see on hwinfo. You can't undervolt either way when this is on, and it doesn't go away unless you disable intel default settings (even setting IA CEP and Undervolt Protection to "disabled" will not work)
NGL, i really don't like Nvidia driver naming scheme. These numbers are not telling. For example AMD currently uses this scheme for public releases. YY.M.V (aka 24.7.1 would mean first driver in July 2024.) Professional usage drivers have naming YY.QX (aka 24.Q3 would mean driver from third quarter of 2024). And test branch drivers have this naming scheme XX.YY.ZZ XX is usually corellates with year, but can lag behind significantly. (Aka 24.20.08, meaning major version 24, branch 20, version 8, or something like that).
Love your detailed content as usual good sir, but i would like to point out that different AIB's are behaving very differently when it comes to CEP & AC_LL including undervolting. Even more so with the x129 Microcode update, Asus for example, at least in my case fails to work whenever you change anything manually regarding AC_LL or even a mild VID alteration of 0.025Mv. Although SVID seems to be alterable, without issue. Very odd... It looks like they've locked it down with CEP enabled (at least from my perspective) when running Intel Baseline on this BETA Bios. Asus Maximus Z790 Formula btw.
I'm not a fan of keeping intel default profiles enabled. The stuff they lock out from motherboard to motherboard is just so random, it's annoying to deal with. Not everything translates well when trying to help others. Just go full manual mode, disable it. I've had great results with Turbo LLC and AC LL as low as possible, small offset on top. Not every CPU likes -0.1V, not even -0.05V unfortunately.
Just all core overclock your CPU then, or put a fixed max on Pcore and Ecore. Like 5700mhz and 4200mhz. Then with undervolting, the VID requests will never be too high.
I have a completely different problem with my new build and I'm pretty much at my wit's end...I have a brand new i9-13900f that has never been used on anything other than the intel default performance settings (i don't have extreme because I'm assuming the i9-13900f doesn't support it) + 0x129 bios. For some reason any time I run a stress test to push the CPU to full load, it only achieves full load with 219w power draw (default pl2 for this chip) for a couple seconds, then throttles itself severely back to 65w (pl1) where all the P cores drop from 5.3 ghz to 2.2-2.4 ghz. 1. I have not touched anything else in the bios other than change to intel default settings, and turn on XMP1 for my 6000 mhz memory. 2. I know I'm not thermal throttling on the CPU at least, because I can see that my temperatures only ever reached maximum 86-87C before throttling. I'm using a Ryujin III 240mm AIO. 3. I have the Asus Tuf gaming B760-PLUS WiFi board, which I know isn't a Z series board but considering the i9-13900f isn't as demanding as the other skus (even less demanding than an i7-13700k), I can't imagine the board's VRM's being the issue? The temperatures seem fine to me. I have an 850w psu, so power is not an issue. I'm starting to suspect that it's the intel default settings that is the issue, since I don't have any thermal issues. Now I'm trying to figure out how to fix this issue while staying within intel's specs. With all the buzz about the stability issues, I'm really concerned about changing too many things within the bios that may potentially harm the CPU. Has any one else experienced this sort of behavior with the intel default profiles?
Huh, that seems strange. Have you checked all temperatures - including VRM, Ram, chipsets? Cause from what you say your CPU should stay at the boost longer and 2.2GHz seems too low for 65W even. it should be more along the lines of 2.7-3GHz. Does the board have any stupid "stability enhancements" or other things like that? @@yarost12 tau should be 56 seconds.
sounds like the intel default settings are enforcing the 65W power limit for the 13900F. As for how to raise the power limit. You might have to disable intel default settings to do that. I unfortunately don't have any high core count non-K CPUs to test with.
@@ActuallyHardcoreOverclocking but even with the power-limit the CPU should not run that low. The few 13900 i have seen - none of them were under 2.5. He'd have to have gotten a real dud there.
thanks for the guide using a i9 14900KF on a aorus z790 elite AX ice with an AIO 360 watercooler i got a 35022 cb score after changing the vcore offset to -100 i got a 35795 cb score setting the vcore loadline calibration to high put my cb score to 36472 still not nearly as high as yours but im happy it improved, was not able to change the AC/DC loadlines as it was unclickable with the latest bios
So, did the new microcode actually do anything besides set limits I could already set in the bios? I'm still waiting for a non-beta release for my board (z790 Aorus elite 1.0 rev)
I updated my board with z790 bios flash, I’m a newb . My pc now runs much harder it throttles all the time like it needs wayyyy more power from the fans . Not sure what’s going on.
I have i7 13700k and with CEP disabled my VID's never go above 1,360V, but with CEP enabled, my VID's jump to 1,440V. My current is limited to 307A and PL1 is set to 125W and PL2 to 253W.
Same here. I have an MSI mobo with a 14700k inside and I have the same results as you. IA CEP disabled, CPU current limit 307A, PL1/PL2 at 253W. Also I changed the value for CPU Lite Load to Mode 7. I never go above 1.36v and my temps are in the high 60s and low 70s during max load in Cinebench 2024.
this will be true on Gigaboards too - a lot of what he shows is very specific to the specific settings he uses - and mostly no one is leaving the things on Auto that he does for his "dont do this" video - but who knows, many lots of noobs/relative-noobs in the viewership.
Hello, I don't really have much knowledge with the bios, but I have an i7 13700k and a gigabyte aorus elite ax ddr5 board, currently my cpu reaches 5.4ghz with a maximum recorded peak voltage of 1.378 on very few occasions but in general the peak maximum is 1.362, I have an lp1 and an lp2. Limited to 200w and maximum temperature of 80 degrees, latest bios for 2023, I don't know if I'm okay with the degradation, or what parameters do you recommend?
@@dizzyfist7 legit same. tried new microcode today too w intels bullshit bios settings today and i was hittin 1.44-1.46v. horrible. going back to OC undervolt on my 14700k.. fuck intel.
MSI doesn't have the voltage limit setting, someone asked on the forum and apparently they've made a decision internally not to offer it. In terms of the CPU core voltage modes, they have Auto, Adaptive mode, Override mode, offset mode, adaptive + offset mode, advanced offset mode (VF point), adaptive + advanced offset mode (vf point), adaptive + offset + advanced offset mode (vf point). I'm not sure what they all mean exactly, but Adaptive + Offset Mode allowed me to lower voltages further without triggering CEP, when I was already on the edge of triggering it with my load line settings.
Same here. I matched my DC LL and LLC, set AC LL TO 15, adaptive + offset, auto and -100 mv. My CPU at 55/43/45 hits 40,800+ on CB23 and maintains 55 on full stress tests. And my VR out and vcore are almost identical
Advanced Offset Mode (VF Point) also works well on MSI, just tested it. And then you can add your undervolt only to VF points 6-11, for example. Or whatever matches your top ratios.
@@janos1312 I RMAd my 13900k and got a 14900k. Here’s my current settings on my MSI board: LLC 6 (50 mohms), ACLL and DCLL 50. Adaptive-auto, -160 mv offset, PL 1/2 260w, 400A. I leave clocks stock. This thing runs great
IDK if it ever happened to you, but it used to be that holding down delete key didn't work. And some settings in startup made entering BIOS window like 0.2 seconds. Ever since I had to deal with mobo like this I do exactly the same thing - smash that delete key as often as possible.
@@jannegrey I think that was the case with lower end boards, I had a b450 that didn't do it, my z790 does tho, I thought that was the case Thanks for educating
@@InfinitycgIN From my personal experience (I had PC's from 1997 and I OC'ed since 2001) most boards had it. It's plausible that boards from the last 5-10 years don't do it as much, but I wouldn't know since I always smash that delete key. It's basically a tick now ;) Depending on the "provider of BIOS/Firmware" (Im not exactly sure now, how this thing was called) - there was AMI and something "star" and couple others, they basically had an overlay that displayed memory in system and all the "important numbers". And often when this "overlay" was engaged you couldn't get into BIOS. You could get into it either before that (when the screen was black) and/or sometimes after it. But CPU's got so much faster that what was initially let's say 3 second window of opportunity, with upgrade could go easily sub second. To turn this off (if there was even an option) - you had to get into BIOS. Something similar can be observed on relatively modern Laptops. They often have overlays and sometimes you either have to click delete key in very precise moment, or you have to use different combination. I'm glad that it isn't now a case that you need to smash that delete key, but it seems to be like inverse of what is happening to Windows and going into safe mode etc. You have to know specific combinations, everything is hidden and stuff like that. Mind you - I'm unsure if the signal from keyboard wouldn't be disconnected (at least on older motherboards) the moment PC turned off for a moment, when you restarted. In such case, you'd need to pull your finger of it and push it again. That might have differed depending on the connector (USB or PS/2). The point is that if you re-booted your PC 5 times in a row and always managed to skip that window of opportunity, you would be smashing that delete key as well ;) And mind you - even then you didn't always manage to get into BIOS (because Window of Opportunity was shorter than keystroke and sometimes keyboard was disengaged prior to this - pain in the ass for troubleshooting). And thank you for teaching me that this is (mostly) no longer the issue. I did notice that it was starting to get better around....2005? But you still got some setups that were..... difficult.
I've posted up a couple times how he can create a simple desktop shortcut that will immediately shutdown the machine and boot into the BIOS. Guess keyboard machine is his thing.
I don't understand something about AC = DC = VRM LL paradigm. If do it this way you basically have almost no v-droop, correct? Because even if you set your VRM loadline very droopy, like 0.9 mohm, and then "tell" the cpu about that vdroop (AC LL) it will just "ask" for more to compensate. Or does it actually makes a difference if we use let's say 20/20, compared to 90/90? Isn't actually something like 50 mohm of diffrence between vrm load line and AC LL considered ideal amount of vdroop or is it different for 13/14th gen?
The function is that if you want to use CEP still, then you have to set the AC/DC to roughly the same impedance the the loadline calibration you have chosen. The best results often come from somewhere in the middle. Medium/High options. You need to know the impedance of the VR though. It is very often different between manufactures. For example my MSI is LLC 6 AC/DC 40 LLC7 AC/DC 69 where as Gigabyte (such as this one and the gaming x etc) are LLC medium AC/DC66 and LLC high AC/DC 55. It allows the caluclations to fall within the tollerance so that when you set the VID lower it isn't seen as an excursion by CEP, so you will get your frequency at that VID instead of it clock stretching or crashing.
@@sp00n hours of research. Found threads where someone had listed the numbers then made my own notes for what I wanted. I didn’t copy them all for every setting unfortunately. They are out there though.
@@sp00n I can confirm the numbers for MSI are correct. You can find out on your own. When your vid = vcore under multicore load it means that you matched your DC Loadline to vrm loadline.
Thanks so much for this content. It's really helpful for people trying to understand what their mobos do. Is it possible for you to make a video with a MSI motherboard?
I get 5-10 degrees hotter and higher voltages with intel defaults no matter what i do on gigabyte z690 gaming x 13700k... which I suspect negates the entire benefit of intel defaults to begin with. So I opted to turn them off. Not that I had any stability problems in the past year other than too aggressive undervolting.
I'm also having higher temperatures with the same processor and motherboard. The problem is that when playing everything becomes unstable. Did you find that removing the default Intel profile solved it?
Which specific motherboard are you using, and when you said the Elite has a different voltage regulator, did you mean the Aorus Elite X Wifi7 model has a different VR?
@@ActuallyHardcoreOverclockingI love how motherboard and other PC component companies make their product naming incomprehensible. Last time I spec'd out a build for a friend, I had to create multiple spreadsheets to keep track of the confusingly similarly named parts. Slap an Elite, Pro, Super, Ti, or a random tech suffix, and watch as consumers with finite time and energy capacity spend too much on parts that don't do what they think they do.
Wait what ?? Did not understand anything you said in that video. lol can you direct me in the right direction to find basics on how to under-volt intel core I5 14600k im a noob and feel weird changing setting when i dont fully understand them. Thanks in advance
Benchmark my i5-14600k. I limt pcore to 35x, and ecore to 26x. With CEP enabled I get Cinebench score 350. With CEP disabled I get Cinebench score 826. Apart from the CPU-frequency, Im running Intel Default settings. I have an Asus z790 motherboard
I also have an Asus z790 board and saw similar behavior where the multicore score goes from 24k to 11k with CEP enabled. Something about Asus' power configuration seems to trigger CEP pretty much automatically.
I think that means your tripping CEP. For example my mobo has the same bios as the one in the video where high llc matches with 55 ac_ll/dc_ll, and medium llc matches with 66 ac_ll/dc_ll. When i tried to reduce ac_ll/dc_ll even by 1, with 0v vcore offset, i.e. ac_ll/dc_ll = 65 with medium llc, my r23 scores dropped hard with cep enabled, and putting it back to 66 restored the score. Also noticed idle vcore was high when doing this with 0v adaptive offset, so i still needed an undervolt to bring that down. Also, i noticed in my bios that the description for CEP states that e-core performance is affected by enabling CEP so that could be another reason for your lower score.
Most of this doesn't really apply to me, but is entertaining to watch. My cheap-a$$ H610M has almost no voltage controls, so DVID or loadline trickery is necessary for
Offtopic but, for the 7000X3D series o/c or whatnot, after all the bios updates with latest AGESA etc., any need for an update video on what should be turned on/off? Maybe some things have improved or changed?
Asus B660-G: can't set negative VID offsets, unless using uCode 104. No undervolting on B660? The Mainboard sets 0.4/1.0 llc4 in synced mode - not going to dig into that even further. 1.4 Imc results in 1.55 CPU vddq - Asus is nuts
my undervolt on my 14600kf cpu on an asus rog strix B660-G motherboard (bios 3601 microcode 0x12B) in cr23 i get 2091 single core/ 25180 multicore. hwinfo e-core VID min 1.141v/max 1.279v/avg 1.22v. temperatures min 31º/max 68º. CPU POWER min 5w/max 122W. Vcore max 1.146v. (all this stressing the CPU) MOBO BIOS -asus advance oc profile: enabled -intel adaptative boost technology: disabled -asus performance enhancement 3.0: disabled -C-state: auto (if I enable it my pc crashes, it doesnt even enter windows) -current cpu core/cache current limit: 225 (default is 200 but it made me clock stretch) -PL1 & PL2: 181 -LLC: 5 -synch acdc loadline with VRM loadline: enabled -IA AC: 0.40 -IA DC: 0.72 -IA CEP: disabled (when enabled it works but if I put xmp profile it destroys my performance) -VRM core voltage: offset - 0.135 (I can lower it more but its a bit unstable)
I was able to adjust the settings for my ASUS ROG STRIX Z790-F GAMING WIFI MB and increased from sub 37,000 to over 39,000 on Cinebench R23 with my 14900k. In the BIOS under the Ai Tweaker tab there is the Global Core SVID Voltage setting which I changed from Auto to Adaptive, then I changed the Offset Mode Sign to -, and finally set the Offset Voltage to 0.10000. This is also after installing the microcode (0x129).
Can you make a video explaining what IA AC Loadline and IA DC Loadline settings actually do? I've watched a bunch of your recent videos that touch on them but I still don't get it. I have a 13900KF that bluescreens and I can't RMA, so I'm trying to get it to be stable. For some reason, setting DC Loadline to 1 and keeping AC on auto with the default LLC stops my CPU from crashing. I've even tried undervolting with adaptive mode and a -70mv offset and it still doesn't crash as long as DC loadline is set to 1. I've capped the voltage to 1.4v, like you mentioned in one of your videos. I'm not sure if running these settings is OK or not because I don't really get what these settings do. There doesn't seem to be a performance impact with these settings and my CPU is running under 80c. I've tried some of the settings you had in another video with CEP on and increasing LLC but that didn't get me stable for whatever reason. Unrelated but my CPU only seems to crash in games. Running benchmarks is stable. I'm currently using Ready or Not to test if the CPU is stable or not since that game seems to cause it to crash most consistently for whatever reason.
I'm so happy i have a gigabyte mobo (so i understand what bz is talking about) and i also undervolted with adaptive mode global offset and it works very well. The only problem is that i can't change the ac loadline limit which says it's set to 0 (?!) but seems like it's working well with extreme llc option. Edit: i forgot to mention that i haven't updated to the new microcode but CEP off otherwise is not 100% stable. Still works very well in terms that the performance is good with no stability issues and even with the 1400 limit it doesn't overshoot much, max 1.452 which is not that high given that i use 5.8ghz all cores with 185 power limit. So basically i found a way to overclock, undervolt and keep temperatures in check where the performance is very good and the voltages are finally "normal" for what we're used to from intel. Btw, this on a 14700k which seems more impressive to me given that is a garbage bin just like bz's 14900k bin.
@@notexactlyrocketscience After some hours of testing i found out that the sweetspot for extreme LLC is AC voltage limit set to 30 or 35 with CEP off. With CEP on you shave off some performance and the voltage shoots a little bit more. Anyway i'm very happy with my config right now, seems that dropping the AC limit from 55 (which was default on my mobo), to 30, keeps the core voltage from overshooting as much (max 1.402 from 1.452 previously), so it seems to respect the 1400 limit that i imposed, better. Very happy about that given that the performance is excellent. I also tested with Ultra Extreme LLC and find out that AC limit works at 20 or 25 best, but the performance isn't better (it's actually worse somehow) and it doesn't seem tp be 100% stable (for example i fail the low thread cpu profile tests in 3DMark), so i decided to stick with Extreme profile and 30 limit.
So, just RMA'ing because the microcode came out. My 14900k is horribly damaged just like yours, it started having issues and crashes just after 3 days of usage, but moving on with a brand new chip how do i ensure i am not damaging it again? i am ofc on the latest bios, and intel defaults, how can we be sure increasing the power limits is gonna be safe?
What's the next steps if you want to continue further the undervolt after you do 55 AC/DC ? Also i have observed on gigabyte motherboard it won't go below -0.080mv offset , it's like is not registering.
You mentioned the aorus elite (and in general the boards with onSemi conductor)...can i ask what are the slopes? Btw, i notice a weird behaviour with my z790 aorus elite. I tried HIGH with 0.5 / 0.5 mOhms (so...50 AC 50 DC) , and even by setting an adaptive offset the load voltage (cinebench) just SKYROCKETS from 1.22 to 1.32 or about that, and it doesn't even seem to be effected by my offset. I'm so confused right now. It just feels like the AC loadline is pushing it up. 13700k though, maybe the i7 behave differently because less current -> less droop? EDIT: fixed by giving a negative offset to Ring voltage...
@@ActuallyHardcoreOverclocking i tried again, resetting the bios to defaults. Just changed power limits to 253/253w/307A with new microcode and intel specs set to 'performance' (the i7 doesn't have the extreme profile). AVX offset to 0. Set vcore loadline to high, ac/dc loadline to 55/55, voltage mode on adaptive, negative offset of -0.150v. Running cinebench single core: 1.22 - 1.23V vcore (but also VIDs) Running cinebench multicore: 1.28V (but also VIDs) at 150A reported in VR IOUT (probably inaccurate?), 240W of power What i think is happening is that the loadline is probably drooping about 40mV from the initial 1.23v but then the AC overvolts for 150A*0.5mOhms = 75mV, effectively jumping from 1.23v to about 1.27/1.28V ... but maybe i'm completely wrong about what's happening (also sensors on this board suck i'm just assuming current reading in hwinfo is somewhat reliable...). EDIT: if i set -0.2 v of offset ... it doesn't matter. Same VIDs and same vcore
@@Profetorum same issue using the same hardware i notice that i'm able to go under 240W by lowering the AC/DC LL the problem is that if i go lower than 40 i will start losing performance but the wattage do decrease using high LLC i'm going to try instead of lowering the vcore offset i will try to reduce the Ring offset to see if this do something i remember i read that on a forum that on the aorus elite after certain point you dont see any improvements by lowering the vcore but i could be wrong
14900K+MSI Z790i, BIOS v19, Microcode 0x120, Cooler Noctua D15 VID Offset (set through TS): Core/P Cache -0.11V CPU LLC: Mode 4 AC/DC Loadline: 12/0.12 mOhm (the same value as LLC Mode 4) IccMax: 400A PL1/PL2: 288W UVP: Disabled IA CEP: Enabled TVB Voltage Opt: Enabled TVB Pts Config: 1 Core at 80℃ -1, at 85℃ -1; 2 Cores at 80℃ -1, at 85℃ -1 (with TVB Voltage Opt enabled, this will prevent high voltage, since there is no voltage limit function) Everything is good. R23 41000 pts (limited by PL2, around 5.5/4.3), CPU-Z 980+/17600+ (fully turbo boosted, 6.0/5.7/4.4)
MSI said on their forums (not directly, but via forum admin proxy) that they currently have no intention of implementing this setting. Quote: "After internal discussion, MSI have decided not to implement “IA VR VOLTAGE LIMIT” to BIOS. MSI suggest user try the latest BIOS with Intel microcode 0x129, which have already improved the CPU VID spiking issue."
@@Golden2Talon Yes, CPU Lite Load is the setting to influence the AC and DC loadlines on MSI. In CPU Lite Load mode "Advanced" you can set them independently, in "Normal" it uses preset combinations. But "IA VR Voltage Limit" is something different entirely. With this, you can basically set a lower limit for the voltage spikes as an additional protection. Meanwhile, CPU Lite Load may lower the AC loadline and thus the voltage, but it can't do much about the voltage spikes, other than to lower the voltage baseline they're occuring from.
@@Mike500 what you can do on MSI Boards is, change the Voltage mode to "Adaptive + offset" and then set a negative global offset. You can combine this with limiting the Boost levels and don't use this 2 core extra boost. If you also set a reasonable lower Loadline and matching Loadline Calibration, you are most likely where you would be with a Voltage Limiter (MSI Boards don't have that). But then again, you have extra warranty, so why bother?
I thought disabling cep was meant to help with lowering temps at least according to msi? Is it not worth disabling for that alone? Never saw a solid answer as to why that was the case they just said disable it better perf lower temps
I tried your config of AC_LL = DC_LL = LLC = 0.55 mOhm with CEP on and an adaptive offset (14700k, F5 bios, IccMax = 307A, no power limit). My previous settings were basically an AC_LL/LLC undervolt. The most bizarre change I noticed is that the CPU gets current-capped at a higher frequency, which results in a performance uplift in some benchmarks. I wonder whether it's a side-effect of CEP or increased voltage under load. Going to run a few more tests.
With CEP being on and following the video's concept, how can I further lower the temps? Which values can I play with so that I can get similar performance without triggering CEP while keeping temps low?
Hey! first of all, thanks a lot for the work you put in! i was wondering if you could do something like this with an asus board. like which LLC should i use and what the AI AC/DC louadline would be. thanks again!
It would be great if you make video with ASUS motherboard instead of GIGABYTE. This company doesn’t update their firmware and BIOS so I would say it’s blacklisted by many people. Thanks!
Why does this feel like Teacher has just handed out lines to all the naughty ones in the class? 🤣 Write out one thousand time - "YOU DON'T HAVE TO TURN OFF CEP to undervolt intel 13/14th gen CPUs"
Hey BZ, sounds like you're saying that there was a communication problem among the employees at Intel that caused the company to be in chaos from a business perspective!😅
Still waiting for someone to address what all of the EVGA owners of z690/790 boards are supposed to do since Evga’s MOBO team doesn’t exist anymore aka no micro code update
I had to turn off CEP to undervolt my 13600k. Tried everything else and CEP off is the only thing that permits an undervolt. Aorus elite ax rev 1.1 z790, 0x129. Clearly, there is an as of yet unknown issue with boards behaving differently. Under Voltage Protection is on. ac dc ll 50/50, vcore auto, dvid -100mv, llc high, 200A max., 1300mV limit, 125w&125w. vrout 1.126v, vcore 1.128, VIDs 1.214V. huge difference there but working on making it disappear by tinkering with ac dc ll.
About windows being slow, while this may be yet another different thing here, I remember this sort of behavior since windows 10 (never happened earlier), and it occurred when windows is absolutely hammering your SSD (or god forbid HDD...) and your poor SSD/HDD can't keep up because windows 10+ IO is just *that* bad. And yes, it is bad enough to even get bottlenecked by an SSD. If, during the slowness, task manager shows 100% HDD/SSD utilization with shockingly low read/write speeds and insane latency, you are hitting that issue (I remember getting 1000ms on an SSD, and 2000+ on HDD)
Had an HDD I was using for media storage that started to fail and it basically gridlocked my entire operating system on Windows 11. Computer wouldn't even shutdown properly with it connected.
should take a look in your device manager and enable hidden objects or whatever its called - anything you know for sure is bunk should be removed, then restart. as for usb sdd's - I only have issues randomly and very infrequently, and its not a Windows issues, its a mobo issue wrt to USB - you can change around which ports are being used and get very different results.
What does MSI Lite Load setting do (as opposed to load line setting)? Ive set mine to mode 3 which gives me much lower temps and no performance degradation on my 13700k. CPU temps under load are around 75c, whereas with the stock setting of 9 it hits 100c. Is this a good way to prevent or reduce the voltage spikes? Never had a crash in 18 months.
If you can go to Mode 1 without crashing it's even better. the lower you can go, the lower voltage the mobo give to your CPU, so it depends on your CPU quality, if it can perform at Mode 1, then you're good to go.
Lite Load DOES set Loadlines (AC_LL and DC_LL) the thing is that it hides this behind some abstract numbers on "Normal" mode. If you go to "Advanced" mode, you can see what it does and you can also set AC_LL and DC_LL on its own. Intel default goes to 110/110. If you use "advanced" mode instead of "Intel default", but enable both CEP settings, you might end up on 80/80 or something (I don't remember exactly).
MSI Lite Loads are basically just load line profiles. For example I believe Mode 3 would be an AC_LL of 15mOhms and a DC_LL of 80mOhms on my MPG Z690 Edge.
Hey BZ, I really appreciate your work. But most of the time I dont understand what you are speaking because stuff is too technical for my small mind. I request you to make a small summary or a follow up video of 5 mins eg what settings should we dial in to save cpu from degradation without much impact on performance.
Great video as always BZ. I did all the settings you have shown but if I try and raise my CPU above 5.4ghz it will start to hit the 253w PL1/PL2 limits. Do you know how I can further tune my 14700k to stay at 1.35ish volts and at or less than 253w? I have a score on my Aorus Master X of 94 so my cpu should be ok as I have not tried to run it past 5.2p cores and 4.0e Cores due to intel's issues.
Hey man great video, i have the B760 and i updated it to the new microcode 0x129, intel default settings is not in the bios before or after i updated the bios. And the CEP setting is the same as u have Auto = disabled. I don't have the "offset" options in my motherboard so im stuck with the dynamic Vcore option. Im getting better score on Cinebench 24 when i apply a negative value in the dynamic Vcore. That means the CEP is actually disabled, right?
B series can't disable CEP, it's forced enabled at the hardware level. I assume your CPU is a lower tier than the 13600K which is the minimum they have programmed for intel default settings
My MSI Z790 ACE has a Renasas controller, and uses the same multipliers for LLC and AC/DC LL as your Gigabyte mobo. Do you think motherboards from different manufacturers with that controller should have the same multipliers?
So I got a terrible 14900k, I can run only the following settings without triggering CEP (even out of the box triggers CEP) on an Asus z790 board. LLC = 3 AC = 0.66 DC = 1.1 On top of that, I slapped a -0.100mv IA Voltage Offset into it, and it's stable on C23, Prime95 and Hogwarts Legacy Shader compilation. At this point I dont even know anymore whats correct or not, but as said, this is the only way for me to run it without hitting CEP (or using the worst ASUS profile, that puts AC = 1).
Great video honestly. helped me tweak my 14700k with CEP back ON and no loss in performance at all. still getting 35-36k on CB23 and 2k+ on CB24. Thank you
@@Inasari2 I just used Intel performance profile on my z790 aorus elite AX board and went ahead and changed AC/DC loadlines to 55 and Loadline calibration is set to high. and put in a 1400 IA VR Voltage limit. and then I undervolted using adaptive voltage like I always have done but this time I put in -125mv vcore offset. I used to have it at -80mv vcore offset but at that time I had CEP off and LLC at low with AC LL at 40. I also went into advanced cpu and put in a TJmax limit of 85c and disabled AVX offset. that's about all I did in terms of cpu tuning.
You mentioned that you find undervolting via VF Point offsets hard to tune. But what if the same offset would be used for undervolting only VF Points 7 to 11 while leaving the lower ratios untouched? Do you see this as an improvement over applying the offset globally? I am a bit concerned about lowering the idle voltage and lower ratio points too much and am not sure how to test stability in that area. That is why I thought of only applying an identical negative offset to upper part of V/F curve.
15:01 wow that is even lower than my 13600K which scores around 20k, which is actually quite a bit lower than what a 13600K should perform. Maybe that's because of degradation or maybe it's because my Deepcool AG620 air-cooler is not enough for this chip.
@@BucifalulR around 24-25k is what I had seen on other people's, including reviewers, benchmarks. It could be that the whole debacle with degradation and the new Intel Default Settings and the latest 0x129 microcode have caused some performance regression, but from what I have seen around the internet, it shouldn't be as bad as 20k on Cinebench R23.
@@Dhruv-qw7jf I got 20802 only :(. If I apply Buildzoid's IA VR Voltage Limit of 1400 (Gigabyte B600M Gaming X DDR4 motherboard), I get even lower, like 19000. Maybe I should try with Intel Default Setting disabled, but doesn't that increase the risk of degradation?
Hello BZ, recently i updated to newest bios version 3401 and i lose too much performance in intel 12 gen cpu, specificly my intel 12400f, what u recomend me to do? i appreciate your job!
It you want to measure something like Die Sense with a 'scope DO NOT connect a piece of co-ax and 50R terminator! Instead just add direct very short wires to the relevant point, plus a ground as near as possible, and connect a 10MOhm X10 scope probe directly to those short wires. As a graduate EE what you are currently doing is painful to watch. Can I also suggest that this could have been literally a 5 minute video showing CORRECT settings for undervolting without triggering false CEP.
25,000 ish points is what I usually got with my i9-12900k lol Do you have a video covering how to safely undervolt/overclock the i9-13900k on the MSI z690a pro WiFi board you have? I just “upgraded” to the 13900k and now come to find out there’s all these issues with these cpus lmao I’m a video producer and editor so time is money for me, I need my system to be fast but reliable. I should probably move up to the z790 board but, the specs didn't seem very different, maybe I'm wrong. Any help? 🤙🏻😎❤️
"So if I understood correctly, the 'Vcore Voltage Mode' should be set to 'Adaptive Vcore', then set 'CPU Vcore Loadline Calibration' to 'high' and lower 'IA AC/DC Loadline' to 55, and then the 'Internal CPU Vcore Offset' should be adjusted from something like -0.080v to the -0.110v to tune the point where the CPU is stable?
se basicly 14k views and no person know anything about my ask :D ... and btw my r23 score with this settings is 39150 and max 1.385v and everything stable ... i tried almost every benchmark from BenchMate and im very sadisfried my temps cant go over 73 degrees but it is to god for me to feal ok :D so i need ask again can someone tell me is it ok what ive done it is described up.
Does Intel XTU no longer work for undervolting 13th and 14th gen CPUs? I have a 12900K and use Intel XTU to undervolt; nothing in BIOS; just default settings.
13900KF with Gigabyte Gaming X AX (1.x revision if it makes any difference), I set my loadlines to 15 and loadline calibration to extreme. with adaptive vcore and vid offset -0.085V. Both CB23 and CB20 run stable with 23 at ~37-38k and cb20 around 9k. peak VID requests in HWiNFO are are around 1.25V but can jump to 1.329 in idle. But, and here is where I stumped, this is with only 5.3 ghz on p-cores and 4.1 on E-cores. It seems I can't get 5.5/4.3 ghz in CB, unless I turn off intel default settings and go with whatever gigabyte uses, which crashes on most U4 shader building/cb20 etc. and another thing i noticed - with these settings if i open intel XTU monitoring there is something called "current/edp limit throttling" constantly being triggered, and no obvious correlation with say p-cores getting boosted to 5.8 ghz during these limit throttlings (or they just don't line up due to only having garbage ITE I/O monitoring). Did anyone encounter something like that? System is stable with current settings, it was stable even before intel default settings except for u4 shader compiling and cb20, and not even all U4 games were affected. But as I don't have a way to figure correct loadline settings except by just going lower than the default ones (90/90) until CB20 starts crashing and I don't have garbage scores (like 25-30k if set to 55/55 and high calibration) I am a little bit nervous if 15/extreme and -0.085v undervolt with adaptive vcore enabled but "current/egp limit throttling" getting triggered (maybe XTU overrides my settings as I see that with it running VID is higher), if it's fine or not.
the current limit is probably getting triggered by ICCMAX. Which happens way before the VRM outputs ICCMAX levels of current because ICCMAX is supposed to include the ripple current flowing through the decoupling capacitors and the VRM can't measure that.
@@ActuallyHardcoreOverclocking did take a look - in hwinfo there is a metric called IA: Electrical Design Point/Other (ICCmax,PL4,SVID,DDR RAPL) it gets triggered when I start CB23, but it's not constantly triggering on/off while idle, so probably intel XTU applies some garbage settings overriding (at least partially) what is set in BIOS. XTU already went into trash, but more interested will I have positive effect if I set ICCMAX manually to what Intel suggests in their documentation? (307A, if I remember correctly) VR VCC Current (SVID IOUT) shows maximum of 206A, but at the same time I have 253w PL1/PL2 default settings, so probably due to 253 max power ICCMAX adjust accordingly?
When I run your settings exactly from previous videos, I get crashes of apps with any kind of undervolt, but cinebench will run perfectly fine for 30 minutes with a 100mv undervolt
Because it's hitting the power limit and running at lower frequencies which won't necessarily be unstable. Something stable in Cinebench could crash in a game which runs closer to 5.7, or the Timespy Extreme CPU test (in a loop) for instance. Or in any number of other scenarios, which is why extended testing is crucial.
@@dd7428I play iracing, the funny thing is the actual game doesn’t crash running at 5.8ghz its the timing apps and discord that crash, the actual game runs without skipping a beat and while its doing this the cpu is pulling around 80w
@@ActuallyHardcoreOverclockingthe game runs fine even with a 100mv undervolt, but the other apps like discord and my timing apps which use a browser instance and an overlay that crashes with the undervolt, even the browser instance of simhub that my screen display uses crashes, I think to get it not to do this the undervolt is -0.020. Also I see you selecting the actual voltage, is there a reason you don’t just type in the offset you want ?
@@PanterlucaI usually only have 1 hand on the keyboard connected to the test bench because it's awkward to reach. So typing things that require non-number characters gets annoying(like decimal points). Also it sounds like you just lost the silicon lottery. I can't imagine that undervolting with CEP turned off would work better(it doesn't help my CPU).
I wish you would attempt these videos on an MSI z790 Board to see what you can and cant do when settings YOUR particular board has but others (such as MSI) doesnt have. Such as IA VR LIMITS we cannot set any because we havnt the ability to do so on MSI boards cant even sync LLCs without full manual tuning. Its terrible how different boards are. There should be a mandatory labeling across all manufacturers and the only difference in settings being between AMD and intel only. If youre on an intel platform the terminology and access to settings should be a standard across all boards. This stuff is ridiculous now.
check the guides on the MSI forum, they helped me forum-en.msi.com/index.php?threads/guide-how-to-set-good-power-limits-in-the-bios-and-reduce-the-cpu-power-draw.400270/ forum-en.msi.com/index.php?threads/different-undervolting-methods-with-ia-cep-enabled-and-how-they-compare-to-lite-load.400984/
Of course, not sure how much undervolt wiggle room you’d have though. Since they are better binned, their VIDs are already lower compared to most K’s and KF’s. Unless you got a really bottom of the barrel KS, I’m sure they exist.
I am on fixed 1.1v with "normal" load line. 4.4ggz locked on a small cooler. Max load at throttle temps 125w. Im wondering if this undervolt method could result in more performance.
How would you decide on optimal AC loadline settings on a platform that doesn't have LLC or a vcore sensor i.e. an Intel HX laptop? Having a lot of fun trying to fine tune my MSI Vector 16 HX with 13980HX (I.E. a 13900k in a laptop) but I don't know how to the optimal ac loadline setting other than to first set a voltage offeet on the edge of stability and then adjusting AC offset and seeing whether heavy loads or light loads cause the first crash. Ideally a properly set AC loadline will allow maximal offset at low loads without crashing either low or high loads, correct? Ironically, my MSI laptop bios DOES have an IA VR limit and I happily run all core 5.6ghz on light loads with a 1.3v limit.
Hey BZ, I think one main reason a lot of your videos are misunderstood is that they they are so long and a lot just gets lost in the sauce lol. If you made a clear summary for people who are new to tweaking, non-native english speakers, or who may not even have the aptitude for this stuff, I bet many more people would understand.
Edit to say: Source... I have experience in teaching. :P
yeah , he dosn't use any chart or tables or a simple notepad that are good for non-native English speakers
videos are too long to understand and there is too much testing this and that . it's like streaming type of content .
dude it's youtube , just give me the juice first then we can sit and watch rest of the video talking about ac/dc/bbc .
his voice and delivery is so good that i can site there and listen for hours
in this video he get started 37k with intel default setting . our goal is to undevolt cpu with CEP turn on . we end up with 39k and we didn't saw how it is with temp or cores frequency or voltage . idk what i just watch tbh .
we already know that you can undervolt a little with CEP on but it doesn't let you get a good temp or low voltage . that is why peoples turn it off .
Not that I completely disagree but I'd say his content is probably not aimed for that target audience.
It is understandable and testing real-time all the possibilities is a great teaching method but:
- you have to do your homework (read/watch other source material - Intel spec, for instance)
- you have to pay attention, like really pay attention and try to anticipate the outcome of the next 'text'
Then he explains why the outcome was this or that on each scenario.
Really I find it to be great content in this format. Not found elsewhere.
don't you dare criticize the ramblings of our lord and savior
I agree with @@DeliriumT - if the goal is to learn, like to gain understanding to the point where you can adapt it to different situations or even come up with some of this yourself, this format is great. At least, short of giving people exercises that make them follow along. It feels a lot like how university courses have lab practical or assignment discussion periods.
But if the goal is more about the bottom line, like "job training", there are plenty of mainstream outlets that can digest things into a tutorial or instructions to follow.
Sorry I don't know the terms in didactics, but this doesn't seem far off from their best practices.
I watch them at x1.5-1.75 speed to not get too distracted while watching them... but overall I don't really see this changing, that seems to be just his style and it's unlikely to change... ever
Based on Intel documentation, CEP is the mechanism by which Fast-V Mode operates. It exists so CPUs can be safely used with low quality motherboards and power supplies that would crap themselves when trying to power a 14900K under heavy load.
I think people with higher-end motherboards and power supplies would be fine with CEP disabled, more safety features is never a bad thing.
In today’s world you need a PhD in electronic engineering to understand and safely operate a piece of silicon that doesn’t even work properly by default 😂
because the default is not 6GHz no matter how hard Intel tries to, it just can't be stable at those clocks
Another reason why you go with AMD, just plop it into the motherboard, update BIOS and enjoy your new PC... None of this rocket scientist crap is needed to just play some damned video games
@@lessdatesmoreonmyplates1457 I remember when that was being said the other way around when AMD CPUs were exploding. Always some corporate fan ready to pounce on an issue to say "Should had bought from this company that doesn't pay me to sell for them, but I do it anyway!". Maybe you do get paid, in which case, fair play. In many cases though, a lot of these people aren't getting paid for their brand tribalism, and it's just pathetic that they hang around their corporate biases competition videos, forums, subreddits just to say "This is why you buy AMD!". They only do it when the other side is down, when their own side is down, they try to lie to everyone that it's actually up. I've seen Intel fans do it too when earlier in the year, you could just turn on a Intel PC and it worked while AMD was having the issues.
yeah
@@lessdatesmoreonmyplates1457 Agreed. I built my PC only 6 months ago and I'm thinking of cutting my losses and getting a 7800X3D just so I don't have to deal with this BS anymore.
Thank you so much for making this and all the other videos as of late!! It has been a INVALUABLE source for me as a novice bios tinkerer. This kind of information and testing is so scarce, not even the board manufactures have a site you can go to that breaks down and explains how a mother board works. You have filled this gap of non information that really shouldn't exist.
Its so weird that there's not comprehensively direct information available when using someone else's product that is designed for you to tinker with it. How exactly are we suppose to learn this stuff on our own without studying to becoming an electrical engineer?!
I've been out of the overclocking game for a long time now - I'm still rocking a Core i7 2600K, and at 4.3 GHz on an air cooler! - so I'm completely lost with all the new configuration options. Maybe I'll get back into it whenever I get around to upgrading my desktop haha. But I'll still chill and listen to Buildzoid talk. Maybe I'll be able to piece together how some of the new silicon is configured.
Thank you so much for your efforts! Been devouring your videos lately cause I was concerned about my i9-14900k which I bought in May 2024.
I'd love to see you doing some AC\DC Loadline testing on motherboards with the OnSemi (NPC81530) controller, cause I have a Gigabyte Z790 UD which uses it.
Keep up the great work :)
For stressing arbitrary combinations of test program + frequency + power, you can limit peak clock frequency with the CPU frequency scaling controls, then set a CPU quota (cgroups in Linux, Battle Programmer Shriase in Windows) to PWM the test program such that it doesn't hit power or thermal limits.
Another thing I see a lot of people incorrectly suggesting has to be turned off in order to undervolt, is "Undervolt Protection". All that setting does is establish a floor for the voltage at POST, it doesn't prevent you from setting a lower voltage at boot in the BIOS
No, you do not only mostly affect high load voltages with AC LL, run Geekbench 6 and compare VID (or vcore, whatever) on light workloads, you'll get significant difference in VID between 1.1 and 0.4 mOhm. You can already see the influence just in the bios, there's tons of voltage pumped regardless of load, since AC LL is a flat offset calculated from AC and probably ICCMAX. The same value is being added to VID all the time. Adaptive 1.435V + 1.1 mOhm LLC + 0.65 mOhm AC LL = up to 1.48V in GB6, 1.435V + 0.06 mOhm LLC + 0.1 mOhm AC LL = 1.41V in GB6.
Also, MSI gives the per ratio voltage offset, but it's completely broken too, so while bIntel gives that opportunity - motherboard manufacturer has to be not idiotic too.
Also2 - at the time of writing this comment I didn't get to 26-ish minute yet, where you've said just what I wrote :P. Which is also contradicting what you said at the 12 minute mark, it's influence is pretty much linear, although by dropping just AC LL you actually hit high load voltage stability limit way sooner, than low load voltage stability.
and yet he says "for some reason people don't understand what I say" - he needs a summary comment pinned at the top, even if it loses some views.
I have enough another issue, if I keep the 1.0/1.0 AC/DC (recommended LLC4 setting with latest ASUS BIOS, then try to undervolt with the SVID (requires a ton of negative offset to get the same voltage as before), it ends up tripping up the 307A ICCMAX on my 13700K. Meanwhile if I use my normal undervolt mode of 0.4/1.1 with LLC3 with SVID offset of -0.05 but with CEP disabled, it doesn't trip the ICCMAX and downclock.
The only way I can somewhat get CEP to work is use LLC5 with 0.4/0.73 AC/DC with SVID offset of -0.11 with CEP enabled. I still end up with about a 200 point lower C23 score, I even tried creeping up the AC LL to .5 and .6 with more SVID offset to compensate, still noticed the loss of roughly 200 points if CEP was enabled vs disabled. I would gain those 200 points back if I left it locked at 0.73/0.73 on AC/DC with roughly a -0.19 offset, but the voltage under load wasn't constant, it drooped heavily in the beginning and raised up over 0.02v by the end of the test. With the other settings, it stayed constant the entire test.
@@ZoneXV I'd expect it tripping PL4 actually.
@@PolskiJaszczomb Could be that too, it's under IA: Electrical Design Point/Other and PL4 is listed there too. It wasn't maxing out PL1 or PL2 though. I'm convinced the stock out of the box settings is trying to kill my CPU. Gave it vcore's in the mid 1.4's and tried to run cinebench23 at 1.32v before immediately throttling. It's trying to run it at full load at the exact VID listed in the VID table with no vdroop at all! I'm not sure if this is on Intel or this is just an ASUS screw up. Really considering going back to my old BIOS I was on with just a voltage limit since it hasn't given me any issues for 2 1/2 years.
I did his settings with the LL to high/55/55/-100v offset, and my 13900k went from running above 1.5v to running up to 1.35v max for what seems to be the same if not better performance. So idk, he says you don't need to do it, but it seems like it really helps. So I'm even more confused now. My clocks are the same, if not better since I seem to hit 5.8Ghz more often, but maybe that's placebo. Temps are waaaay better, 20 C difference on cores that were hitting 70 C on startup alone. Now my highest core is currently 53 C. Also I'm running a Gigabyte Z790 Aorus Elite AX gen 1.
We really lucked out with Buildzoid having a Gigabyte board.
Yeah, almost everyone that make this type of videos uses either asus or msi
I know, right?
Gigabyte has the most hard to use bios that owner can proud of their tuning ability.
@@yuan.pingchen3056 asrock would like a word
I tell myself this all the time
I think one of the big takeaways from this whole debacle is that Intel and board partner documentation is an absolute joke. They are shipping products and features with no clear or comprehensive explanation for how they work and leaving it up to the consumer to figure it out through trial and error. It's actually insane that we have to rely on what are the equivalent of unverified community forum posts for information on how the commercial products of these multi-billion dollar corporations are supposed to be operated instead of having a detailed manual we can refer to.
and it will stay that way because the percentage of people who even bother do this are like a rounding error, for worse and not for better
Thanks Buildzoid!!! Running these settings on same board and 14900KS with 8000mt RAM!!!!
I agree the video is a tad long, but I need to offer confirmation that your recommendations are spot-on. I have the Z690 AORUS Elite AX with F30 BIOS (0x12B Microcode) and a brand new i9-14900K (never subjected to faulty default voltage settings, BTW). With your settings: Adaptive Vcore, Legacy mode and -100mv offset, Loadline Calibration = High, VR Voltage Limit=1400, and IA AD/DC Loadline=55, I consistently get CBR23 score of 39500-39700. And highest vcore never exceeds 1.34v under all core load. And I'm still seeing 2 Perf-cores hit 6 GHz periodically and multi CPU ALL-core loads at 5.7 GHz on the 8 Pcores. I suppose I could tweak at the margins, but I really see no need to as these are rock stable under gaming loads, and CPU voltages drop to .77-.78 v under idle or low loads. Thanks for your explanation of this most-confusing topic.
for my system; 13600kf>msi z790 gaming pro wifi, setting AC/DC loadline at 100, llc8 and setting cpu vcore to adaptive+offset with a negative offset seems to be giving the best results. i dont have fancy tools to test exact voltages, but in hwinfo and cpu-z my vcore and vid are the same (+/- 3mv) and i havent had any instability issues. excited for you to test the msi bios and hear your thoughts. thanks for the great content!
Honestly, i dont see the point in using LLC8, for me anything above 4 is just unnecessary heat
i tried two ways:
1. using LLC6 with undervolting (no CEP)
just surfing an internet browser causes the CPU to go all the way up to 80+ degrees
2. using LLC3 no undervolting (no CEP)
doing same silly test, will max at 65+- degrees
more LLC = more heat = less performance in real life things
sure there is still the silicon lottery, but for most CPUs, i would stick to default settings with the minimal LLC that is still stable (LLC3 in my case)
also i would suggest using ac and dc load line to be exactly same as your VRM load line, 1.1 for Asus
@@CASIOMAN20 For MSI, the higher the level, the higher the Vdroop. So level 8 (which is almost the same as Auto, but only for Auto voltage (resp Auto voltage +Offset)) would give the lowest Vcore and temps.
If you set LLC to "auto" (1.1 mOhms), don't set an high enough Voltage Offset and then limit the VRM to 1.4 V, you will likely see lower boosting P-Cores. Just because it needs a higher VID and if that doesn't fit under the 1.4 V ceiling, the next lower boost stage is used that fits beneath that ceiling. I've tested this on an Gigabyte Board lately. Lower LLCs (and matched AC and DC Loadlines) obviously can fix that - depends on the quality of the CPU.
but then 23:45 "dont touch the loadline settings", and more confusion
@@notexactlyrocketscience and yet he wonders why people don't get what he says - he needs less fly-by-seat videos and to actually plan it out and less of the "well.... [ramble on some disclaimer that doesn't really matter to the main point]" moments.
@@xBINARYGODx i actually wrote down my own personal, cleaned up transcript and i'm going through it with a highlighter
Legacy offset is entirely ignored on a 13600K AND 13600KF with the same motherboard and settings in my experience. Even tried CEP disabled. DVID has been the only way for me to undervolt on a 13600K/F on both the Z690 Aorus Master and the Z790 Aorus Master X.
Oddly though, HWInfo detects the offset correctly when it is applied, but the VID and VRVOUT do not change.
I am pretty sure there are weird quirks going on here from motherboard to motherboard. Or same board, different CPU. I've experienced some weird stuff as well. It's supposed to work in a certain way but it really doesn't.
have you tried disabling intel default settings?
Did you by any chance switch the profile from Intel Base Line to something else? I found by doing so you introduce a lot of weird stuff. Also there is a feature called undervolt protection which I did not have to Disable for my i9 but for an i3 I owned I had to with led to a bunch of problems!
TLDR; What are you current BIOS settings?
@@interceptor001 Yes, I've tried with CEP on and off, yes I've turned intel's default settings off. I've mimicked BZ's exact settings and it's not respected. I've tried on default and with custom settings. I believe it is a 13600K issue. I feel the need to reiterate that this behavior was present 2 years ago on the Z690 Master with a launch batch 13600KF, and is still present.
@@Zhunter5000 I am not talking about CEP I am talking about the Undervolt Protection
what does xtu use? I thought it worked in there with undervolting, but not sure. And could you do a video on a asus z790 bios? It's so different than others, I don't know where I find things, because they have other names, are other places and things like that. I just don't know how to apply your showed settings 😅
I'm tired of all this.
I installed the new microcode. You recommend enabling CEP, does disabling it remove the 1.55v hard limit for VID voltage included in the new microcode? In the meantime I disabled CEP and set an IA VR limit of 1430mV (below that number my cpu clocks decrease) in addition to the hard limit which I hope is active...
(I have a 13700k, an Asus Z790 mobo)
Vai aumentando o LLC que resolve
i wish i could have seen the temps with your tests, nice video AHO
I was able to drop voltages by -0.175 on my 14700kf with CEP enabled and using your loadline settings. Also I was able to increase clocks by 100 MHz, so they are 5.6 GHz all core, 5.7 GHz 1-2 cores. In HWINFO, the VID doesn't go above 1.199 V and VR VOUT 1.190 V. R23 ~37000 with 253w power limit. Also put 1.4 V limit just in case. Have to use my PC more to be 100% sure about stability but seems to be doing good.
loadline high 0.55 ac dc?
@@mexicano-claudiu6802 yup
@@mexicano-claudiu6802 or 45
you are missing an R in undeRvolt in the title, as always great video, whoever doesn't appreciate this type of content or the duration can go for eye candy content with 0 value in plenty other channels and thats fine
MSI seems to have this on lock. No brainer with their boards. Lock the cores at 5.6 or 5.7 and set CPU Lite Load to mode 7 - 12. Lower is better as long as it doesn't crash. Too low and it will crash just from the CPU not getting enough voltage. You will see performance loss as you work your way down from Mode 12 once you get to low. When you hit that performance loss go back up one mode and enjoy your cooler, more power efficient, fast @$$ CPU! Don't let 2 cores boost to 6Ghz it over volts ever single core. Even the E Cores.
The thing is that MSI Lite Load is exactly what BZ explains as the second wrong way to undervolt. It lowers AC_LL and DC_LL. You can see what it does if you set your "normal" LiteLoad, safe & reboot and then change LiteLoad mode to "advanced". You can honor CEP with the advanced mode, if you bring AC_LL, DC_LL and LLC together manually. You can find a table online that shows you which LLC equals to which AC_LL/DC_LL.
@@bauerns5er MSI's current solution to that is to set the Lite Load (and so AC/DC) stupid high under Intel defaults. So you can drop it a long way from its default value and still not hit the CEP tolerance limit.
🤦
@@Raxiel497 Intel Default is 1.1 mOhm on my Z790 MSI for AC/DC LL (110 in MSI terms). That's not really stupid high, it's the default. 😉
hour long videos for something you can do in 10 minutes on an MSI board with Lite Load. Lock P cores to 5.5. Lock E cores to 4.1. Set a negative 0.055 voltage offset. -2 AVX ratio. BCLK 100mhz lock enabled. Speed Shift enabled. Intel Turbo Boost Max 3 disabled. Enhanced Turbo disabled. CPU Current Limit 307a. Power Limits 250w. GT Current Limit 30. CPU Lite Load 9, or 10 if you want to be conservative. CEP will probably never activate anyway so you might as well leave it on. This has got me between 38k & 39k in Cinebench R23 runs, voltage will be below 1.4v (I've done runs where the max was 1.309v) and you won't go near throttling with decent cooling and a contact frame (360mm aio, max core temp 93c on my hot core #5, rest below 90c, e cores below 80c), and CPU power doesn't go above 230w. Using Z790 Carbon Wifi board, 2x32gb ram and 14900k with the new bios installed.
CEP was introduced due to silent corruption since 12th Gen I assume. I udervolted a few builds without CEP and they eventually corrupted the windows installation. This has never happened with 11th Gen or older, they would just freeze if voltage wasn't enough.
Update: The root cause was SA CEP, ALWAYS ENABLE SA CEP even if SA voltage is left on auto
On my gigabyte board, intel default settings force something called "Dynamic Overclocking Undervolt Protection" which you can see on hwinfo. You can't undervolt either way when this is on, and it doesn't go away unless you disable intel default settings (even setting IA CEP and Undervolt Protection to "disabled" will not work)
same here
Model ?
@@jst8922 Gigabyte Z790 GAMING X AX (rev. 1.x)
@@jst8922 i think my reply got auto deleted, i have z790 gaming x ax
Sure enough, I have undervolt protection enabled and I can undervolt perfectly fine on my MSI. Looks like motherboards are allover the place LOL!
33:22
nvidia 375.70 (3rd row from top to bottom, right side) is from 2016.10.28.
nvidia 474.64 (1st row, 3rd column from right) is from 2023.10.31.
NGL, i really don't like Nvidia driver naming scheme. These numbers are not telling.
For example AMD currently uses this scheme for public releases.
YY.M.V (aka 24.7.1 would mean first driver in July 2024.)
Professional usage drivers have naming YY.QX (aka 24.Q3 would mean driver from third quarter of 2024).
And test branch drivers have this naming scheme
XX.YY.ZZ XX is usually corellates with year, but can lag behind significantly. (Aka 24.20.08, meaning major version 24, branch 20, version 8, or something like that).
@@DimkaTsv who cares - number higher is all that matters - but really, for most its "I was told my software X I have new drivers"
Love your detailed content as usual good sir, but i would like to point out that different AIB's are behaving very differently when it comes to CEP & AC_LL including undervolting. Even more so with the x129 Microcode update, Asus for example, at least in my case fails to work whenever you change anything manually regarding AC_LL or even a mild VID alteration of 0.025Mv. Although SVID seems to be alterable, without issue. Very odd...
It looks like they've locked it down with CEP enabled (at least from my perspective) when running Intel Baseline on this BETA Bios. Asus Maximus Z790 Formula btw.
I'm not a fan of keeping intel default profiles enabled. The stuff they lock out from motherboard to motherboard is just so random, it's annoying to deal with. Not everything translates well when trying to help others. Just go full manual mode, disable it. I've had great results with Turbo LLC and AC LL as low as possible, small offset on top. Not every CPU likes -0.1V, not even -0.05V unfortunately.
Please yes! Do the Z790 Godlike! I have a Z790 Ace Max which is very close.
Setting IA VR config is not available on MSI boards, a rep in their forums straight up said they wont enable it.
Just all core overclock your CPU then, or put a fixed max on Pcore and Ecore. Like 5700mhz and 4200mhz. Then with undervolting, the VID requests will never be too high.
just use lite load and do it that way.
@@bigpoppa1234 No. Lite load only adjusts the AC/DC Loadline.
@@Infi1337 Check the second page. 😐
Lite load is the way to go for MSI
I have a completely different problem with my new build and I'm pretty much at my wit's end...I have a brand new i9-13900f that has never been used on anything other than the intel default performance settings (i don't have extreme because I'm assuming the i9-13900f doesn't support it) + 0x129 bios. For some reason any time I run a stress test to push the CPU to full load, it only achieves full load with 219w power draw (default pl2 for this chip) for a couple seconds, then throttles itself severely back to 65w (pl1) where all the P cores drop from 5.3 ghz to 2.2-2.4 ghz.
1. I have not touched anything else in the bios other than change to intel default settings, and turn on XMP1 for my 6000 mhz memory.
2. I know I'm not thermal throttling on the CPU at least, because I can see that my temperatures only ever reached maximum 86-87C before throttling. I'm using a Ryujin III 240mm AIO.
3. I have the Asus Tuf gaming B760-PLUS WiFi board, which I know isn't a Z series board but considering the i9-13900f isn't as demanding as the other skus (even less demanding than an i7-13700k), I can't imagine the board's VRM's being the issue? The temperatures seem fine to me. I have an 850w psu, so power is not an issue.
I'm starting to suspect that it's the intel default settings that is the issue, since I don't have any thermal issues. Now I'm trying to figure out how to fix this issue while staying within intel's specs. With all the buzz about the stability issues, I'm really concerned about changing too many things within the bios that may potentially harm the CPU.
Has any one else experienced this sort of behavior with the intel default profiles?
Maybe try undervolting?
Is TAU period still a thing?
Huh, that seems strange. Have you checked all temperatures - including VRM, Ram, chipsets? Cause from what you say your CPU should stay at the boost longer and 2.2GHz seems too low for 65W even. it should be more along the lines of 2.7-3GHz.
Does the board have any stupid "stability enhancements" or other things like that?
@@yarost12 tau should be 56 seconds.
sounds like the intel default settings are enforcing the 65W power limit for the 13900F. As for how to raise the power limit. You might have to disable intel default settings to do that. I unfortunately don't have any high core count non-K CPUs to test with.
@@ActuallyHardcoreOverclocking but even with the power-limit the CPU should not run that low. The few 13900 i have seen - none of them were under 2.5. He'd have to have gotten a real dud there.
thanks for the guide
using a i9 14900KF on a aorus z790 elite AX ice with an AIO 360 watercooler i got a 35022 cb score
after changing the vcore offset to -100 i got a 35795 cb score
setting the vcore loadline calibration to high put my cb score to 36472
still not nearly as high as yours but im happy it improved, was not able to change the AC/DC loadlines as it was unclickable with the latest bios
Hoi BZ any 9000 series coverage? easy 8400 settings on 9700X perhaps :D
I want that too
IO die is the same as in ryzen 7000, so same 6000 MT sweet spot and no "easy 8400" possible sadly
don't have one yet.
So, did the new microcode actually do anything besides set limits I could already set in the bios? I'm still waiting for a non-beta release for my board (z790 Aorus elite 1.0 rev)
I updated my board with z790 bios flash, I’m a newb . My pc now runs much harder it throttles all the time like it needs wayyyy more power from the fans . Not sure what’s going on.
I have i7 13700k and with CEP disabled my VID's never go above 1,360V, but with CEP enabled, my VID's jump to 1,440V. My current is limited to 307A and PL1 is set to 125W and PL2 to 253W.
Same here. I have an MSI mobo with a 14700k inside and I have the same results as you. IA CEP disabled, CPU current limit 307A, PL1/PL2 at 253W. Also I changed the value for CPU Lite Load to Mode 7. I never go above 1.36v and my temps are in the high 60s and low 70s during max load in Cinebench 2024.
this will be true on Gigaboards too - a lot of what he shows is very specific to the specific settings he uses - and mostly no one is leaving the things on Auto that he does for his "dont do this" video - but who knows, many lots of noobs/relative-noobs in the viewership.
Hello, I don't really have much knowledge with the bios, but I have an i7 13700k and a gigabyte aorus elite ax ddr5 board, currently my cpu reaches 5.4ghz with a maximum recorded peak voltage of 1.378 on very few occasions but in general the peak maximum is 1.362, I have an lp1 and an lp2. Limited to 200w and maximum temperature of 80 degrees, latest bios for 2023, I don't know if I'm okay with the degradation, or what parameters do you recommend?
@@elvinsontaveras7125 Just for the sake of reference my 13700k is stable running 100mV less than that.
@@dizzyfist7 legit same. tried new microcode today too w intels bullshit bios settings today and i was hittin 1.44-1.46v. horrible. going back to OC undervolt on my 14700k.. fuck intel.
MSI doesn't have the voltage limit setting, someone asked on the forum and apparently they've made a decision internally not to offer it. In terms of the CPU core voltage modes, they have Auto, Adaptive mode, Override mode, offset mode, adaptive + offset mode, advanced offset mode (VF point), adaptive + advanced offset mode (vf point), adaptive + offset + advanced offset mode (vf point). I'm not sure what they all mean exactly, but Adaptive + Offset Mode allowed me to lower voltages further without triggering CEP, when I was already on the edge of triggering it with my load line settings.
Same here. I matched my DC LL and LLC, set AC LL TO 15, adaptive + offset, auto and -100 mv. My CPU at 55/43/45 hits 40,800+ on CB23 and maintains 55 on full stress tests. And my VR out and vcore are almost identical
And…I may hit low 90’s on a 30 min OCCT run. I usually maintain upper 80’s
Advanced Offset Mode (VF Point) also works well on MSI, just tested it. And then you can add your undervolt only to VF points 6-11, for example. Or whatever matches your top ratios.
@@kpm4620which llc mode are you using and what value for dc?
@@janos1312 I RMAd my 13900k and got a 14900k. Here’s my current settings on my MSI board:
LLC 6 (50 mohms), ACLL and DCLL 50. Adaptive-auto, -160 mv offset, PL 1/2 260w, 400A. I leave clocks stock. This thing runs great
I just love the fact that he keeps pressing delete even while the pc is shutting down 😂😂, Just holding down delete works greatly (and accurately)
IDK if it ever happened to you, but it used to be that holding down delete key didn't work. And some settings in startup made entering BIOS window like 0.2 seconds. Ever since I had to deal with mobo like this I do exactly the same thing - smash that delete key as often as possible.
@@jannegrey I think that was the case with lower end boards, I had a b450 that didn't do it, my z790 does tho, I thought that was the case
Thanks for educating
@@InfinitycgIN From my personal experience (I had PC's from 1997 and I OC'ed since 2001) most boards had it. It's plausible that boards from the last 5-10 years don't do it as much, but I wouldn't know since I always smash that delete key. It's basically a tick now ;)
Depending on the "provider of BIOS/Firmware" (Im not exactly sure now, how this thing was called) - there was AMI and something "star" and couple others, they basically had an overlay that displayed memory in system and all the "important numbers". And often when this "overlay" was engaged you couldn't get into BIOS. You could get into it either before that (when the screen was black) and/or sometimes after it. But CPU's got so much faster that what was initially let's say 3 second window of opportunity, with upgrade could go easily sub second. To turn this off (if there was even an option) - you had to get into BIOS.
Something similar can be observed on relatively modern Laptops. They often have overlays and sometimes you either have to click delete key in very precise moment, or you have to use different combination.
I'm glad that it isn't now a case that you need to smash that delete key, but it seems to be like inverse of what is happening to Windows and going into safe mode etc. You have to know specific combinations, everything is hidden and stuff like that.
Mind you - I'm unsure if the signal from keyboard wouldn't be disconnected (at least on older motherboards) the moment PC turned off for a moment, when you restarted. In such case, you'd need to pull your finger of it and push it again. That might have differed depending on the connector (USB or PS/2). The point is that if you re-booted your PC 5 times in a row and always managed to skip that window of opportunity, you would be smashing that delete key as well ;) And mind you - even then you didn't always manage to get into BIOS (because Window of Opportunity was shorter than keystroke and sometimes keyboard was disengaged prior to this - pain in the ass for troubleshooting).
And thank you for teaching me that this is (mostly) no longer the issue. I did notice that it was starting to get better around....2005? But you still got some setups that were..... difficult.
I've posted up a couple times how he can create a simple desktop shortcut that will immediately shutdown the machine and boot into the BIOS. Guess keyboard machine is his thing.
I remember old mobos would error or ignore a stuck key like that. (Might have been PS/2 era?)
I don't understand something about AC = DC = VRM LL paradigm. If do it this way you basically have almost no v-droop, correct? Because even if you set your VRM loadline very droopy, like 0.9 mohm, and then "tell" the cpu about that vdroop (AC LL) it will just "ask" for more to compensate. Or does it actually makes a difference if we use let's say 20/20, compared to 90/90? Isn't actually something like 50 mohm of diffrence between vrm load line and AC LL considered ideal amount of vdroop or is it different for 13/14th gen?
He doesn't know how any of this works. You're asking the wrong person.
The function is that if you want to use CEP still, then you have to set the AC/DC to roughly the same impedance the the loadline calibration you have chosen. The best results often come from somewhere in the middle. Medium/High options. You need to know the impedance of the VR though. It is very often different between manufactures. For example my MSI is LLC 6 AC/DC 40 LLC7 AC/DC 69 where as Gigabyte (such as this one and the gaming x etc) are LLC medium AC/DC66 and LLC high AC/DC 55. It allows the caluclations to fall within the tollerance so that when you set the VID lower it isn't seen as an excursion by CEP, so you will get your frequency at that VID instead of it clock stretching or crashing.
@@theboomfish Where did you get this numbers from, they seem pretty precise?
@@sp00n hours of research. Found threads where someone had listed the numbers then made my own notes for what I wanted. I didn’t copy them all for every setting unfortunately. They are out there though.
@@sp00n I can confirm the numbers for MSI are correct. You can find out on your own. When your vid = vcore under multicore load it means that you matched your DC Loadline to vrm loadline.
Thanks so much for this content. It's really helpful for people trying to understand what their mobos do. Is it possible for you to make a video with a MSI motherboard?
I get 5-10 degrees hotter and higher voltages with intel defaults no matter what i do on gigabyte z690 gaming x 13700k... which I suspect negates the entire benefit of intel defaults to begin with. So I opted to turn them off. Not that I had any stability problems in the past year other than too aggressive undervolting.
I'm also having higher temperatures with the same processor and motherboard. The problem is that when playing everything becomes unstable. Did you find that removing the default Intel profile solved it?
Which specific motherboard are you using, and when you said the Elite has a different voltage regulator, did you mean the Aorus Elite X Wifi7 model has a different VR?
No I meant the non-X elite
@@ActuallyHardcoreOverclockingI love how motherboard and other PC component companies make their product naming incomprehensible. Last time I spec'd out a build for a friend, I had to create multiple spreadsheets to keep track of the confusingly similarly named parts. Slap an Elite, Pro, Super, Ti, or a random tech suffix, and watch as consumers with finite time and energy capacity spend too much on parts that don't do what they think they do.
Wait what ?? Did not understand anything you said in that video. lol can you direct me in the right direction to find basics on how to under-volt intel core I5 14600k im a noob and feel weird changing setting when i dont fully understand them. Thanks in advance
Benchmark my i5-14600k. I limt pcore to 35x, and ecore to 26x.
With CEP enabled I get Cinebench score 350.
With CEP disabled I get Cinebench score 826.
Apart from the CPU-frequency, Im running Intel Default settings.
I have an Asus z790 motherboard
I also have an Asus z790 board and saw similar behavior where the multicore score goes from 24k to 11k with CEP enabled. Something about Asus' power configuration seems to trigger CEP pretty much automatically.
I think that means your tripping CEP. For example my mobo has the same bios as the one in the video where high llc matches with 55 ac_ll/dc_ll, and medium llc matches with 66 ac_ll/dc_ll. When i tried to reduce ac_ll/dc_ll even by 1, with 0v vcore offset, i.e. ac_ll/dc_ll = 65 with medium llc, my r23 scores dropped hard with cep enabled, and putting it back to 66 restored the score. Also noticed idle vcore was high when doing this with 0v adaptive offset, so i still needed an undervolt to bring that down. Also, i noticed in my bios that the description for CEP states that e-core performance is affected by enabling CEP so that could be another reason for your lower score.
Most of this doesn't really apply to me, but is entertaining to watch. My cheap-a$$ H610M has almost no voltage controls, so DVID or loadline trickery is necessary for
12100F doesn't have CEP anyways
Luckily, the MSI board I have just has a dropdown that says by PWM or by CPU for the offset mode. And then you can just set your offset from there.
Offtopic but, for the 7000X3D series o/c or whatnot, after all the bios updates with latest AGESA etc., any need for an update video on what should be turned on/off? Maybe some things have improved or changed?
Asus B660-G: can't set negative VID offsets, unless using uCode 104. No undervolting on B660?
The Mainboard sets 0.4/1.0 llc4 in synced mode - not going to dig into that even further. 1.4 Imc results in 1.55 CPU vddq - Asus is nuts
same with my Asus TUF Gaming B760m, it has to go back to microcode 0x104 in order to do any sort of undervolting.
Same with Gigabyte B760M Aorus Elite (1.x).
my undervolt on my 14600kf cpu on an asus rog strix B660-G motherboard (bios 3601 microcode 0x12B) in cr23 i get 2091 single core/ 25180 multicore. hwinfo e-core VID min 1.141v/max 1.279v/avg 1.22v. temperatures min 31º/max 68º. CPU POWER min 5w/max 122W. Vcore max 1.146v. (all this stressing the CPU)
MOBO BIOS
-asus advance oc profile: enabled
-intel adaptative boost technology: disabled
-asus performance enhancement 3.0: disabled
-C-state: auto (if I enable it my pc crashes, it doesnt even enter windows)
-current cpu core/cache current limit: 225 (default is 200 but it made me clock stretch)
-PL1 & PL2: 181
-LLC: 5
-synch acdc loadline with VRM loadline: enabled
-IA AC: 0.40
-IA DC: 0.72
-IA CEP: disabled (when enabled it works but if I put xmp profile it destroys my performance)
-VRM core voltage: offset - 0.135 (I can lower it more but its a bit unstable)
I was able to adjust the settings for my ASUS ROG STRIX Z790-F GAMING WIFI MB and increased from sub 37,000 to over 39,000 on Cinebench R23 with my 14900k. In the BIOS under the Ai Tweaker tab there is the Global Core SVID Voltage setting which I changed from Auto to Adaptive, then I changed the Offset Mode Sign to -, and finally set the Offset Voltage to 0.10000. This is also after installing the microcode (0x129).
keep CEP on?
@@AdventureStreamingGuysyes
Also, realized those weren’t my CPU temps but rather my GPU.
Can you make a video explaining what IA AC Loadline and IA DC Loadline settings actually do? I've watched a bunch of your recent videos that touch on them but I still don't get it. I have a 13900KF that bluescreens and I can't RMA, so I'm trying to get it to be stable. For some reason, setting DC Loadline to 1 and keeping AC on auto with the default LLC stops my CPU from crashing. I've even tried undervolting with adaptive mode and a -70mv offset and it still doesn't crash as long as DC loadline is set to 1. I've capped the voltage to 1.4v, like you mentioned in one of your videos. I'm not sure if running these settings is OK or not because I don't really get what these settings do. There doesn't seem to be a performance impact with these settings and my CPU is running under 80c.
I've tried some of the settings you had in another video with CEP on and increasing LLC but that didn't get me stable for whatever reason.
Unrelated but my CPU only seems to crash in games. Running benchmarks is stable. I'm currently using Ready or Not to test if the CPU is stable or not since that game seems to cause it to crash most consistently for whatever reason.
I would really appreciate a guide for MSI boards
msi desktop boards dont have IA VR voltage limit and looks like they don't planning to adding that so rip with that
Can't wait for an Asrock guide for the coming weeks.
I'm so happy i have a gigabyte mobo (so i understand what bz is talking about) and i also undervolted with adaptive mode global offset and it works very well.
The only problem is that i can't change the ac loadline limit which says it's set to 0 (?!) but seems like it's working well with extreme llc option.
Edit: i forgot to mention that i haven't updated to the new microcode but CEP off otherwise is not 100% stable. Still works very well in terms that the performance is good with no stability issues and even with the 1400 limit it doesn't overshoot much, max 1.452 which is not that high given that i use 5.8ghz all cores with 185 power limit.
So basically i found a way to overclock, undervolt and keep temperatures in check where the performance is very good and the voltages are finally "normal" for what we're used to from intel. Btw, this on a 14700k which seems more impressive to me given that is a garbage bin just like bz's 14900k bin.
type type in a number, it'll work that way
@@notexactlyrocketscience right, i've done this before with other settings idk why i didn't think about it, i'll try. thanks for the suggestion.
@@notexactlyrocketscience After some hours of testing i found out that the sweetspot for extreme LLC is AC voltage limit set to 30 or 35 with CEP off. With CEP on you shave off some performance and the voltage shoots a little bit more. Anyway i'm very happy with my config right now, seems that dropping the AC limit from 55 (which was default on my mobo), to 30, keeps the core voltage from overshooting as much (max 1.402 from 1.452 previously), so it seems to respect the 1400 limit that i imposed, better. Very happy about that given that the performance is excellent.
I also tested with Ultra Extreme LLC and find out that AC limit works at 20 or 25 best, but the performance isn't better (it's actually worse somehow) and it doesn't seem tp be 100% stable (for example i fail the low thread cpu profile tests in 3DMark), so i decided to stick with Extreme profile and 30 limit.
Thank you Dr Zoidberg
Read by Kermit.....
can't unhear that now
So, just RMA'ing because the microcode came out. My 14900k is horribly damaged just like yours, it started having issues and crashes just after 3 days of usage, but moving on with a brand new chip how do i ensure i am not damaging it again? i am ofc on the latest bios, and intel defaults, how can we be sure increasing the power limits is gonna be safe?
What's the next steps if you want to continue further the undervolt after you do 55 AC/DC ?
Also i have observed on gigabyte motherboard it won't go below -0.080mv offset , it's like is not registering.
Keep up the great content bro 👏
You mentioned the aorus elite (and in general the boards with onSemi conductor)...can i ask what are the slopes?
Btw, i notice a weird behaviour with my z790 aorus elite. I tried HIGH with 0.5 / 0.5 mOhms (so...50 AC 50 DC) , and even by setting an adaptive offset the load voltage (cinebench) just SKYROCKETS from 1.22 to 1.32 or about that, and it doesn't even seem to be effected by my offset. I'm so confused right now. It just feels like the AC loadline is pushing it up. 13700k though, maybe the i7 behave differently because less current -> less droop?
EDIT: fixed by giving a negative offset to Ring voltage...
what voltage sensor are you looking at?
@@ActuallyHardcoreOverclockingvcore
@@ActuallyHardcoreOverclocking i tried again, resetting the bios to defaults.
Just changed power limits to 253/253w/307A with new microcode and intel specs set to 'performance' (the i7 doesn't have the extreme profile). AVX offset to 0.
Set vcore loadline to high, ac/dc loadline to 55/55, voltage mode on adaptive, negative offset of -0.150v.
Running cinebench single core: 1.22 - 1.23V vcore (but also VIDs)
Running cinebench multicore: 1.28V (but also VIDs) at 150A reported in VR IOUT (probably inaccurate?), 240W of power
What i think is happening is that the loadline is probably drooping about 40mV from the initial 1.23v but then the AC overvolts for 150A*0.5mOhms = 75mV, effectively jumping from 1.23v to about 1.27/1.28V ... but maybe i'm completely wrong about what's happening (also sensors on this board suck i'm just assuming current reading in hwinfo is somewhat reliable...).
EDIT: if i set -0.2 v of offset ... it doesn't matter. Same VIDs and same vcore
@@Profetorum same issue using the same hardware i notice that i'm able to go under 240W by lowering the AC/DC LL the problem is that if i go lower than 40 i will start losing performance but the wattage do decrease using high LLC i'm going to try instead of lowering the vcore offset i will try to reduce the Ring offset to see if this do something i remember i read that on a forum that on the aorus elite after certain point you dont see any improvements by lowering the vcore but i could be wrong
I'm literally scouring the internet for a z790 dark hero undervolt guide. Please make the videos for other motherboards!
13:00 - 13:30 was very helpful
14900K+MSI Z790i, BIOS v19, Microcode 0x120, Cooler Noctua D15
VID Offset (set through TS): Core/P Cache -0.11V
CPU LLC: Mode 4
AC/DC Loadline: 12/0.12 mOhm (the same value as LLC Mode 4)
IccMax: 400A
PL1/PL2: 288W
UVP: Disabled
IA CEP: Enabled
TVB Voltage Opt: Enabled
TVB Pts Config: 1 Core at 80℃ -1, at 85℃ -1; 2 Cores at 80℃ -1, at 85℃ -1 (with TVB Voltage Opt enabled, this will prevent high voltage, since there is no voltage limit function)
Everything is good. R23 41000 pts (limited by PL2, around 5.5/4.3), CPU-Z 980+/17600+ (fully turbo boosted, 6.0/5.7/4.4)
can you translate IA VR Voltage Limite to MSI z690/z790 motherboards? I think there is no such option there
Yeah, and from what I read on there forums, there is no plans to add it in the future.
MSI said on their forums (not directly, but via forum admin proxy) that they currently have no intention of implementing this setting. Quote: "After internal discussion, MSI have decided not to implement “IA VR VOLTAGE LIMIT” to BIOS. MSI suggest user try the latest BIOS with Intel microcode 0x129, which have already improved the CPU VID spiking issue."
isnt it cpu lite load? the modes change the DC AC thing
@@Golden2Talon Yes, CPU Lite Load is the setting to influence the AC and DC loadlines on MSI. In CPU Lite Load mode "Advanced" you can set them independently, in "Normal" it uses preset combinations. But "IA VR Voltage Limit" is something different entirely. With this, you can basically set a lower limit for the voltage spikes as an additional protection. Meanwhile, CPU Lite Load may lower the AC loadline and thus the voltage, but it can't do much about the voltage spikes, other than to lower the voltage baseline they're occuring from.
@@Mike500 what you can do on MSI Boards is, change the Voltage mode to "Adaptive + offset" and then set a negative global offset. You can combine this with limiting the Boost levels and don't use this 2 core extra boost. If you also set a reasonable lower Loadline and matching Loadline Calibration, you are most likely where you would be with a Voltage Limiter (MSI Boards don't have that). But then again, you have extra warranty, so why bother?
On my motherboard the loadline is set to high but I have to set Ac45 Dc45 so that Cep doesn't intervene. (Gigabyte Z790 Aorus Elite AX)
I hate mainboards so much... why is it called "Vcore offset" when in reality its "VID offset"
I thought disabling cep was meant to help with lowering temps at least according to msi? Is it not worth disabling for that alone? Never saw a solid answer as to why that was the case they just said disable it better perf lower temps
I tried your config of AC_LL = DC_LL = LLC = 0.55 mOhm with CEP on and an adaptive offset (14700k, F5 bios, IccMax = 307A, no power limit). My previous settings were basically an AC_LL/LLC undervolt. The most bizarre change I noticed is that the CPU gets current-capped at a higher frequency, which results in a performance uplift in some benchmarks. I wonder whether it's a side-effect of CEP or increased voltage under load. Going to run a few more tests.
With CEP being on and following the video's concept, how can I further lower the temps? Which values can I play with so that I can get similar performance without triggering CEP while keeping temps low?
Have to keep vcore higher than vid im pretty sure, or it gets triggered.
Hey! first of all, thanks a lot for the work you put in! i was wondering if you could do something like this with an asus board. like which LLC should i use and what the AI AC/DC louadline would be. thanks again!
Dude, this video was worth watching every second. thank you so so much
It would be great if you make video with ASUS motherboard instead of GIGABYTE. This company doesn’t update their firmware and BIOS so I would say it’s blacklisted by many people. Thanks!
Why does this feel like Teacher has just handed out lines to all the naughty ones in the class? 🤣
Write out one thousand time -
"YOU DON'T HAVE TO TURN OFF CEP to undervolt intel 13/14th gen CPUs"
Hey BZ, sounds like you're saying that there was a communication problem among the employees at Intel that caused the company to be in chaos from a business perspective!😅
Still waiting for someone to address what all of the EVGA owners of z690/790 boards are supposed to do since Evga’s MOBO team doesn’t exist anymore aka no micro code update
I had to turn off CEP to undervolt my 13600k. Tried everything else and CEP off is the only thing that permits an undervolt. Aorus elite ax rev 1.1 z790, 0x129. Clearly, there is an as of yet unknown issue with boards behaving differently. Under Voltage Protection is on. ac dc ll 50/50, vcore auto, dvid -100mv, llc high, 200A max., 1300mV limit, 125w&125w. vrout 1.126v, vcore 1.128, VIDs 1.214V. huge difference there but working on making it disappear by tinkering with ac dc ll.
I'd love to see you do a video with the 14900K and Aorus Elite AX mobo
About windows being slow, while this may be yet another different thing here, I remember this sort of behavior since windows 10 (never happened earlier), and it occurred when windows is absolutely hammering your SSD (or god forbid HDD...) and your poor SSD/HDD can't keep up because windows 10+ IO is just *that* bad. And yes, it is bad enough to even get bottlenecked by an SSD. If, during the slowness, task manager shows 100% HDD/SSD utilization with shockingly low read/write speeds and insane latency, you are hitting that issue (I remember getting 1000ms on an SSD, and 2000+ on HDD)
My laptop has this issue. Latency sometimes in tens of thousands of miliseconds and 100% use. It's basically frozen until it finishes.
Had an HDD I was using for media storage that started to fail and it basically gridlocked my entire operating system on Windows 11. Computer wouldn't even shutdown properly with it connected.
@@AzwraithPL 3 SSD's connected via USB - same.
should take a look in your device manager and enable hidden objects or whatever its called - anything you know for sure is bunk should be removed, then restart.
as for usb sdd's - I only have issues randomly and very infrequently, and its not a Windows issues, its a mobo issue wrt to USB - you can change around which ports are being used and get very different results.
What does MSI Lite Load setting do (as opposed to load line setting)? Ive set mine to mode 3 which gives me much lower temps and no performance degradation on my 13700k. CPU temps under load are around 75c, whereas with the stock setting of 9 it hits 100c. Is this a good way to prevent or reduce the voltage spikes? Never had a crash in 18 months.
If you can go to Mode 1 without crashing it's even better. the lower you can go, the lower voltage the mobo give to your CPU, so it depends on your CPU quality, if it can perform at Mode 1, then you're good to go.
Lite Load DOES set Loadlines (AC_LL and DC_LL) the thing is that it hides this behind some abstract numbers on "Normal" mode. If you go to "Advanced" mode, you can see what it does and you can also set AC_LL and DC_LL on its own. Intel default goes to 110/110. If you use "advanced" mode instead of "Intel default", but enable both CEP settings, you might end up on 80/80 or something (I don't remember exactly).
MSI Lite Loads are basically just load line profiles. For example I believe Mode 3 would be an AC_LL of 15mOhms and a DC_LL of 80mOhms on my MPG Z690 Edge.
Hey BZ, I really appreciate your work. But most of the time I dont understand what you are speaking because stuff is too technical for my small mind. I request you to make a small summary or a follow up video of 5 mins eg what settings should we dial in to save cpu from degradation without much impact on performance.
Great video as always BZ. I did all the settings you have shown but if I try and raise my CPU above 5.4ghz it will start to hit the 253w PL1/PL2 limits. Do you know how I can further tune my 14700k to stay at 1.35ish volts and at or less than 253w? I have a score on my Aorus Master X of 94 so my cpu should be ok as I have not tried to run it past 5.2p cores and 4.0e Cores due to intel's issues.
Hey man great video, i have the B760 and i updated it to the new microcode 0x129, intel default settings is not in the bios before or after i updated the bios.
And the CEP setting is the same as u have Auto = disabled.
I don't have the "offset" options in my motherboard so im stuck with the dynamic Vcore option.
Im getting better score on Cinebench 24 when i apply a negative value in the dynamic Vcore.
That means the CEP is actually disabled, right?
B series can't disable CEP, it's forced enabled at the hardware level. I assume your CPU is a lower tier than the 13600K which is the minimum they have programmed for intel default settings
@@pixels_per_inch i have the 14700F , and I think it is disabled , because my Cinebench score is higher after i undervolt it using the Dynamic Vcore.
@@7MDv Ah right Intel started allowing disabling CEP on newer 14th gens steppings
Is there a chance you can link to the video where you show how to properly undervolt 13/14 gen cpus? I can't seem to find it based on video titles.
My MSI Z790 ACE has a Renasas controller, and uses the same multipliers for LLC and AC/DC LL as your Gigabyte mobo.
Do you think motherboards from different manufacturers with that controller should have the same multipliers?
So I got a terrible 14900k, I can run only the following settings without triggering CEP (even out of the box triggers CEP) on an Asus z790 board.
LLC = 3
AC = 0.66
DC = 1.1
On top of that, I slapped a -0.100mv IA Voltage Offset into it, and it's stable on C23, Prime95 and Hogwarts Legacy Shader compilation.
At this point I dont even know anymore whats correct or not, but as said, this is the only way for me to run it without hitting CEP (or using the worst ASUS profile, that puts AC = 1).
did you change svid behavior ?
@@nemesis762 no, it doesn't matter, it only changes AC/DC values. Auto is AC/DC = 1, typical changes AC to 0.4
Great video honestly. helped me tweak my 14700k with CEP back ON and no loss in performance at all. still getting 35-36k on CB23 and 2k+ on CB24. Thank you
I have the same cpu, what settings did you end up using/changing? I still struggle with performce loss with CEP enabled
@@Inasari2 I just used Intel performance profile on my z790 aorus elite AX board and went ahead and changed AC/DC loadlines to 55 and Loadline calibration is set to high. and put in a 1400 IA VR Voltage limit. and then I undervolted using adaptive voltage like I always have done but this time I put in -125mv vcore offset. I used to have it at -80mv vcore offset but at that time I had CEP off and LLC at low with AC LL at 40. I also went into advanced cpu and put in a TJmax limit of 85c and disabled AVX offset. that's about all I did in terms of cpu tuning.
You mentioned that you find undervolting via VF Point offsets hard to tune. But what if the same offset would be used for undervolting only VF Points 7 to 11 while leaving the lower ratios untouched? Do you see this as an improvement over applying the offset globally? I am a bit concerned about lowering the idle voltage and lower ratio points too much and am not sure how to test stability in that area. That is why I thought of only applying an identical negative offset to upper part of V/F curve.
That's what I use. 0.068 undervolt on points 7-11 on my 14900K Z790 Dark Hero.
15:01 wow that is even lower than my 13600K which scores around 20k, which is actually quite a bit lower than what a 13600K should perform. Maybe that's because of degradation or maybe it's because my Deepcool AG620 air-cooler is not enough for this chip.
What score should an i5-13600K get ideally?
@@BucifalulR around 24-25k is what I had seen on other people's, including reviewers, benchmarks. It could be that the whole debacle with degradation and the new Intel Default Settings and the latest 0x129 microcode have caused some performance regression, but from what I have seen around the internet, it shouldn't be as bad as 20k on Cinebench R23.
@@BucifalulRMy 13600kf gets 25k+ on my old settings and slightly under 24k with Intel default settings.
@@RandomGuy-ll9uz On CInebench R23, right?
@@Dhruv-qw7jf I got 20802 only :(. If I apply Buildzoid's IA VR Voltage Limit of 1400 (Gigabyte B600M Gaming X DDR4 motherboard), I get even lower, like 19000. Maybe I should try with Intel Default Setting disabled, but doesn't that increase the risk of degradation?
Hello BZ, recently i updated to newest bios version 3401 and i lose too much performance in intel 12 gen cpu, specificly my intel 12400f, what u recomend me to do? i appreciate your job!
It you want to measure something like Die Sense with a 'scope DO NOT connect a piece of co-ax and 50R terminator! Instead just add direct very short wires to the relevant point, plus a ground as near as possible, and connect a 10MOhm X10 scope probe directly to those short wires. As a graduate EE what you are currently doing is painful to watch.
Can I also suggest that this could have been literally a 5 minute video showing CORRECT settings for undervolting without triggering false CEP.
Voting for a loadline video on the Asus TUF Z690
same, all ACDC LL settings that I try triggers CEP on ASUS TUF z690 D4
This any Asus Board please =)
25,000 ish points is what I usually got with my i9-12900k lol
Do you have a video covering how to safely undervolt/overclock the i9-13900k on the MSI z690a pro WiFi board you have? I just “upgraded” to the 13900k and now come to find out there’s all these issues with these cpus lmao
I’m a video producer and editor so time is money for me, I need my system to be fast but reliable. I should probably move up to the z790 board but, the specs didn't seem very different, maybe I'm wrong. Any help? 🤙🏻😎❤️
maybe I won the silicon lottery, I got a stable -0.175mv with the same video configuration on an i7 14700KF
"So if I understood correctly, the 'Vcore Voltage Mode' should be set to 'Adaptive Vcore', then set 'CPU Vcore Loadline Calibration' to 'high' and lower 'IA AC/DC Loadline' to 55, and then the 'Internal CPU Vcore Offset' should be adjusted from something like -0.080v to the -0.110v to tune the point where the CPU is stable?
se basicly 14k views and no person know anything about my ask :D ... and btw my r23 score with this settings is 39150 and max 1.385v and everything stable ... i tried almost every benchmark from BenchMate and im very sadisfried my temps cant go over 73 degrees but it is to god for me to feal ok :D so i need ask again can someone tell me is it ok what ive done it is described up.
If anything this just made my CPU run hotter
Thank you for taking the time to explain this, Buildzoid.
Does Intel XTU no longer work for undervolting 13th and 14th gen CPUs? I have a 12900K and use Intel XTU to undervolt; nothing in BIOS; just default settings.
13900KF with Gigabyte Gaming X AX (1.x revision if it makes any difference), I set my loadlines to 15 and loadline calibration to extreme. with adaptive vcore and vid offset -0.085V. Both CB23 and CB20 run stable with 23 at ~37-38k and cb20 around 9k.
peak VID requests in HWiNFO are are around 1.25V but can jump to 1.329 in idle.
But, and here is where I stumped, this is with only 5.3 ghz on p-cores and 4.1 on E-cores.
It seems I can't get 5.5/4.3 ghz in CB, unless I turn off intel default settings and go with whatever gigabyte uses, which crashes on most U4 shader building/cb20 etc.
and another thing i noticed - with these settings if i open intel XTU monitoring
there is something called "current/edp limit throttling" constantly being triggered, and no obvious correlation with say p-cores getting boosted to 5.8 ghz during these limit throttlings (or they just don't line up due to only having garbage ITE I/O monitoring).
Did anyone encounter something like that?
System is stable with current settings, it was stable even before intel default settings except for u4 shader compiling and cb20, and not even all U4 games were affected. But as I don't have a way to figure correct loadline settings except by just going lower than the default ones (90/90) until CB20 starts crashing and I don't have garbage scores (like 25-30k if set to 55/55 and high calibration) I am a little bit nervous if 15/extreme and -0.085v undervolt with adaptive vcore enabled but "current/egp limit throttling" getting triggered (maybe XTU overrides my settings as I see that with it running VID is higher), if it's fine or not.
the current limit is probably getting triggered by ICCMAX. Which happens way before the VRM outputs ICCMAX levels of current because ICCMAX is supposed to include the ripple current flowing through the decoupling capacitors and the VRM can't measure that.
@@ActuallyHardcoreOverclocking did take a look - in hwinfo there is a metric called IA: Electrical Design Point/Other (ICCmax,PL4,SVID,DDR RAPL) it gets triggered when I start CB23, but it's not constantly triggering on/off while idle, so probably intel XTU applies some garbage settings overriding (at least partially) what is set in BIOS.
XTU already went into trash, but more interested will I have positive effect if I set ICCMAX manually to what Intel suggests in their documentation? (307A, if I remember correctly)
VR VCC Current (SVID IOUT) shows maximum of 206A, but at the same time I have 253w PL1/PL2 default settings, so probably due to 253 max power ICCMAX adjust accordingly?
When I run your settings exactly from previous videos, I get crashes of apps with any kind of undervolt, but cinebench will run perfectly fine for 30 minutes with a 100mv undervolt
Because it's hitting the power limit and running at lower frequencies which won't necessarily be unstable. Something stable in Cinebench could crash in a game which runs closer to 5.7, or the Timespy Extreme CPU test (in a loop) for instance. Or in any number of other scenarios, which is why extended testing is crucial.
@@dd7428I play iracing, the funny thing is the actual game doesn’t crash running at 5.8ghz its the timing apps and discord that crash, the actual game runs without skipping a beat and while its doing this the cpu is pulling around 80w
you crash even with a -50mv undervolt?
@@ActuallyHardcoreOverclockingthe game runs fine even with a 100mv undervolt, but the other apps like discord and my timing apps which use a browser instance and an overlay that crashes with the undervolt, even the browser instance of simhub that my screen display uses crashes, I think to get it not to do this the undervolt is -0.020. Also I see you selecting the actual voltage, is there a reason you don’t just type in the offset you want ?
@@PanterlucaI usually only have 1 hand on the keyboard connected to the test bench because it's awkward to reach. So typing things that require non-number characters gets annoying(like decimal points).
Also it sounds like you just lost the silicon lottery. I can't imagine that undervolting with CEP turned off would work better(it doesn't help my CPU).
I wish you would attempt these videos on an MSI z790 Board to see what you can and cant do when settings YOUR particular board has but others (such as MSI) doesnt have. Such as IA VR LIMITS we cannot set any because we havnt the ability to do so on MSI boards cant even sync LLCs without full manual tuning. Its terrible how different boards are. There should be a mandatory labeling across all manufacturers and the only difference in settings being between AMD and intel only. If youre on an intel platform the terminology and access to settings should be a standard across all boards. This stuff is ridiculous now.
check the guides on the MSI forum, they helped me
forum-en.msi.com/index.php?threads/guide-how-to-set-good-power-limits-in-the-bios-and-reduce-the-cpu-power-draw.400270/
forum-en.msi.com/index.php?threads/different-undervolting-methods-with-ia-cep-enabled-and-how-they-compare-to-lite-load.400984/
Would these settings also apply to a 14900KS on same board?
Of course, not sure how much undervolt wiggle room you’d have though. Since they are better binned, their VIDs are already lower compared to most K’s and KF’s. Unless you got a really bottom of the barrel KS, I’m sure they exist.
@@MhillPlays I'm gonna mess around in bios and give the setting a try.
I am on fixed 1.1v with "normal" load line. 4.4ggz locked on a small cooler. Max load at throttle temps 125w.
Im wondering if this undervolt method could result in more performance.
How would you decide on optimal AC loadline settings on a platform that doesn't have LLC or a vcore sensor i.e. an Intel HX laptop?
Having a lot of fun trying to fine tune my MSI Vector 16 HX with 13980HX (I.E. a 13900k in a laptop) but I don't know how to the optimal ac loadline setting other than to first set a voltage offeet on the edge of stability and then adjusting AC offset and seeing whether heavy loads or light loads cause the first crash. Ideally a properly set AC loadline will allow maximal offset at low loads without crashing either low or high loads, correct?
Ironically, my MSI laptop bios DOES have an IA VR limit and I happily run all core 5.6ghz on light loads with a 1.3v limit.