They are straight up scams tbh (except if you are an extreme overclocker) regardless of budget very few people have reason to go above the "Master" or "Hero" and equivalent lines. I'm sure buildzoid would say the same thing lol.
"Scams to be honest" Wow have you got some bias let alone misusing the English language. Some people appreciate and understand better vrms more pcie and lan performance others just live in ignorancce and make simplistic if not purile comments like the quote above.
23:45 - I noticed the same. When I was on the edge of stability I had these symptoms: - Task manager being slow to load (sometimes it took 5s to switch between "processes" and "performance") - Scrolling through youtube comment section caused stuttering. - Dragging windows around was choppy. All these symptoms were gone when I raised the voltage by 30 mV (-120 mV to -90mV).
@@Savigo. You can also check in your Windows Event Log for WHEA errors. For ease of use, it's best if you set up a Custom View for that. Either google how to, or the short version is to use "By source" and then select "WHEA-Logger" as the "Event source" when setting up the filter for the Custom View.
Why am I getting ~2250 pts in Cinebench R15 with i7 14700k. What should be the optimal score for this specific CPU on 5.5Ghz(Default) with intel optimized settings. PL1&pL2 253w
Applied these settings with 0.150 undervolt on my 13900KF and so far looks amazing! Gained 2K on R23, better temps and voltage goes to maximum 1.2 Love your content my friend, thanks for your teachings. Edit: after a day or two it gave some BSDOS so I changed to -0.135 to be safe.
Thank you for doing this with an MSI board! It's much appreciated!! Applied this to my Z690 Unify and an MSI Z790I Edge, of course with some slight variances. Great results on both! Thanks again!
Thanks Buildzoid! I was able to undervolt my 13700k on the Pro Z690-A DDR5 with CPU Lite Load Mode 3 and CPU Load Line Calibration Control Mode 3 and Adaptive + Offset of -0.170. Runs smoothly in benchmarks R23 (Score: 30624), R15 (Score: 4570) and all my games. With an offset of -0.180 I had crashes in R15. My stats in HWinfo while running the benchmarks with the -0.170 offset: max. Vcore: 1.178, max. Power consumption: 193 W, max. Temperature: 81°C.
As a MSI User I want to thank you for your Testing and Insights. You've found out yourself that we are plagued with the limited possibilities like the for some reason not opened IA VR Voltage Limiter. May I ask you why you did leave the undervoltage protection "enabled" does it do nothing like the AVX guardband setting that I hinted you (and was wrong) on the last Video? edit: Wait a minute it IS disabled on your last Screens of the advanced Menue (Minute 38:41), but when you look at around Minute 11:17, undervoltage Protection is "enabled" (both times settings are on "Auto"). MSI things...
Undervolt protection just sets a floor to the voltage settings at boot. You can still set them wherever you like in the BIOS, but once its booted, software can only leave it alone or increase it. XTU requires UV protection be on (if virtualization is enabled). While I was tuning, I'd set an aggressive UV in the bios, and then bump it up a bit in windows with XTU when trying various raised multipliers, and then go back to the BIOS to commit those final values when I was happy with it.
That voltage spike at start-up is scary. Luckily MSI was setting their AC loadline to 50mOhms at default previously which would have helped curb the voltage up until now, but since I'm pretty sure the new BIOS updates just set AC_LL = DC_LL there's going to be a lot of people who updated their BIOS in the hopes of prolonging their CPU's life only to be slowly frying it every time they turn on their computer. And the scariest part is you have no way of knowing if you're affected or not unless you have an oscilloscope to hook up to your board.
@@ActuallyHardcoreOverclocking True, but I think you just have a particularly bad sample. And in this specific case exposing bad chips is probably preferable to frying good chips.
@@ActuallyHardcoreOverclocking Why my i7 13700k never hits 5.3ghz in gaming after microcode update? Now is 4.9 max.My sp score is 92.I have gigabyte mobo,i use intel default
Funny enough, for me, the best settings are the ASUS OC optimized ones. Highest voltages ive seen are about 1.31 (including a +0.02V offset). 13600K, 5.4GHz 1-2 core, 5.3Ghz all core load. Im lazy, i just use intel extreme tuning app's with its Speedoptimizer 2.0 turned on at boot. If i do the same on the intel default settings profile, it goes crazy. Pushing 1.3V on all core load CB23, when it should do 1.2V. Results in lower clocks and 100C on some cores. ASUS profile it just works fine, temps 75-80C, all core voltages 1.21V on the same test.
There's yet another bios released for this board, not sure what the real difference is, 8/21 version "7D25v1J" -- doesn't have beta so not sure what exactly the difference is though. edit, not sure why, but MSI has a really weird naming structure, the earlier "beta" one is listed as version 1J.1, where as the newest (listed on their site as released 8/21) shows in flash as 1J.0. "build date" is literally the same between the two in flash as well (8/15).. which is stupid to me why they would listed as that. God MSI is really starting to piss me off these days.
Anyway my biggest compliment to you Buildziod. Is i really appreciate you acknowledging what you might have learned from all of this in terms of what you might choose to eval in the future. Who would have thought of this at that time about all this? I think going forward it would be nice to add this analysis (voltage control) to the mix. What is the actual voltage, how is it control, does it make sense et, not just what the board can do in terms of peak, sustained power, but almost like power quality--- smoothness etc. I do kind of wish they just had a hard voltage limit. I'd just set it to 1.35. I was looking at this over the weekend, and the llc, and lite load is this complex balance. You try to stop the cpu from over shooting or dropping, and stopping it for over responding to that drop. I basically had to put this in terms of a hose and trying to fill water, and what happens when you plug the hose quickly. and all other kinds of things.
Good video, really helped me tune my 13700K. I'm running the DDR5 Wifi version of this board and installed the newest, non-beta BIOS they pushed on the 21st with microcode 0x129. I dont' have my own probe so I can't monitor the voltage during boot, but I was noticing some pretty high spikes on stock settings, pretty close to, or at 1.5V when running Cinebench R23, and I was thermal throttling. I have a 360m AIO sand it's mounted fine, so it wasn't that. Following your advice, I dropped the CPU Load Line down to Mode 8, anything lower than that and I was seeing performance loss, with a negative offset of 0.07 and now I'm getting better temps and a bit of a performance boost from stock. Bumping up the offset to a -0.08 also results in some performance loss too, but I may have to play with it a tad more. I think I'm dialed in quite nicely now, Cinebench R23 and R15 are only getting me spikes of 1.3V in single and multi core workloads. I've had this CPU since launch and the only issue I had with this PC since I built it at the end of 2022 was the first 4080 I got ended up being a dud and causing me all sorts of problems, but replacing it with a new one resolved everything and it's been great since. This video was very informative and did a good job explaining all the CPU Load Line stuff that I was otherwise getting confused about. Thanks!
Thanks a ton!! Exactly what I was looking for.. Just to add - Someone posted exact impedence values of LLC levels and Lite Load AC_LL/DC_LL values for this motherboard on some forum recently. Apparently, you wanted to set 0.3ohms, which is why you selected Level 3. But Lite Load Level 3 and LLC level 3 correspond to different values, not AC_LL of 30 and 0.3 ohms respectively
Massive massive thank you. Followed the guide got my 13900Ks stable 1.27v under load while retaining my 6ghz single core boost highest hwinfo voltage 1.31v. Really happy with this thank you. Vrr
Great info!! The way the CPU calculates instructions varies from one app to another too. This can be why some games work and others don’t, and even might be why some CPUs have degraded faster than others. All this should be considered when UV’ing and stability testing.
Intel has a serious yield issue that they try to compensate by loosening binning specs. There was that top performing R-batch, the reason why we went to the stores, but since then bins have become a box of chocolates.
Thank you! This helped my VCore and temperatures on my i7 13700k. When I ran Cinebench R23 at stock with 0x129 microcode, I would get a max of 1.464V and immediately get 100 degrees C. Now with Lite Load 3 and a 0.08V undervolt, I get a max of 1.248V and it immediately goes to 91 degrees C and slowly creeps to 100 degrees. That’s much better and increased its score!
Just tried this on my 13900K, got 35,700-ish points in R23, not too bad, no different than stock for the score though, my chip probably sucks, dunno, I was hoping for 37,000 points because I was able to do 39,000 on previous BIOS'. But I do have lower voltages now which is very nice for longevity though. Thanks Buildzoid.
If I remember correctly with these boards I believe LLC3 is 0.10 mOhms and Lite Load 3 is 0.15 mOhms on the AC_LL so you'd actually want to set your LLC to 4 if you wanted it to match your AC_LL at Lite Load 3. Although I think the Lite Load DC_LL is set to 80 mOhms from modes 12 to 2 so your power readings will probably also be thrown off doing it this way. Personally I just suggest manually setting your AC and DC loadlines rather than using the Lite Load presets.
@@ActuallyHardcoreOverclocking Actually it seems like MSI Lite Loads have different impedances depending on whether CEP is enabled or disabled so what I said wouldn't necessarily apply in your config.
Please help us to know why Ring Limit Reasons is always YES even in this video ? Is there any issue with it ? Any way to fix it ? Whether setting ring ratio and ring down bin will solve this issue ? 65w non K CPUs didn't have this issue but K series having it(checked 12400 and 14500 CPUs). Ring may be getting high voltage. Is the initial high voltage causing it ? It will be very helpful if you can add a video about this issue. A lot of people are confused about it and thinks this may be one of the reasons for degradation.
Hello, I have the same issue, but came to the conclusion that it is just an unfortunatly named hardware limit flag. The reason it's 99% on yes even in idle is that the ringbus frequency is linked to the maximum P-core frequency minus a 5-600 Mhz offset. So it constantly "throttles" depending on the P-Core clocks, but that's how it's supposed to work. Not all these flags indicate a real issue with the system, "IA: Max Turbo limit" for example and the "RING: MAX VR Voltage, ICCMax, PL4" is one of them.
Would love to see you test a board with the OnSemi controller next. My i9-14900k runs on a Gigabyte Z790 UD and I'm pretty happy with my current settings. It's fast and cool even when boosting to 6Ghz (running a Liquid Freezer II AIO). Cinebench R23: 39k, 2.3k single - 85°c max *My settings* Intel Default Settings: Extreme CEP: Enabled AVX Settings->AVX Optimum: Enabled (this removes the Gigabyte Default 7x AVX2 ratio offset) Adaptive Vcore Offset: -100mV CPU Vcore Loadline Calibration: High IA AC Loadline: 0.45mOhm IA DC Loadline: 0.45mOhm IA Voltage Limit: 1400mV Been running like a charm for over a week now, no stability problems in all BenchMate tests as well as Prime95 or games like Cyberpunk 2077 (been playing for many continous hours lately). I'm kind of a noob and your videos helped me a LOT while figuring stuff out. I was kinda scared my CPU will degrade and I'm much more comfortable now. Thank you so much for making them!
29:01 I have observed the same thing, but like you I am on Windows 10 without thread director. So I wonder, if Windows 11 can run CBr15 with these settings.
MSI annoyed me with that 'automatic extreme settings' thing too. I have a 13600KF, so thermal throttling was never really an issue, but it did spend a lot of time throttling due to the new 200A current limit in the intel defaults. Undervolting did eliminate that without having to go too far. Then I wanted to raise the clocks from the stock 5.1/3.9 P/E to 5.4/4.3 that I knew it would do at that voltage, but MSI insisted on switching the power profile when I did so. As far as I can tell, the only thing it changes is setting PL1/2 and max current to unlimited and I was able to change it back, but still annoying. With the overclock It still hits the current limit (up to 220A) in all core, but performance is no worse than stock and its better in general use applications that don't throw 20 threads at it (the 3d mark CPU test even showed a small increase in 16 thread)
Why use adaptive+offset instead of just offset? I thought adaptive+offset would only offset voltages at high load, hence a single core workload would still be able to request a high voltage.
I appreciate how this generation of home computer hardware is making it more mainstream to learn to tinker with your PC and never trusting manufacturer default settings out of the box.
Hey guys, can someone clarify something for me? Is having a Vdroop a bad thing? I'm sitting on z790 with 14900k with Intel defaults, and LLC 8 with AC80 DC110, adaptive offset of -0.12 and according to HWinfo64 under full load it sits around 1.20, in gaming avg 1.27v with peaks to 1.33v iirc. Here Buildzoid used far more flat LLC with lower AC DC which if I understand correctly leads to smaller Vdroop. On his earlier videos I'm pretty sure he didn't went as far down with those numbers. What's the benefit of doing this? Thanks in advance
A couple of things here. First off, a lower, droopier, LLC has better voltage regulation and helps with the voltage spikes. The thing you have to worry about is at idle or when it boosts to its 1-2 core max, the voltage requests go through the roof. The other part is, why did he choose a higher, flatter, LLC and go against what he said in previous videos? Well, back when he made this video, MSI boards didn’t have the option to set a voltage limit. So, if you were using a lower LLC with no limiter, it could potentially degrade your cpu. However, with a higher LLC, voltages come way down. Yes, you’ll get spikes. But even at those peak spikes, they’ll be less than the top voltage requested with a lower LLC. So, he’s saying to do this if you can’t limit your voltages.
In my case on MSI mb if I only selected "intel default profile" I lost 50% points in Cinebench r23 multicore... So I swapped back on MSI profile and do undervolt, but I don't know if I'm doing well or not now.
In the File drop-down menu, on the upper left corner, is an Advanced Benchmark option. This enables different run times, including single pass benchmarks. This works on all versions of Cinebench.
Just updated to the bios released on 21 of august,had no prior problems and this seems to work fine as well. Using default intel settings and before that limit to 253W. I had to update the bios when I first received this mb, also MSI pro z670-a ddr4, or the 13900k wouldn't work. Updated it a few weeks ago to the most recent full release (April this year) and now 7d25v1j non beta. Only thing that's sometimes happens is that disk io freezes for 15-20 seconds, then continues like normal. Could be a qlc flash thing I have as data storage (Samsung 870 qvo 8TB, Samsung 980 pro 1TB OS drive)
I just have one question, why are you not using the same or similar LLC settings like switching frequency etc. that you used in the MSI Pro Z690-A voltage regulation and VRM settings video.
Mmm so.. You are doing your own Curve Optimizer for Intel?... 5.1 to 5.7Ghz all core workloads you get with -130mv That is pretty nice... If it works a hahaha... Then, with the new Curve Shaper of AMD That specific problem of undervolting can be FIX?
you probably already saw many asking, but just wanted to leave another request for an asus bios optmization. 13900k/14900k. im a noob that does not want to mess with settings im not qualified to take risks with
I had a weird bug on my MSI board, if I switch to the 4096W power settings, it basically massively reduces the performance to like 16,000 R23 points, you can even switch the power profile back to Intel Default Settings and it doesn't register it as the default profile, it stays stuck bugged till you clear CMOS. Beta BIOS' are so much fun!
My under volt was stable for a month or more, using the 0x129 bios and verifying with all the usual stability tests and gaming, until I had to install nVidia drivers. It didn't really crash because it was just on the edge, but installing the driver was unusual slow, to the point I though it was frozen.
Not too sure what my issue is, i have the same motherboard but DDR5 and if i run Lite Load mode 3 i get blue screens and all my games say "lacking video memory" brand new 14900k with less than 100 hours on it. I did -0.070 offset and that works but im still hitting 1.550v in game. Microcode 0x129 (0x12b isnt available to my motherboard yet)
What can I do to find out if the VRM is actually showing correct voltage is? Do I need a special instrument? Also, after flashing the latest micro code, I was able to set a voltage cap on my MSI B760-Vc WiFi 7 board.
The best way to overclock a GPU on MSI afterburner is to make a custom curve where you use the highest offset at the high frequencies that are needed at heavy workload, and gradually lower and no offset at lower frequencies. With the curve you can also cap the frequency at a level you know isn't stable, but that the GPU might opportunistically shoot into for a split second with erratic loads, but isn't practically useful. Can't believe we still don't have this for CPUs, it would be a huge selling point for a motherboard.
Thanks for the video. Got my MSI Z690 down from 75% voltage throttle to 10-15%. Going from .100 to .110 negative completely blows the system up, CPU and GPU bound on Iracing. Maybe I'll try 0.105 next to see...
Would it be possible that you do a video on undervolting with cep on on asus boards z790? There is no vcore voltage, IA ac and ia DC are in different places, happy I found them, load line has different settings and something with synching vrm voltage with ac DC loadline, an there are level 1-3 loadline and then also load line 1 to 8, I took 5 bacause I counted from your video the options on gigabyte and then used it in number form. but then there is vrm voltage offset, svid voltage ofset, and the v/f curve, but no general v/f offset voltage, only per ratio offset voltage, and there is no dynamic voltage offset and I don't know what to change, I have some crashes but don't know the reason, both with vrm offset and svid offset, and the v/f curve is more confusing, because I don't want to do something wrong, I don't know if I can kill my cpu with undervolting and I don't know how to adjust ratio and tvb oc, and things like that with Intel default, because if I enable the 2 best cores to 60, in tvb it says all cores have a 60 ratio, even with all core, and then I can't fine tune it really, and what does xtu do, can I use xtu for things like that and what does xtu affect? Asus calls things different and there are some things completely missing (generally v/f offset, dynamic offset, vcores) and also I don't know what does what (svid offset and vrm offset and things like that, cpu svid, cpu power request, and so on) and a video on a asus board z790 would be very helpful if you could do that, because there are no explanations manuals for the bios and no guides with "translations" to other board naming schemes, and so I am at a los and a bit scared trying it on my own without a real guide on it.
If you could also probe into the behavior when TVB voltage optimization (enabled by default) is disabled, that would be great. Even if AC_LL has the same value, there is a difference in the voltage the CPU gets when it is enabled and when it is not.
Would this apply to other MSI motherboards especially after the most recent update? say MSI Z790? I've tried this just now with a 14700k and yes it's a little slower than it was at factory standard but my voltages are 200 millivolts less during helldivers 2. I used the exact CPU Liteload and LLC you did here. Along with a negative variable offset. Nothing else. Some people are spooking me saying a LLC of 3 might be too high and could cause voltage spikes that even my sensors cant pick up (using HWinfo 64). My last CPU was a 4670K so I am a bit overwhelmed by everything going on.
I was running happily at 1.25 volts thereabouts, even upping the power limit PL1=PL2 to 253W as a test. I've lowered it back to 125W to 175W just to keep it running a little less hard. I don't need all the power all the time. Stability so far seems good.
Applies this Prroblem also to an msi Z790 carbon WiFi Mobo? Usually LLC 3 is to high for my understanding so far I was going with 5 or 6, but frying that buddy every time.windows boots is also not what I want...
Looking forward to your asus video. Still a bit confused over LLC and its relation with AC / DC load line. I miss the simplicity of overclocking in the old days (athlon xp / p3). Tweak the vcore, increase the cpu multipler, pass prime95 and you were golden 😂
I find it interesting that with a -40 mV offset, AC 60, DC 82, and LLC 7, my average core VID and average VR Vout readings are within 3 mV difference, when running R23 multi core. I thought AC != DC would make the readings off. This with a 14700k and Z790 ACE
Hey so using these settings on my MSI Pro Z690-A Wifi DDR4 with the i9-13900K I get an AC/DC load line of 0.2/0.2 With PL1 253W PL2 253W ICCmax 400a and a -0.080mv under load the VCore stays right around 1.25V and pops up everynow and then to around 1.35. For the most part it seems like things are mostly stable, temperatures are fine except for 1-2 PCores CONSTANTLY pissing me off at 100C while the rest are chilling at 91-94C under heavy load. Anyway, I do get some weird slow downs when working on things not at full load but no hard crashing. I'm thinking the load line might be too low would that be a correct assumption? And on the MSI board, going from Mode 3 to Mode 4 etc increases the load line setting right?
The AsRock Z790 boards have a hard limit of 100mv negative offset. Also they don’t have a setting to limit the voltage vid request. Not that I can find anyway.
You can undervolt further by using OC Tweaker/FIVR configuration instead. Set core voltage mode to adaptive, VF offset mode to legacy, VF configuration scope to All-core. Then under core voltage offset you can enter the amount of undervolting. Hope this helps.
I did the 150mv undervolt offset on my 13900K with MSI Z790 Edge WIFI DDR4, R15/R32 Score are now 6123/38409 up from about 5500/36000. The CPU isn't hitting TjMAX with a temp of 103C anymore, it's running at about 1.200v with a max temp of 93C. I think it needs repasting, but I'll be going with a thermal pad soon. Anyway, I'll change the undervolt offset to 140/130/120mv, etc. if I find any instability. Thanks for the video!
What if I'm getting 1.4v on my current BIOS (7D91VHB has 0x123 microcode) and averaging around 1.3v in Z790 tomahawk WiFi. And what if I underclock to 5.1Ghz and also apply -0.035v adaptive+offset it doesn't exceed 1.21v. Do I need to update BIOS and worry about voltage issues? In HWinfo AC load line show 0.4/1.1 mOhms as I've not limit anything thing in BIOS. CPU lite load set to auto(Mode 09). Just tell me should I update and tweak a little bit or keep my current settings 5.1Ghz and slight undervolt.
I have a z690 force wifi and a 13700k right now. I was able to offset my voltage by a decent bit and I don't hit anything above 1.25 on single and 1.2 on multi using the new intel default.
What values does CEP actually compare to decide if it should trigger or not? From what I understand (excuse any blatant ignorance please) there is lets call it default VID, Which is what is written on the cpu vid table, the AC LL adjusted VID, which is what it actually requests, the VID readout, which is the AC adjusted VID further adjusted by DC LL (?? or original VID djusted by DC LL???) and the vcore. So... what does CEP compare to what? Like, if the vcore matches VID in HWinfo does that mean the CEP is happy? or is it going to still do some -magic- weird math and still freak out because oh no mercury is in retrograde today and also vid readout doesn't match vcore divided by AC LL multiplied by the square root of current or whatever tf? Really sorry if you already covered this and I missed it.
Is it better to have AC/DC 0.1/1.1 (Lite Load 3) or 0.17/0.17? In some previous video you suggested to have the same AC and DC, but it wasn’t MSI motherboard. I’ve been using Loadline Calibration 5 with 0.130 offset and AC/DC set to 17/17. In HWinfo I had VID and VCore average almost the same (>0.05 difference). Do u think it’s better to go to LoadLine Level 3 and LLC3 with 0.135 Adaptive offset? It’s also pretty stable by far, but hits 96C with 185 PL1/PL2 when the previous ones max was 89. I’ve got i7 14700KF and MSI MAG Z790 Tomahawk Wi-Fi DDR5
on the new BIOS Intel defaults, my MSI motherboard must have been doing what yours is doing, because it would hit in excess of 1.55v on windows startup, according to HWinfo64. (don't have an oscilloscope, so I can't know for sure, I guess.) Great video, as always.
Hey Buildzoid, can I ask you a favour? Can you do a "quick" comparison between the Raptor Lake voltage behaviours and Alder Lake? To demonstrate how even if both are set to the same voltages (and clocks?), Raptor Lake has higher excursions due to the overshoot/over aggressive voltage drop compensation? Preferably a 13600K or some other lower-clocked/power Raptor Lake CPU that's more comparable to Alder Lake? Like a 13700 vs 12900K? Just enough to see the boot behaviour + single-threaded R15. A default/stock VF curve comparison would be cool too.
@@godnamedtay ? It's not impossible that he does it. Get a grip. He's clearly interested in investigating this sort of thing or he wouldn't do videos on it. There's nothing BS about asking. You never win if you never roll the dice. It's not like he has public rules about what he does or has stated he will never do a video about something someone has asked about. In reality, he has made many such videos, specifically because people keep asking.
Oddly enough I find RPCS3 when it compiles PPUs at the start of a PS3 game is a great way of testing high frequency stability. If it crashes or logs a WHEA error in windows you are unstable. You can quickly find instability after starting a game 3 times.
I wonder what the actual impact is of an overshoot voltage of 1.59V. what percentile is over/at that voltage? If it's only 50us every second or so that can't really cause heat damage. Unless it rises above clamping voltage like static electricity I don't see how this can degrade the CPU.
@@imqqmi It's what damages the ring, so most definitely. All electronics have operating values, a specification, and if you go out of spec, well the behavior cannot be controlled anymore.
@@trixniisama According to the datasheet Vcore max is 1.72V. max overshoot is 10% for 500us at TDP. An overshoot of 1.59V for 50us is well within that range and I believe his CPU is around 70⁰C under load but not sure he hits TDP limit. Of course if the chips are faulty the specs can't be trusted. They state at TDP and a time limit, suggesting that over voltage main concern is thermal damage. What other failure modes could there be at or below 1.72 + 10% is 1.892V?
@@imqqmi I mean, what do you want me to tell you? Concerns about too much voltages have been lasting for years. The fact that Intel does not actually have a proper specification is what led to the recent motherboard debacle. Now they've published guidelines but anyone knows that pushing 1.5V is already insane, let alone 1.7V! This "specification" was delusional to begin with. Board partners are trying to reduce voltage and not TDP or thermal limits for a reason. Temperature hasn't been a concern in modern CPUs for a while.
Hi BuildZoid, if you haven't already done so, can you test the MSI 0x129 BIOS to see if it disables the 1.55V limiter when you leave (or get kicked out of) Intel Defaults. MSI tech support refuses to give any guidance on this matter because the people we spoke with either didn't understand the question or didn't know the answer.
I did a quick check and both of the MSI profiles seem to have the 1.55V limit in place. However the unlimited profile disables CEP and drops the AC loadline to 0.5 which with my CPU destabilizes cinebench R15(so that's something to keep in mind if the unlimited profile is unstable for you). But with MSI's BIOS you can still change the ICCMAX and power limits even with the intel defaults enabled.
Thanks! I tried your settings and the temperatures seem to be a bit lower. Will see if it's stable. Before I tried AC/DC LL with 55 and LLC4 which was not 100% stable.
Hi BuildZoid, my interaction is generally very limited in the comments, but I have a concern. I have an MSI Z790 Edge WIFI DDR4 with an Intel I7 13700K, since I built it I have never had temperature problems, game or program crashes, BSOD with synthetic benchmarks, CPU stress, nothing. Should I do the update anyway? However, I have my CPU with certain configurations, such as all P-cores at 5.5ghz and E-cores at 4.3Ghz with 1.27v, LLC lvl 3 and XMP at 3866mhz.
GTG, Thanks BZ. Quick question, is there a reason for Level3 vs say level5? I've tested it on a 13900k, and 13700k now, seems to work well. But all and all, I like this tutorial/settings better than the one a couple years back, when I first setup the board.
Hello. I am using 14900kf and msi z690 force Intel default profile I'v done negative offset to 0.100 and lite load to 3 as you mentioned. I am getting 36600 in cinebench 23, and i think it is rather low. Is it? Should i rise power limit a bit, or do something else? Thank you
sync all cores to whatever all core spec your cpu has (to get rid of the stupid 2 core boost stuff), IA VR 1450~, pl2 253w (pl1 whatever you want, i set mine to 253w aswell), then just undervolt while keeping stability ESPECIALLY in UE5 shader compilations. If you wish to do so you can also mess around with the loadline settings as buildzoid showed in previous vids. Keep in mind that ASUS uses 0.00 for lowest AC/DC and not just 0 like some other boards
Is there a difference between "Offset Mode" and "Adaptive + Offset Mode" in the CPU Core Voltage Mode? As far as I have understood it so far, when you leave the Core Voltage at Auto with Adaptive, it's the same? And if your set a voltage there, it will only apply if the CPU boosts above X (5.7 Ghz?), and the entered voltage value is above (not below) the entry in the internal VID table of the highest VF point entry (also 5.7Ghz?)? There are so many entries on MSI boards for the Vcore setting for Intel that it's really confusing.
Adaptive makes the vid request match the vcore for me, so i think its better. If i just use offset, then i end up with 2 different values. Example for offset is, the vid requests 1.4v and my vcore is 1.2. if i use adaptive they end up being the same number. Not sure if theres anything wrong with using regular offset though.
@@Keaweahe Did you use both times use LLC Auto and AC/DC Auto values? I'll have to check again, but I don't think Adaptive + Offset does automagically set the correct AC/DC loadline for any selected LLC value.
@@Keaweahe I have the same mobo, will have to check later if I can see any difference. Yesterday I painstakingly tried to identify the mOhm values for each level of LLC (disable enough cores to not run into the temp limit, then run Prime95 without AVX and compare the VID request to the VR VOUT value), I really hope the Offset mode didnt skew these measurements and the Adaptive + Offset mode shows the same. 😵💫
Thanks! With latest BIOS, i tried your method on my 13700KF and it works perfectly. Also i went with mode 2 on both Loadline and Lite Load, with better scores on CB23. Is it safe to do? Could you suggest how to achieve a mild OC?
I tamed my 14900k heating issues by 15 degrees or so. I just replaced the stock mobo contact frame, to a 3rd party CPU contact frame. I stuck my 420mm aio, ran cinibench with stock Intel extreme 5.7 with 3 cores being at 6, and stayed steady in the 70's with two cores in the mid 80's. Wish I would have done this 7 months ago. Best 20$ I spent. No more thermal throttling. Yaaay!
We all sincerely appreciate the step by step play by play. So awesome & 100% needed from ur comment. We were all wondering about every detail of info u gave us, thanks!
@@godnamedtay I was pulling my hair out trying to figure how to keep it cool. I had 6 cores hitting over 💯 degree range. I was passed the return due date, so I was stuck with it. I had little hope thinking that the contact cpu frame was gonna work. But I decided to try it anyway. Thank God I did. 😆 it worked. I was hitting 💯 even with the new microcode before changing it. That 7 months sucked.
In the latest alpha version of CoreCycler (v0.10.0.0) I've added an option to do an automated test. Which could be helpful in finding a stable undervolt, as you can define various stress test programs (Prime95, y-cruncher, Linpack, Aida64). For Intel processors, this automated test means that it will automatically adjust the voltage offset when an error has occurred, so it could help you find this (somewhat) stable undervolt setting by e.g. letting it start with an offset value you think might be somewhat stable, like the mentioned -150mv in this video, and then letting it run overnight and see what the outcome is in the morning. It currently does *not* adjust the offset more into the negative, only more towards the "stable" side (i.e. up to an upper limit you define), so it's not a fully automated tool (yet?), but it can help you automate manual tasks. It's also worth noting that this voltage offset is only temporary, and will have been reset to your previous value after a reboot (unless you also enable to resume after an unexpected exit, in which case CoreCycler will try to restart the testing process after the computer hard crashed). And as the name suggests this is still very much an alpha, and I think the only person having tested this with an Intel CPU right now is me, as the tool was originally designed to test your Ryzen Curve Optimizer settings, and so the current user base is also mainly Ryzen owners.
It's not a sustainable model to swear off a motherboard brand because of one lemon model like the guy above me said. There's like 5, you'll run out sooner rather than later lol.
ASROCK intel motherboards are bad(especially budget ones).They were even selling multiple B series motherboards (mainly b660m-itx) with 100w limit as one compatible with 13900k.But their Ryzen motherboards are pretty good.
Is MSI Z690 Tomahawk DDR5 Has this issue of Spiking voltage at startup? It's MY first Ever Build and This Happen to INTEL MY BUILD i7 14700k Z690 MAG TOMAHAWK DDR5 I'm Tracking Voltages using HWINFO64 MAX VCore is 1.43 MAX CORE VIDs 1.51 CInebench 23,200 33000 Multi Cores 10min test I have zero experience with voltage and over or under volting
Is there anyway I can set to ALWAYS go to bios when restarting? I do loose a lot of time because eventhough i hit DEL, sometimes I cannot go to bios, and I have to reset the computer again to go to bios... Waste of time and patience
How much can I actually trust HWINFO VIDs and vcore reading from the motherboard? I have a 14900k and stays mostly at 1.4V even with a super single threaded load like minecraft. According to my Asrock Z690 Taichi BIOS the VID for 6Ghz is 1.404V, which from what I can see is exceedingly low but I do still worry about getting too much voltage
The HWiNFO VIDs are influcened by the DC_LL setting. They should be matched to the acutal VRM loadline, but MSI does not do this, so unless you match them manually they will be wrong.
i cannot thank you enough for making this video. budget boards need more exposure like this. not everybody can afford masters or apexes
They are straight up scams tbh (except if you are an extreme overclocker) regardless of budget very few people have reason to go above the "Master" or "Hero" and equivalent lines. I'm sure buildzoid would say the same thing lol.
@@FreelancerVideoDumpthe Apex type boards used to be cheaper than Hero/Master, hate that you have to pay a big premium now to get better mem OC
"Scams to be honest" Wow have you got some bias let alone misusing the English language. Some people appreciate and understand better vrms more pcie and lan performance others just live in ignorancce and make simplistic if not purile comments like the quote above.
23:45 - I noticed the same. When I was on the edge of stability I had these symptoms:
- Task manager being slow to load (sometimes it took 5s to switch between "processes" and "performance")
- Scrolling through youtube comment section caused stuttering.
- Dragging windows around was choppy.
All these symptoms were gone when I raised the voltage by 30 mV (-120 mV to -90mV).
You may have had WHEA errors, I noticed them popping up and being displayed in HWiNFO with a too aggressive undervolt
@@sp00n My HWInfo always shows 0, so I probably didn't go that far.
@@Savigo. You can also check in your Windows Event Log for WHEA errors. For ease of use, it's best if you set up a Custom View for that. Either google how to, or the short version is to use "By source" and then select "WHEA-Logger" as the "Event source" when setting up the filter for the Custom View.
@@sp00n Interesting, I only have 2 WHEA events (not even errors, just info) from the first day I booted up my PC. And then nothing.
Why am I getting ~2250 pts in Cinebench R15 with i7 14700k. What should be the optimal score for this specific CPU on 5.5Ghz(Default) with intel optimized settings. PL1&pL2 253w
Thank you Buildziod. I've digested about 6 of your vids on this topic now, and have my 13700k running very nicely. I truely appreciate it.
I'm scared to do a bios update I haven't had problems yet but haven't update bios since I got my pc last year
I have a 13700kf
@@IamWeezyHDIf you have stable power, then it should be okay. Me and 2 friends have updated with no problems multiple times.
Applied these settings with 0.150 undervolt on my 13900KF and so far looks amazing! Gained 2K on R23, better temps and voltage goes to maximum 1.2
Love your content my friend, thanks for your teachings.
Edit: after a day or two it gave some BSDOS so I changed to -0.135 to be safe.
Thank you for doing this with an MSI board! It's much appreciated!!
Applied this to my Z690 Unify and an MSI Z790I Edge, of course with some slight variances. Great results on both!
Thanks again!
bz out here doing the Lord's work
Great video and you saved my performance after updating the new 0x12b microcode. Now I’m back to within 200 points on Cinebench. Thank you
Thanks Buildzoid! I was able to undervolt my 13700k on the Pro Z690-A DDR5 with CPU Lite Load Mode 3 and CPU Load Line Calibration Control Mode 3 and Adaptive + Offset of -0.170. Runs smoothly in benchmarks R23 (Score: 30624), R15 (Score: 4570) and all my games. With an offset of -0.180 I had crashes in R15.
My stats in HWinfo while running the benchmarks with the -0.170 offset:
max. Vcore: 1.178, max. Power consumption: 193 W, max. Temperature: 81°C.
If you run prime 95 do you get WHEA errors in hwinfo? I thought my CPU was stable until I did prime 95 and saw WHEA errors there
Did you have to disable CEP to get a proper undervolt?
As a MSI User I want to thank you for your Testing and Insights. You've found out yourself that we are plagued with the limited possibilities like the for some reason not opened IA VR Voltage Limiter. May I ask you why you did leave the undervoltage protection "enabled" does it do nothing like the AVX guardband setting that I hinted you (and was wrong) on the last Video? edit: Wait a minute it IS disabled on your last Screens of the advanced Menue (Minute 38:41), but when you look at around Minute 11:17, undervoltage Protection is "enabled" (both times settings are on "Auto"). MSI things...
Undervolt protection just sets a floor to the voltage settings at boot. You can still set them wherever you like in the BIOS, but once its booted, software can only leave it alone or increase it.
XTU requires UV protection be on (if virtualization is enabled). While I was tuning, I'd set an aggressive UV in the bios, and then bump it up a bit in windows with XTU when trying various raised multipliers, and then go back to the BIOS to commit those final values when I was happy with it.
That voltage spike at start-up is scary. Luckily MSI was setting their AC loadline to 50mOhms at default previously which would have helped curb the voltage up until now, but since I'm pretty sure the new BIOS updates just set AC_LL = DC_LL there's going to be a lot of people who updated their BIOS in the hopes of prolonging their CPU's life only to be slowly frying it every time they turn on their computer. And the scariest part is you have no way of knowing if you're affected or not unless you have an oscilloscope to hook up to your board.
yeah but MSI's 0.5 mOhm ACLL literally causes some CPUs(like mine) to crash.
@@ActuallyHardcoreOverclocking True, but I think you just have a particularly bad sample. And in this specific case exposing bad chips is probably preferable to frying good chips.
@@ActuallyHardcoreOverclocking Why my i7 13700k never hits 5.3ghz in gaming after microcode update?
Now is 4.9 max.My sp score is 92.I have gigabyte mobo,i use intel default
We are waiting asus motherboard optimised video (0x129)
Yes ❤❤❤
Please!!!
Funny enough, for me, the best settings are the ASUS OC optimized ones. Highest voltages ive seen are about 1.31 (including a +0.02V offset). 13600K, 5.4GHz 1-2 core, 5.3Ghz all core load. Im lazy, i just use intel extreme tuning app's with its Speedoptimizer 2.0 turned on at boot. If i do the same on the intel default settings profile, it goes crazy. Pushing 1.3V on all core load CB23, when it should do 1.2V. Results in lower clocks and 100C on some cores. ASUS profile it just works fine, temps 75-80C, all core voltages 1.21V on the same test.
This
Set the VID limit, same as gigabyte
There's yet another bios released for this board, not sure what the real difference is, 8/21 version "7D25v1J" -- doesn't have beta so not sure what exactly the difference is though.
edit, not sure why, but MSI has a really weird naming structure, the earlier "beta" one is listed as version 1J.1, where as the newest (listed on their site as released 8/21) shows in flash as 1J.0. "build date" is literally the same between the two in flash as well (8/15)..
which is stupid to me why they would listed as that. God MSI is really starting to piss me off these days.
Same
Anyway my biggest compliment to you Buildziod. Is i really appreciate you acknowledging what you might have learned from all of this in terms of what you might choose to eval in the future. Who would have thought of this at that time about all this? I think going forward it would be nice to add this analysis (voltage control) to the mix. What is the actual voltage, how is it control, does it make sense et, not just what the board can do in terms of peak, sustained power, but almost like power quality--- smoothness etc.
I do kind of wish they just had a hard voltage limit. I'd just set it to 1.35. I was looking at this over the weekend, and the llc, and lite load is this complex balance. You try to stop the cpu from over shooting or dropping, and stopping it for over responding to that drop.
I basically had to put this in terms of a hose and trying to fill water, and what happens when you plug the hose quickly. and all other kinds of things.
Good video, really helped me tune my 13700K. I'm running the DDR5 Wifi version of this board and installed the newest, non-beta BIOS they pushed on the 21st with microcode 0x129. I dont' have my own probe so I can't monitor the voltage during boot, but I was noticing some pretty high spikes on stock settings, pretty close to, or at 1.5V when running Cinebench R23, and I was thermal throttling. I have a 360m AIO sand it's mounted fine, so it wasn't that. Following your advice, I dropped the CPU Load Line down to Mode 8, anything lower than that and I was seeing performance loss, with a negative offset of 0.07 and now I'm getting better temps and a bit of a performance boost from stock. Bumping up the offset to a -0.08 also results in some performance loss too, but I may have to play with it a tad more. I think I'm dialed in quite nicely now, Cinebench R23 and R15 are only getting me spikes of 1.3V in single and multi core workloads. I've had this CPU since launch and the only issue I had with this PC since I built it at the end of 2022 was the first 4080 I got ended up being a dud and causing me all sorts of problems, but replacing it with a new one resolved everything and it's been great since. This video was very informative and did a good job explaining all the CPU Load Line stuff that I was otherwise getting confused about. Thanks!
I have a 13700kf and I'm afraid to update the bios
Thanks a ton!! Exactly what I was looking for..
Just to add - Someone posted exact impedence values of LLC levels and Lite Load AC_LL/DC_LL values for this motherboard on some forum recently.
Apparently, you wanted to set 0.3ohms, which is why you selected Level 3. But Lite Load Level 3 and LLC level 3 correspond to different values, not AC_LL of 30 and 0.3 ohms respectively
Massive massive thank you. Followed the guide got my 13900Ks stable 1.27v under load while retaining my 6ghz single core boost highest hwinfo voltage 1.31v.
Really happy with this thank you.
Vrr
Great info!! The way the CPU calculates instructions varies from one app to another too. This can be why some games work and others don’t, and even might be why some CPUs have degraded faster than others. All this should be considered when UV’ing and stability testing.
If I don't turn off CEP from Auto[ON] it will lower the score by 50% on default load line and & lite load "Mode 9" so I've to manually turn CEP off.
Another legendary video thanks bro! Looking forward to your asus video!
Intel has a serious yield issue that they try to compensate by loosening binning specs. There was that top performing R-batch, the reason why we went to the stores, but since then bins have become a box of chocolates.
Thank you! This helped my VCore and temperatures on my i7 13700k. When I ran Cinebench R23 at stock with 0x129 microcode, I would get a max of 1.464V and immediately get 100 degrees C. Now with Lite Load 3 and a 0.08V undervolt, I get a max of 1.248V and it immediately goes to 91 degrees C and slowly creeps to 100 degrees. That’s much better and increased its score!
Shouldn't I stick to LLC 6 or 7 because of the higher vdroop? Why u chose mode 3?
Just tried this on my 13900K, got 35,700-ish points in R23, not too bad, no different than stock for the score though, my chip probably sucks, dunno, I was hoping for 37,000 points because I was able to do 39,000 on previous BIOS'. But I do have lower voltages now which is very nice for longevity though. Thanks Buildzoid.
What programs (where multithreading is important) do you use the most?
If I remember correctly with these boards I believe LLC3 is 0.10 mOhms and Lite Load 3 is 0.15 mOhms on the AC_LL so you'd actually want to set your LLC to 4 if you wanted it to match your AC_LL at Lite Load 3. Although I think the Lite Load DC_LL is set to 80 mOhms from modes 12 to 2 so your power readings will probably also be thrown off doing it this way. Personally I just suggest manually setting your AC and DC loadlines rather than using the Lite Load presets.
For my MSI Z790 board, Lite Mode 3 is AC 10 (0.10 mOhm) & DC 110 (1.10 mOhm). Z690 may be different of course.
LLC Mode3 with 0.1 ACLL sees a slight performance regression.
@@ActuallyHardcoreOverclocking Actually it seems like MSI Lite Loads have different impedances depending on whether CEP is enabled or disabled so what I said wouldn't necessarily apply in your config.
How’s about 13/14600k and 13/14700k bios optimization???
Please help us to know why Ring Limit Reasons is always YES even in this video ? Is there any issue with it ? Any way to fix it ? Whether setting ring ratio and ring down bin will solve this issue ? 65w non K CPUs didn't have this issue but K series having it(checked 12400 and 14500 CPUs). Ring may be getting high voltage. Is the initial high voltage causing it ? It will be very helpful if you can add a video about this issue. A lot of people are confused about it and thinks this may be one of the reasons for degradation.
Hello,
I have the same issue, but came to the conclusion that it is just an unfortunatly named hardware limit flag.
The reason it's 99% on yes even in idle is that the ringbus frequency is linked to the maximum P-core frequency minus a 5-600 Mhz offset.
So it constantly "throttles" depending on the P-Core clocks, but that's how it's supposed to work.
Not all these flags indicate a real issue with the system, "IA: Max Turbo limit" for example and the "RING: MAX VR Voltage, ICCMax, PL4" is one of them.
Good job! thanks! I wait for the next video "asus" from you
Would love to see you test a board with the OnSemi controller next.
My i9-14900k runs on a Gigabyte Z790 UD and I'm pretty happy with my current settings. It's fast and cool even when boosting to 6Ghz (running a Liquid Freezer II AIO).
Cinebench R23: 39k, 2.3k single - 85°c max
*My settings*
Intel Default Settings: Extreme
CEP: Enabled
AVX Settings->AVX Optimum: Enabled (this removes the Gigabyte Default 7x AVX2 ratio offset)
Adaptive Vcore Offset: -100mV
CPU Vcore Loadline Calibration: High
IA AC Loadline: 0.45mOhm
IA DC Loadline: 0.45mOhm
IA Voltage Limit: 1400mV
Been running like a charm for over a week now, no stability problems in all BenchMate tests as well as Prime95 or games like Cyberpunk 2077 (been playing for many continous hours lately).
I'm kind of a noob and your videos helped me a LOT while figuring stuff out. I was kinda scared my CPU will degrade and I'm much more comfortable now.
Thank you so much for making them!
i have the same motherboard too
how much tempa do you get?
i just enabled energy efficient & cpu doesnt exceed 5.7ghz
@37:52 for anyone who can't be bothered watching the entire thing.
This lowered by temperature by a lot and raised its performance at the same time. Thanks!
Nice useful video thanks for all the effort and information. Cheers
new video after microcode update?
29:01 I have observed the same thing, but like you I am on Windows 10 without thread director. So I wonder, if Windows 11 can run CBr15 with these settings.
MSI annoyed me with that 'automatic extreme settings' thing too. I have a 13600KF, so thermal throttling was never really an issue, but it did spend a lot of time throttling due to the new 200A current limit in the intel defaults. Undervolting did eliminate that without having to go too far. Then I wanted to raise the clocks from the stock 5.1/3.9 P/E to 5.4/4.3 that I knew it would do at that voltage, but MSI insisted on switching the power profile when I did so. As far as I can tell, the only thing it changes is setting PL1/2 and max current to unlimited and I was able to change it back, but still annoying. With the overclock It still hits the current limit (up to 220A) in all core, but performance is no worse than stock and its better in general use applications that don't throw 20 threads at it (the 3d mark CPU test even showed a small increase in 16 thread)
Why use adaptive+offset instead of just offset? I thought adaptive+offset would only offset voltages at high load, hence a single core workload would still be able to request a high voltage.
I appreciate how this generation of home computer hardware is making it more mainstream to learn to tinker with your PC and never trusting manufacturer default settings out of the box.
Hey guys, can someone clarify something for me? Is having a Vdroop a bad thing? I'm sitting on z790 with 14900k with Intel defaults, and LLC 8 with AC80 DC110, adaptive offset of -0.12 and according to HWinfo64 under full load it sits around 1.20, in gaming avg 1.27v with peaks to 1.33v iirc.
Here Buildzoid used far more flat LLC with lower AC DC which if I understand correctly leads to smaller Vdroop.
On his earlier videos I'm pretty sure he didn't went as far down with those numbers.
What's the benefit of doing this? Thanks in advance
A couple of things here. First off, a lower, droopier, LLC has better voltage regulation and helps with the voltage spikes. The thing you have to worry about is at idle or when it boosts to its 1-2 core max, the voltage requests go through the roof.
The other part is, why did he choose a higher, flatter, LLC and go against what he said in previous videos? Well, back when he made this video, MSI boards didn’t have the option to set a voltage limit. So, if you were using a lower LLC with no limiter, it could potentially degrade your cpu. However, with a higher LLC, voltages come way down. Yes, you’ll get spikes. But even at those peak spikes, they’ll be less than the top voltage requested with a lower LLC.
So, he’s saying to do this if you can’t limit your voltages.
Thank you for this video. My 13900k voltages dropped dramatically following your info here.
In my case on MSI mb if I only selected "intel default profile" I lost 50% points in Cinebench r23 multicore... So I swapped back on MSI profile and do undervolt, but I don't know if I'm doing well or not now.
14:36 how do we configure cinebench to just run a single pass? I hate having to wait 10 minutes everytime i want to benchmark.
In the File drop-down menu, on the upper left corner, is an Advanced Benchmark option. This enables different run times, including single pass benchmarks.
This works on all versions of Cinebench.
Just updated to the bios released on 21 of august,had no prior problems and this seems to work fine as well. Using default intel settings and before that limit to 253W. I had to update the bios when I first received this mb, also MSI pro z670-a ddr4, or the 13900k wouldn't work. Updated it a few weeks ago to the most recent full release (April this year) and now 7d25v1j non beta.
Only thing that's sometimes happens is that disk io freezes for 15-20 seconds, then continues like normal. Could be a qlc flash thing I have as data storage (Samsung 870 qvo 8TB, Samsung 980 pro 1TB OS drive)
I just have one question, why are you not using the same or similar LLC settings like switching frequency etc. that you used in the MSI Pro Z690-A voltage regulation and VRM settings video.
Mmm so.. You are doing your own Curve Optimizer for Intel?...
5.1 to 5.7Ghz all core workloads you get with -130mv That is pretty nice... If it works a hahaha... Then, with the new Curve Shaper of AMD That specific problem of undervolting can be FIX?
you probably already saw many asking, but just wanted to leave another request for an asus bios optmization. 13900k/14900k. im a noob that does not want to mess with settings im not qualified to take risks with
Patiently waiting =) ❤
I had a weird bug on my MSI board, if I switch to the 4096W power settings, it basically massively reduces the performance to like 16,000 R23 points, you can even switch the power profile back to Intel Default Settings and it doesn't register it as the default profile, it stays stuck bugged till you clear CMOS. Beta BIOS' are so much fun!
My under volt was stable for a month or more, using the 0x129 bios and verifying with all the usual stability tests and gaming, until I had to install nVidia drivers. It didn't really crash because it was just on the edge, but installing the driver was unusual slow, to the point I though it was frozen.
Are there any motherboard manufacturers that you do like the power delivery / stock settings on? Who has the most sane stock settings in your mind?
good question
None
Mode 3 on loadline and mode 3 on lite loadline,ends up in 100/1100 in ac/dc but do not trigger CEP, how come
Not too sure what my issue is, i have the same motherboard but DDR5 and if i run Lite Load mode 3 i get blue screens and all my games say "lacking video memory" brand new 14900k with less than 100 hours on it. I did -0.070 offset and that works but im still hitting 1.550v in game. Microcode 0x129 (0x12b isnt available to my motherboard yet)
What can I do to find out if the VRM is actually showing correct voltage is? Do I need a special instrument? Also, after flashing the latest micro code, I was able to set a voltage cap on my MSI B760-Vc WiFi 7 board.
The best way to overclock a GPU on MSI afterburner is to make a custom curve where you use the highest offset at the high frequencies that are needed at heavy workload, and gradually lower and no offset at lower frequencies. With the curve you can also cap the frequency at a level you know isn't stable, but that the GPU might opportunistically shoot into for a split second with erratic loads, but isn't practically useful. Can't believe we still don't have this for CPUs, it would be a huge selling point for a motherboard.
Isn't that *exactly* what AMD's new curve shaper is?
after 4 years stable on i9-10980XE started to BSOD, turned off XMP profile, all good! used OCCT to check memory and was finding errors with XMP on.
Thanks for the video. Got my MSI Z690 down from 75% voltage throttle to 10-15%. Going from .100 to .110 negative completely blows the system up, CPU and GPU bound on Iracing. Maybe I'll try 0.105 next to see...
Yeah i7's! thanks for the shout 😅
Does the voltage spike on restart happen on 13th and 14th gen only on this board or does it also do it when you have a 12th gen cpu in it
Would it be possible that you do a video on undervolting with cep on on asus boards z790? There is no vcore voltage, IA ac and ia DC are in different places, happy I found them, load line has different settings and something with synching vrm voltage with ac DC loadline, an there are level 1-3 loadline and then also load line 1 to 8, I took 5 bacause I counted from your video the options on gigabyte and then used it in number form. but then there is vrm voltage offset, svid voltage ofset, and the v/f curve, but no general v/f offset voltage, only per ratio offset voltage, and there is no dynamic voltage offset and I don't know what to change, I have some crashes but don't know the reason, both with vrm offset and svid offset, and the v/f curve is more confusing, because I don't want to do something wrong, I don't know if I can kill my cpu with undervolting and I don't know how to adjust ratio and tvb oc, and things like that with Intel default, because if I enable the 2 best cores to 60, in tvb it says all cores have a 60 ratio, even with all core, and then I can't fine tune it really, and what does xtu do, can I use xtu for things like that and what does xtu affect? Asus calls things different and there are some things completely missing (generally v/f offset, dynamic offset, vcores) and also I don't know what does what (svid offset and vrm offset and things like that, cpu svid, cpu power request, and so on) and a video on a asus board z790 would be very helpful if you could do that, because there are no explanations manuals for the bios and no guides with "translations" to other board naming schemes, and so I am at a los and a bit scared trying it on my own without a real guide on it.
If you could also probe into the behavior when TVB voltage optimization (enabled by default) is disabled, that would be great.
Even if AC_LL has the same value, there is a difference in the voltage the CPU gets when it is enabled and when it is not.
Can you do a video on the Asus Motherboard please?
Would this apply to other MSI motherboards especially after the most recent update? say MSI Z790? I've tried this just now with a 14700k and yes it's a little slower than it was at factory standard but my voltages are 200 millivolts less during helldivers 2.
I used the exact CPU Liteload and LLC you did here. Along with a negative variable offset. Nothing else.
Some people are spooking me saying a LLC of 3 might be too high and could cause voltage spikes that even my sensors cant pick up (using HWinfo 64). My last CPU was a 4670K so I am a bit overwhelmed by everything going on.
I was running happily at 1.25 volts thereabouts, even upping the power limit PL1=PL2 to 253W as a test. I've lowered it back to 125W to 175W just to keep it running a little less hard. I don't need all the power all the time. Stability so far seems good.
Applies this Prroblem also to an msi Z790 carbon WiFi Mobo? Usually LLC 3 is to high for my understanding so far I was going with 5 or 6, but frying that buddy every time.windows boots is also not what I want...
Looking forward to your asus video. Still a bit confused over LLC and its relation with AC / DC load line.
I miss the simplicity of overclocking in the old days (athlon xp / p3). Tweak the vcore, increase the cpu multipler, pass prime95 and you were golden 😂
I find it interesting that with a -40 mV offset, AC 60, DC 82, and LLC 7, my average core VID and average VR Vout readings are within 3 mV difference, when running R23 multi core. I thought AC != DC would make the readings off.
This with a 14700k and Z790 ACE
Hey so using these settings on my MSI Pro Z690-A Wifi DDR4 with the i9-13900K I get an AC/DC load line of 0.2/0.2
With PL1 253W PL2 253W ICCmax 400a and a -0.080mv under load the VCore stays right around 1.25V and pops up everynow and then to around 1.35.
For the most part it seems like things are mostly stable, temperatures are fine except for 1-2 PCores CONSTANTLY pissing me off at 100C while the rest are chilling at 91-94C under heavy load.
Anyway, I do get some weird slow downs when working on things not at full load but no hard crashing. I'm thinking the load line might be too low would that be a correct assumption? And on the MSI board, going from Mode 3 to Mode 4 etc increases the load line setting right?
The AsRock Z790 boards have a hard limit of 100mv negative offset. Also they don’t have a setting to limit the voltage vid request. Not that I can find anyway.
You can undervolt further by using OC Tweaker/FIVR configuration instead. Set core voltage mode to adaptive, VF offset mode to legacy, VF configuration scope to All-core. Then under core voltage offset you can enter the amount of undervolting. Hope this helps.
I did the 150mv undervolt offset on my 13900K with MSI Z790 Edge WIFI DDR4, R15/R32 Score are now 6123/38409 up from about 5500/36000. The CPU isn't hitting TjMAX with a temp of 103C anymore, it's running at about 1.200v with a max temp of 93C. I think it needs repasting, but I'll be going with a thermal pad soon. Anyway, I'll change the undervolt offset to 140/130/120mv, etc. if I find any instability.
Thanks for the video!
i am using microcode 0x12b and just swapped my 13600kf for this 14900kf and when booting in windows it just resets for some reason black screen.
What if I'm getting 1.4v on my current BIOS (7D91VHB has 0x123 microcode) and averaging around 1.3v in Z790 tomahawk WiFi. And what if I underclock to 5.1Ghz and also apply -0.035v adaptive+offset it doesn't exceed 1.21v. Do I need to update BIOS and worry about voltage issues? In HWinfo AC load line show 0.4/1.1 mOhms as I've not limit anything thing in BIOS. CPU lite load set to auto(Mode 09). Just tell me should I update and tweak a little bit or keep my current settings 5.1Ghz and slight undervolt.
have you tested an i5 13600k to see if it goes to high voltages too?
Lmao, nah bruh. Dude ain’t testing no i5’s, Foh
I have a z690 force wifi and a 13700k right now. I was able to offset my voltage by a decent bit and I don't hit anything above 1.25 on single and 1.2 on multi using the new intel default.
What values does CEP actually compare to decide if it should trigger or not?
From what I understand (excuse any blatant ignorance please) there is lets call it default VID, Which is what is written on the cpu vid table, the AC LL adjusted VID, which is what it actually requests, the VID readout, which is the AC adjusted VID further adjusted by DC LL (?? or original VID djusted by DC LL???) and the vcore.
So... what does CEP compare to what?
Like, if the vcore matches VID in HWinfo does that mean the CEP is happy? or is it going to still do some -magic- weird math and still freak out because oh no mercury is in retrograde today and also vid readout doesn't match vcore divided by AC LL multiplied by the square root of current or whatever tf?
Really sorry if you already covered this and I missed it.
7D25v1J1 is beta bios,
There is 7D25v1J version after beta bios.
Thanks a bunch sir
Is it better to have AC/DC 0.1/1.1 (Lite Load 3) or 0.17/0.17? In some previous video you suggested to have the same AC and DC, but it wasn’t MSI motherboard.
I’ve been using Loadline Calibration 5 with 0.130 offset and AC/DC set to 17/17. In HWinfo I had VID and VCore average almost the same (>0.05 difference). Do u think it’s better to go to LoadLine Level 3 and LLC3 with 0.135 Adaptive offset? It’s also pretty stable by far, but hits 96C with 185 PL1/PL2 when the previous ones max was 89.
I’ve got i7 14700KF and MSI MAG Z790 Tomahawk Wi-Fi DDR5
on the new BIOS Intel defaults, my MSI motherboard must have been doing what yours is doing, because it would hit in excess of 1.55v on windows startup, according to HWinfo64. (don't have an oscilloscope, so I can't know for sure, I guess.) Great video, as always.
Hey Buildzoid, can I ask you a favour?
Can you do a "quick" comparison between the Raptor Lake voltage behaviours and Alder Lake? To demonstrate how even if both are set to the same voltages (and clocks?), Raptor Lake has higher excursions due to the overshoot/over aggressive voltage drop compensation? Preferably a 13600K or some other lower-clocked/power Raptor Lake CPU that's more comparable to Alder Lake? Like a 13700 vs 12900K?
Just enough to see the boot behaviour + single-threaded R15. A default/stock VF curve comparison would be cool too.
Why do people keep asking him these bs questions like he’s gonna do it? Cmon people
@@godnamedtay ? It's not impossible that he does it. Get a grip. He's clearly interested in investigating this sort of thing or he wouldn't do videos on it.
There's nothing BS about asking. You never win if you never roll the dice. It's not like he has public rules about what he does or has stated he will never do a video about something someone has asked about. In reality, he has made many such videos, specifically because people keep asking.
Oddly enough I find RPCS3 when it compiles PPUs at the start of a PS3 game is a great way of testing high frequency stability. If it crashes or logs a WHEA error in windows you are unstable. You can quickly find instability after starting a game 3 times.
I wonder what the actual impact is of an overshoot voltage of 1.59V. what percentile is over/at that voltage? If it's only 50us every second or so that can't really cause heat damage. Unless it rises above clamping voltage like static electricity I don't see how this can degrade the CPU.
You know that heat isn't the only point of failure of electronic devices right? Over-voltage itself can be damaging.
@@trixniisama yeah I know about static voltage degrading barriers/junctions in MOSFETs and capacitors but at such low voltages is that still a thing?
@@imqqmi It's what damages the ring, so most definitely. All electronics have operating values, a specification, and if you go out of spec, well the behavior cannot be controlled anymore.
@@trixniisama According to the datasheet Vcore max is 1.72V. max overshoot is 10% for 500us at TDP. An overshoot of 1.59V for 50us is well within that range and I believe his CPU is around 70⁰C under load but not sure he hits TDP limit. Of course if the chips are faulty the specs can't be trusted.
They state at TDP and a time limit, suggesting that over voltage main concern is thermal damage.
What other failure modes could there be at or below 1.72 + 10% is 1.892V?
@@imqqmi I mean, what do you want me to tell you? Concerns about too much voltages have been lasting for years. The fact that Intel does not actually have a proper specification is what led to the recent motherboard debacle. Now they've published guidelines but anyone knows that pushing 1.5V is already insane, let alone 1.7V! This "specification" was delusional to begin with. Board partners are trying to reduce voltage and not TDP or thermal limits for a reason. Temperature hasn't been a concern in modern CPUs for a while.
Great content, new sub here cheers m8
What does load line calibration mode 3 and cpu lite load mode 3 do? What happens if you change one but not the other?
Hi BuildZoid, if you haven't already done so, can you test the MSI 0x129 BIOS to see if it disables the 1.55V limiter when you leave (or get kicked out of) Intel Defaults. MSI tech support refuses to give any guidance on this matter because the people we spoke with either didn't understand the question or didn't know the answer.
Why don't you test it yourself? Do you have an MSI motherboard and a 14900k?
I did a quick check and both of the MSI profiles seem to have the 1.55V limit in place. However the unlimited profile disables CEP and drops the AC loadline to 0.5 which with my CPU destabilizes cinebench R15(so that's something to keep in mind if the unlimited profile is unstable for you).
But with MSI's BIOS you can still change the ICCMAX and power limits even with the intel defaults enabled.
@@ActuallyHardcoreOverclockingwhen i run r15 or r15 extreme it only loads my e-cores, even if i set it to 32 threads whats the trick?
@@Jolinator Is this on Windows 10 or 11?
@@Jolinator need to use the benchmate version which fixes the scheduling
Legend. Thank you ❤
Thanks! I tried your settings and the temperatures seem to be a bit lower. Will see if it's stable. Before I tried AC/DC LL with 55 and LLC4 which was not 100% stable.
Hi BuildZoid, my interaction is generally very limited in the comments, but I have a concern. I have an MSI Z790 Edge WIFI DDR4 with an Intel I7 13700K, since I built it I have never had temperature problems, game or program crashes, BSOD with synthetic benchmarks, CPU stress, nothing.
Should I do the update anyway? However, I have my CPU with certain configurations, such as all P-cores at 5.5ghz and E-cores at 4.3Ghz with 1.27v, LLC lvl 3 and XMP at 3866mhz.
Wait, i can just undervolt my cpu using the IA VR voltage limit on gigabyte board, do i need to set anything else?
GTG, Thanks BZ. Quick question, is there a reason for Level3 vs say level5? I've tested it on a 13900k, and 13700k now, seems to work well. But all and all, I like this tutorial/settings better than the one a couple years back, when I first setup the board.
Hello. I am using 14900kf and msi z690 force
Intel default profile
I'v done negative offset to 0.100 and lite load to 3 as you mentioned.
I am getting 36600 in cinebench 23, and i think it is rather low. Is it?
Should i rise power limit a bit, or do something else? Thank you
Why my mouse start lagging, when i switch w_pump speed from full speed to standard mode?
We need asus motherboard setting with i9 13900k/ i9 13900kf with 0x129 microcode
Should be the same, lol. 14th gen is basically a new stepping, within silicon variance of 13th gen.
sync all cores to whatever all core spec your cpu has (to get rid of the stupid 2 core boost stuff), IA VR 1450~, pl2 253w (pl1 whatever you want, i set mine to 253w aswell), then just undervolt while keeping stability ESPECIALLY in UE5 shader compilations.
If you wish to do so you can also mess around with the loadline settings as buildzoid showed in previous vids.
Keep in mind that ASUS uses 0.00 for lowest AC/DC and not just 0 like some other boards
Is there a difference between "Offset Mode" and "Adaptive + Offset Mode" in the CPU Core Voltage Mode? As far as I have understood it so far, when you leave the Core Voltage at Auto with Adaptive, it's the same? And if your set a voltage there, it will only apply if the CPU boosts above X (5.7 Ghz?), and the entered voltage value is above (not below) the entry in the internal VID table of the highest VF point entry (also 5.7Ghz?)?
There are so many entries on MSI boards for the Vcore setting for Intel that it's really confusing.
Adaptive makes the vid request match the vcore for me, so i think its better. If i just use offset, then i end up with 2 different values. Example for offset is, the vid requests 1.4v and my vcore is 1.2. if i use adaptive they end up being the same number. Not sure if theres anything wrong with using regular offset though.
@@Keaweahe Did you use both times use LLC Auto and AC/DC Auto values? I'll have to check again, but I don't think Adaptive + Offset does automagically set the correct AC/DC loadline for any selected LLC value.
@@sp00n yes i did.
@@sp00n msi z790 carbon wifi
@@Keaweahe I have the same mobo, will have to check later if I can see any difference.
Yesterday I painstakingly tried to identify the mOhm values for each level of LLC (disable enough cores to not run into the temp limit, then run Prime95 without AVX and compare the VID request to the VR VOUT value), I really hope the Offset mode didnt skew these measurements and the Adaptive + Offset mode shows the same. 😵💫
Thanks! With latest BIOS, i tried your method on my 13700KF and it works perfectly. Also i went with mode 2 on both Loadline and Lite Load, with better scores on CB23. Is it safe to do?
Could you suggest how to achieve a mild OC?
What lite load and other changes in BIOS would you set for i7-13700K cpu?
Is there any advantage to setting that voltage offset in bios versus in XTU?
I tamed my 14900k heating issues by 15 degrees or so. I just replaced the stock mobo contact frame, to a 3rd party CPU contact frame. I stuck my 420mm aio, ran cinibench with stock Intel extreme 5.7 with 3 cores being at 6, and stayed steady in the 70's with two cores in the mid 80's. Wish I would have done this 7 months ago. Best 20$ I spent. No more thermal throttling. Yaaay!
We all sincerely appreciate the step by step play by play. So awesome & 100% needed from ur comment. We were all wondering about every detail of info u gave us, thanks!
@@godnamedtay I was pulling my hair out trying to figure how to keep it cool. I had 6 cores hitting over 💯 degree range. I was passed the return due date, so I was stuck with it. I had little hope thinking that the contact cpu frame was gonna work. But I decided to try it anyway. Thank God I did. 😆 it worked. I was hitting 💯 even with the new microcode before changing it. That 7 months sucked.
In the latest alpha version of CoreCycler (v0.10.0.0) I've added an option to do an automated test. Which could be helpful in finding a stable undervolt, as you can define various stress test programs (Prime95, y-cruncher, Linpack, Aida64).
For Intel processors, this automated test means that it will automatically adjust the voltage offset when an error has occurred, so it could help you find this (somewhat) stable undervolt setting by e.g. letting it start with an offset value you think might be somewhat stable, like the mentioned -150mv in this video, and then letting it run overnight and see what the outcome is in the morning. It currently does *not* adjust the offset more into the negative, only more towards the "stable" side (i.e. up to an upper limit you define), so it's not a fully automated tool (yet?), but it can help you automate manual tasks.
It's also worth noting that this voltage offset is only temporary, and will have been reset to your previous value after a reboot (unless you also enable to resume after an unexpected exit, in which case CoreCycler will try to restart the testing process after the computer hard crashed).
And as the name suggests this is still very much an alpha, and I think the only person having tested this with an Intel CPU right now is me, as the tool was originally designed to test your Ryzen Curve Optimizer settings, and so the current user base is also mainly Ryzen owners.
Do an ASROCK board. Im done with the other 3 brands
don't have one
Gee, what will you do when they mess up. Calculator? 😂
Just curious, do you stop buying a brand of car because they put a bad model out one year?
It's not a sustainable model to swear off a motherboard brand because of one lemon model like the guy above me said. There's like 5, you'll run out sooner rather than later lol.
ASROCK intel motherboards are bad(especially budget ones).They were even selling multiple B series motherboards (mainly b660m-itx) with 100w limit as one compatible with 13900k.But their Ryzen motherboards are pretty good.
yes. the z690 taichi went from $600 to $200
Have you tried Mode 7 + prime95 - i dont think it will reboot, it might crash but it wont reboot.
Is MSI Z690 Tomahawk DDR5 Has this issue of Spiking voltage at startup?
It's MY first Ever Build and This Happen to INTEL
MY BUILD
i7 14700k
Z690 MAG TOMAHAWK DDR5
I'm Tracking Voltages using HWINFO64
MAX VCore is 1.43
MAX CORE VIDs 1.51
CInebench 23,200
33000 Multi Cores 10min test
I have zero experience with voltage and over or under volting
Is there anyway I can set to ALWAYS go to bios when restarting? I do loose a lot of time because eventhough i hit DEL, sometimes I cannot go to bios, and I have to reset the computer again to go to bios... Waste of time and patience
Hold shift when you click restart in the start menu and it takes you to bios automatically
- 0.070 offset im getting 1.18 vcore but then my PC crashes in the middle of cinebench
did you do a mode 3 cpu loadline calibration and a cpu lite load mode 3 as well?
@@rleekc yes
How much can I actually trust HWINFO VIDs and vcore reading from the motherboard?
I have a 14900k and stays mostly at 1.4V even with a super single threaded load like minecraft.
According to my Asrock Z690 Taichi BIOS the VID for 6Ghz is 1.404V, which from what I can see is exceedingly low but I do still worry about getting too much voltage
The HWiNFO VIDs are influcened by the DC_LL setting. They should be matched to the acutal VRM loadline, but MSI does not do this, so unless you match them manually they will be wrong.