29:35 optimization settings 55:11 when to RMA CPU 35:55 undervolting settings - recommended for performance and even safer voltages but optional bc of instability risk: 38:00; also listen to 42:40 regarding undervolting! Like he said, every CPU is different. The process of finding out how much you can undervolt is the following: - Download Hardware Info 64 and BenchMate. - Make sure you set the settings BZ listed (29:35, LLC6 is the most important even if you change other stuff). - Set a negative VID offset, start with -0.07 f.a. (45:05). - Boot, start BenchMate, start Cinebench, set a custom number of threads: top left corner click on "Preferences" to change this. - Run multiple benchmark runs with different numbers of threads active (f.a. 1, 4, 8, 12, 16,...) -> this ensures that different frequencies are stable. If fewer threads are active, the cores that are active have a greater "power budget" before the power-limit is reached. Therefore they can clock higher. You want to make sure that the highest clock is stable at the lowest voltage. - If the CPU is stable, increase the negative offset value. I'd recommend increasing by 0.02 increments. Safes time. If you crash, lower the negative offset value by 0.01 and test again. After you found your stable offset (f.a. -0.11 like BZ has in this video), I'd personally lower it by at least 0.01 again (to -0.100) as an additional safety margin. Especially because cinebench is not a real stability test. YCruncher would be better f.a. ... or just google "how to undervolt"
Im on a Z690-A and my LLC only goes up to 7, his goes up to 8, should I still use 6? My AC/DC in HWInfo with his Optimized settings is 1.7/1.7, his ends up being 0.49/0.49 Edit turns out the "Intel Fail Safe" SVID which is default on F5, caused it, using auto solved it. However my CPU only goes up to 4.8p, 3,7e ghz now. Unrelated: For the Global SVID do you not have to lower Cache aswell to match? Everyone else seem to say they should be equal. With the undervolt, VID is about 0.05 lower than Vcore, is this correct? I get 1.288 allcore R23, vid is requesting around 1.23, this is with global adaptive -0.07 Actually changing global svid to -0.09 doesnt seem to do anything at all its still the same 1.288 vcore, changing to -0.1 also makes no difference to vcore.
Right? I killed my 13900KF with a bad delid just as all of this stuff about VID blew up. bought, delidded, and installed a 14900KF a couple of weeks ago, and am only now getting around to tuning. Good guy Buildzoid has a vid for me to watch about just that thing :P
Because even though he does ramble , he gets annoyed about the same stuff we do. Actually shows configurations and what they do and how he got there. Does not hold back when criticizing something just like how i usually think about these things that annoy me.
Awesome. Another reason to update-I’ve seen several 1.65+ VID requests on pre-0x125 with a 14700k on undervolted setups. This is when the ETVB formulas were messed up badly. If not constantly monitoring for it, I never would have caught it. 0x129 also includes that fix as well.
Black Myth Wukong Benchmark Tool is free on steam if you really want to stress via the UE5 shader compilation. Personally, I think that’s the best way to tell if you have instability or degradation beyond R15. UE5 after all was the first indicator that 13th/14th gen had problems to begin with.
Thanks buildzoid. I finally took the courage to put intel default back on and play with load-lines after this. Got LLC 6 stable like you with an undervolt and ac loadline set to 0.4 with CEP enabled. Performance now near same within margin of 5-10% of asus oc profile
Thanks for the video buildzoid. Just wanted to report that ASUS has done a poor job supporting laptops from as late as 2023. My laptop with a 13980HX (essentially a 13900K with a wattage cap) latest bios revision from this past month is using microcode 0x123, and can easily voltage spike between 1.55V - 1.6V
With an ASUS Tuf Z790 and 13700k I get slightly better numbers under ASUS OC than yours under intel settings. With ASUS they use LLC 3 with 0.4/1.1 ACDC. Under this setting I get around 29600 R23 score and 85 C max temps. But with Intel settings at LLC 6 0.49/0.49 ACDC and a - 0.100 v offset I was getting 30000 R23 score but temps were 95 C. Both options were pulling around the same wattage but the ASUS one has very slightly higher vcore voltages. Using IA VR limit of 1450 in both options too. Not sure which would be the better option to go with but for now I'm sticking with the lower temps.
Intel fixed the overvolting issue, so the difference between stock (LLC 3 + 1.1 Ohms) and LLC 6 + 0.49 ohms is very minor now. Of course if you do unmatched LLC and AC/DC you have to disable CEP which may reduce stability. Otherwise you can still do the SVID override (adaptive voltage offset) with stock LLC and AC/DC
Question to clear up some confusion: Most Asus "guides" online (for AC undervolting) suggest set LLC = 4, AC = 0.40 and reduce AC until unstable, if you can reach 0.01 AC, reduce LLC = 3, repeat with AC til unstable. This obviously triggers CEP in most configurations since LLC 3 is 1.1 mohms but having an AC that is way lower, makes CEP unhappy. So lets say the system is set to LLC = 3 / AC = 0.20 do you just set DC to also 0.20 or just go with a higher LLC (like 6 as shown in the video, (just as an example)). I think that is where (for Asus) lots of the confusion is coming from for undervolting because everyone says a more droopy setup is good and don't use a high LLC (meaning above 4).
That is why I am confused about Buildzoid's chip. He says his isn't stable at the mobo default settings, but pretty much everyone I see have chips that not only are stable at the 0.4ohm, usually they can go much lower. My 13700K can go all the way down to 0.01ohm on the AC even and still be stable. I wouldn't think all the major motherboard makers would set their default settings to voltages that weren't at least stable, in the past if anything they have errored on the side of giving it too much voltage. I really just wonder if Buildzoid has a broken CPU from the factory. I tried to get my same undervolt with the Intel Default Settings just using a voltage offset, but at the certain point, it just doesn't go any lower unless I start messing with the cache offset even though my uncore VID is already way lower than it was with my LLC undervolt.
@@ZoneXV I thought my 14900k was stable just AC LL undervolting down to 0.1, passed everything til I ran OCCT, Insta core crashes/WHEA errors. I plan to test out his settings here. All I really care about is gaming performance and temps during that gaming. The OP does have a good question though
@@ZoneXV Regarding your last paragraph: The reason it won't go any lower is because not only the cache vid but also the e-core vid can limit the VCORE at a certain point. The VID offset is only for the p-cores. And regarding your first paragraph: luckily BZ chip is one of the lowest end chips on the curve. That way we get first hand experience and guidance of the worst case! His chip isn't broken, just very bad.
Hi Mr, please create new version according to new bios update released 2 days ago. I think there will be 2-3% lower performance :( I got same CPU and MOBO as on this video, and before new bios 1703 I had 39.6k points with similar settings as on movie, but after update I set same settings and have "only" 39k points... New BIOS version have new microcode 0x12B and looks like it lower performance even more...
damn i was here to see how SVID behavior effect cpu , there is a "intel fail safe" vs "typical" scenario war rn but still got some other ideas . thanks for the video
@@unhappymeal5652 I just tested this on my Z690 Board, and if i leave it on "Intel Fail Safe" the AC/DC LL is set to 1.100 mOhm not matter what i set my LLC Level to (AC/DC LL sync is enabled). So the SVID Behavior on ASUS Boards does indeed influence what the other SVID/Load-Line related settings actually do. If i set it to "Auto" and the LLC to Level 6, i also get 0.49 mOhm AC/DC LL
Thanks. I am just upgrading from 13900k to 14900ks and switching from Z790 Aorus Master to Z790 Apex Encore for that 7800 MHz stable RAM and 1% FPS boost :) (no cap, really, just for fun) Great informative video, as usual. Cheers!
IA VR Voltage limits are for peace of mind, although bere in mind if your CPU is a potato setting 1.4v will prevent your CPU even maintaining 5.7Ghz even in gaming. You’ll suddenly be fluctuating between 5.3Ghz to 5.7Ghz. An undervolt addresses this since it moves you up the VID boost table. Although I’m still unsure about level 6 LLC, since you’re then moving into AIB AC/DC ranges, and technically speaking the existence of CEP prevents the instability you would otherwise get. Think I’m getting that right….
10:00 that's not true. It is also the ICC_MAX for the 13900K and 14900K in the Extreme profile, just like the 13900KS and 14900KS. The only difference between the K and KS in Intel Default Settings spec is that KS has PL2 and PL4 of 320W and 550W respectively, while the K has PL2 and PL4 of 253W and 380W. This is all just assuming that you use the Extreme profile (which is "Recommended" while the "Performance" profile is "Default" according to Intel Default Settings spec, on all of the mentioned CPUs). So you should be using the Extreme profile on the 14900K, which would set the ICC_MAX to 400A.
Buildzoid, could you investigate microcode update 10E? Tech Yes City put out a video in August that just seems like nonsense to me but suggests that he had lower voltages on that patch and that newer microcodes have been cranking them up, particularly VDD and IMC values that he believes should be at 1.3V. Pretty sure these are just memory-related though...
I was working on finalizing my undervolt and after watching this video decided to try to run P95 Small FFT with only 8 cores and no HT to really push the P-cores and check for instablity. I received the out of memory error and noticed that MSI AB crashed. This does not happen with 24-cores. Current settings when MSI AB crashed: Asus Z790 Dark Hero MB Vcore to 1.4v (IA VR LIMIT Bios setting), set 253w/400A limits, use LLC#6 with ACLL@0.49 and DCLL@0.54(calibrated), enable CEP, use V/F curve 6,7,8,10 undervolt -0.13. I reduced my undervolt in increments of -0.05 and each step MSI AB would still crash within Test 2 of the Small FFT run. When i finally hit -0.11v MSI AB no longer crashes and I am able to get through many more runs of P95(3 cycles of 6 tests each). Does the out of memory error only happen when the chip has degraded or can it also happen with a too severe undervolt? In other words, is the out of memory error simplly an i/o issue with too little Vcore, or does this only happen once the chip has degraded and more voltage allows you to hide it for a while? Prior to this I was R23/R15 stable and have not received any any WHEA errors (weeks), but had not yet tried P95. I am curious if i need to be concerned.
So to sum up, i want to change load line calibration to 6 and then change global svid to offset - and set for example as you did 0.11 And thats it? The IA VR Voltage Limit i do not know what works best for my CPU maybe just try 1400 as youre example and see how it goes? With further inspection i did set load line calibration to 6, set global svid to adaptive and - but the voltage offset was greyed out. Couldnt be changed. So i guess there is a setting that isnt turned on for it to be changed? I did not have "offset" as setting in drop down. Only auto, adaptive and manual. So our bioses for our motherboards look different. Gained 400 points in cinebench r23 with those settings. What im i missing and what should i try? Using Z790, 14700K and 1503 bios with 0x129 microcode, intel default settings, performance profile. XMP I enabled. Without the undervolt and that tweaks i get about 33.000 points in cinebench r23
I've been overclocking/undervolting for about 6 weeks now and this video just really made so many things make sense for me. I've watched it now 4x's and some things just clicked. One thing that still confuses me is Intel's pl1/pl2. Is this intended for people with minimal cooling abilities wether lack of fans or small AIO? Or is this a hard limit no matter the system? Anyways, thanks so much for this video.
How much can I apply this to the Asus B760-A? First thing I noticed is that "sync ACDC loadline to VRM" doesn't apply, it stays at 1.1 mOms regardless of VRM Loadline. Secondly, there is no level 8 for VRM LL, does this mean Level 6 may not be 0.49 mOhms ACDC for my board? Lastly, I can't do adaptive SVID offset, it's either Auto or Manual. V/F points are also fixed. The only thing I can touch is actual VRM core voltage, which means I'm unable to undervolt without CEP kicking in, and that is forced enabled on my B series board. So I'm totally out of any way to undervolt?
This is great as I run a 13900KS on this board. Just wondering if your thought process would do anything differently for the 13900KS off the new intel defaults than what you show/describe here.
At least Asus shows the VID table when you want to set a VF point offset. MSI only shows the fields to enter the offset, with no VID values whatsoever. 😒 I haven't found a way yet to display the VID table on an MSI motherboard, the info seems to be so low-level that it's only available in the BIOS and not accessible by regular software.
If i limited my Vcore to 1.4v (IA VR LIMIT Bios setting), set 253w/400A limits, use LLC#6 with ACLL@0.49 and DCLL@0.54(calibrated), enable CEP, use V/F curve 6,7,8,10 undervolt -0.13. Is there any danger to using ASUS OC Profile over the Intel one? I feel like the VRM limit being 1.4v supersedes the Intel 1.55v limit from the microcode. is this correct?
I have also discovered another odd thing that using the "Intel Default Settings" profile does that I'm not sure is specific to ASUS or not. Using the Intel settings, it forces a really aggressive level of C-States. Aggressive enough that I'm having certain p-cores drop down to 800mhz during gaming. That is either the balanced or high performance power plans and c-states set to auto in the BIOS.
Just go to bios and look for C1E and set to disabled. Leave regular C states enabled. If you want clockspeeds to drop during balanced power plan, set speedshift to Disabled and EIST (Enhanced Intel Speedstep) to Enabled. This way balanced power plan will function properly. 0x129 enables C1E by default, this causes the random 800mhz drops in my experience.
Hi, what do you think about my undervolting on a 14900kf. I have LLC4 with Intel Default Extreme with -0.07V on Global Svid Voltage on an Asus Rog Strix Z90-F Gaming Wifi. I also have TVB (on) and TVB Voltage (auto = on). In most games I have a Vcore of 1.38V (+- 0.03V) with 5.7ghz on the P-Cores and 4.4 on the E-Cores. In Cinebench R23 I have 38,500 points but in Time Spy a CPU score of 20,500 points (possibly for G-Sync. The temperatures are 50-65 degrees depending on the game. Microcode is 0x12B
Hey @buildzoid, I'm facing an issue with constant current/EDP throttling ever since my CPU arrived last month. No matter what I do, the throttling persists. I tried lowering my PL1/PL2 limit below 220W and increasing ICCMax to 307A, even up to 370A or 400A, but it's still throttling. I followed your 0x129 microcode video and set LLC to level 5, which lowered the resistance to 0.730/0.730 mOhm from 1/1 mOhm at level 4. However, after applying these settings, the throttling continues, even when idle and while playing CPU-heavy games. I feel stuck and frustrated. As you said "i want to use my pc not fix it". I've read tons of articles, but most of it goes over my head (not very tech-savvy). Here's my setup: i7 13700K (Cooler Master 360 Atmos AIO) Asus ROG Strix Z790-H Gaming WiFi ROG Strix 750W PSU Any help or insight into why this is happening would be really appreciated. Thanks for your time, and sorry for the long message!
Are you setting your higher iccmax in xtu or the bios? I had a weird issue with xtu showing my pl1/2 and iccmax were raised but would never go above default valuesm I had to go to the bios and set them in the bios to actually get different values for some reason.
@@xvjaqen how hard are you undervolting? Not sure how much it applies to computer cpu's but basic electrical theory says when doing the same amount of work, amperage increases if voltage decreased and vice versa. So if you have a giant undervolt it needs more amps to do the same amount of work. If you don't have a big undervolt then it's something I'll have to keep thinking about or wait for someone smarter to comment.
Whats your opinion on Sync all Cores to prevent the 2 p-cores from running away from you? My problem with the new microcode is that my idling voltages and temps are stupid high.
I just set a max ratio to 57 on my 14900k, as 6Ghz is the culprit for silly spikes to begin with. And let’s be honest, it’s pretty useless and you’re not going to miss it. There was no need for 6Ghz boosts apart from Intel claiming the “fastest gaming cpu crown”. Then set an IA VR Voltage to 1.42, and a 0.070 undervolt and forget about it imo.
Can anyone go into more detail on the accuracy of the oscilloscope readings. I have V Latch and VR VOUT monitoring on this board the Asus Apex Encore and wondered how close this figure was to the reading from the scope? I believe BZ mentioned that the reading could be off by several millivolts under full load and close to actual at idle. Not sure if that would mean reading more or less volts. Any comments? Thanks!
What I did was a delid with an EK Nucleus Direct Die aio, my chip is not the greatest. The latest bios/microcode locked my two preferred cores to 57, then to get the 5.7ghz consistent on all P cores I set my IA VR Maximum Voltage to 1420, Vcore hovers around 1.35v and my VID request do not exceed 1.35-138Xv per core. Now if I want to fully micromanage my system, I cannot use the new baseline bios with microcode, I have to use the bios before the first baseline and an older microcode version, otherwise tuning/fully micromanaging the cpu becomes a painful nightmare. 6ghz+ requires too much voltage they should have just limited their chips to 5.7-5.8ghz maybe made a 12-16P core no E core chip and called it a day...
Is it just me or loadline undervolting works better than this? I have a 13900k on an Asus Prime Z790P and an Arctic Liquid Freezer II 360. If I undervolt in this way I still get thermal throttling in Cinebench R23. If I undervolt disabling CEP and setting LLC level 6, AC 0.19 and DC 0.5 I no longer thermal throttle. EDIT: I use LLC5 not 6
Are you unstable below 0.19? Are you testing with anything aside from R23? So you're saying by using his settings including a - 110mv offset you thermal throttle in R23? I would also say try his exact settings but disable IA CEP and see what happens also, be curious to know. Maybe even try lowering the offset slightly more?
@@wardeagle1015 Yes I'm unstable at AC 0.17, cannot pass R15, so I set it at 0.19 to have a buffer. I test with R23, R15 and R15 extreme. Using his settings including the offset I thermal throttle, yes. An offset higher than -120mV is not stable for me, and I still thermal throttle even at -120mV. So unfortunately his way of setting things don't work for my CPU, at least if I want to get the most out of the undervolt.
@@IlCode85 That's so surprising actually I would think it should yield similar results at worst. I wonder if there's something else changing that's not been made aware of. Not insinuating there is but still just surprises me. Whats your SP and Cooler score? Just out of curiosity
Also have you tried something more droopy just out of curiosity like LLC4 and then just drop AC LL as low as you can, turn CEP off and test that. Be interested to see if the results differ
@@wardeagle1015 Could it be that undervolting with AC and DC loadlines has a different effect depending on the load compared to undervolting with an offset? Just my thought, I'm not an expert at all. Unfortunately I cannot check my SP and cooler score as my motherboard is just a Prime Z790P, not one of those high end ones.
Im on a Z690-A and my LLC only goes up to 7, his goes up to 8, should I still use 6? My AC/DC in HWInfo with his Optimized settings is 1.7/1.7, his ends up being 0.49/0.49 Edit turns out the "Intel Fail Safe" SVID which is default on F5, caused it, using auto solved it. However my CPU only goes up to 4.8p, 3,7e ghz now. Unrelated: For the Global SVID do you not have to lower Cache aswell to match? Everyone else seem to say they should be equal. With the undervolt, VID is about 0.05 lower than Vcore, is this correct? I get 1.288 allcore R23, vid is requesting around 1.23, this is with global adaptive -0.07 Actually changing global svid to -0.09 doesnt seem to do anything at all its still the same 1.288 vcore.
Asus board is very easy to optimise on but unfortunately in the intel profile mode we cant cap our p cores only in the asus oc profile mode we can. I was doing that for eztra safety to not run at higher clock frequency for an automatic undervolt on top on the slight adaptive undervolt for the svids at -0.01
Please tell me: how to make an undervolt when you have the 13th generation and a ASUS B760 motherboard has no function "Adaptive Mode" in Global Core SVID Voltage setting?
Thank you very much for this great explanation, I'm glad to see that my board handles the new microcode correctly. I was still getting apps crashing with 0x0000005 and RMA'd my 14900KS, the replacement will arrive tomorrow. The board is already updated and I'm thinking about if I should undervolt using the method shown in the Video or if I just leave everything on factory defaults - I've chosen the 14900KS because I don't wanted to bother with overclocking at all but still wanted get the best possible performance out of the box, but now I'm thinking; The undervolt seems to be even safer than the Intel defaults AND having a higher score. Are you using a contact frame or the Intel ILM? Question to all: Which contact frame would you recommend?
If I'm not mistaken, undervolting can only damage your sanity, if your app or computer crashes when you don't want it to. Other than that, it's trading a bit of time in setting it up and testing that it's stable, for better everything. That is, the crashing from undervolting, from what I know, don't degrade the CPU. So in a way, it would make it even safer. That being said, I'm not an overclocker (well, probably I'd be, if I wasn't on a laptop) so, do check with others too.
I manually set all the default power limits to intel extreme on my 14900ks with the Asus OC profile but with some safety nets turned back on . On a z790 Maximus Dark Hero. Set Svid behavior to auto, CPU LLC Lvl at 5 and matched the internal CPU power management to intel settings. Core usage to Auto. I have various other small settings changed for stability and gaming performance, like turning off PCIE Aspm and other random stuff. But the main thing is I use to fine tune the voltage curve is actually their AI features tab and I set my cooler training to Manual and use that to fine tune the voltage curve. My peak performance is at setting cooler score to 178 and leaving it alone.
@@schneezockt about 32c Idle (same as liquid temps). About 55-60c running something cpu heavy like Read Dead 2 or Spiderman Mile Morales, 40c running Black Myth Wukong. I'm using an Asus Ryujin III 360 Aio with some Noctua 3000rpm IPPC fans with some pro-active fan curves as well.
My 14900K (and Asus Prime Z790) is only running at 5000 as it's hitting around 253W and therefore will not send more voltage to the CPU (my guess). With CPU under full load, I'm hitting an average of 1.14V, with an -0.01V offset. When I try to increase PL1 to 300W in Bios, it's still showing PL1 max of 253W. When I then set it in Intel ETU, it is changing the PL1 max to 300W, but only drawing up to 264W. Are there any other bottleneck in settings I have to change?
Running heavy computation on 14900k can tell that these settings are unstable, not even default constantly get segmentation fault after updating to 129 microcode, doing heavy computation on Intel become a nightmare (which wasn’t before update, still can’t find stable settings or CPU degraded to the point where it can’t be stable anymore)
0x12B was just the past two Microcodes rolled into one, basically. I'm sure there were a few more minor underlying changes but all in all shouldn't be much different than 129. All of my benchmarks and undervolts still work and bench the exact same. Just my personal experience, though
Idk i have a 14700 and a b760 i cant set an offset on vid only on actual vrm voltage, so i have to turn cep off to get my cores above 4.6ghz what do you think?
So my cpu was "fine" except that it had some mayor problems with training tertiary ram timings on all my different ram dies, after the update my screen goes black in certain games, game crashes and 5 beeps, any advise?
58:00 I have a question for you, is this bellow safe ? I'm on the version before that but I see no reason to update, the difference is I do not use the offset, I use static voltage and I have it set to 1.25V, my load line is 0.500 and my load calibration is on level 4, my real vid is 1.220V on idle. My vid request on load are max 1.340V and the "real" vid drops to 1.190V ( the request go to 1.340V ). My CPU (13700KF) stays at 5.4Ghz all P cores, board is Asus Z790-P WIFI, tested with OCCT extreme stability test, also test with Ezbench (unreal 5) and keeps the CPU ratio of 54 fully stable (1.195V/1.197V), temp at max 72C, idle temps are around 34C to 36C I actually mess with the V/F and the max I allow is 1.35V for 54, have my IA VR set to 140, also I have the Intel default settings set. Now if I understand the video right you also have a max request around 1.35V ish , but I see on the C1 vid max 1.45V ish , now if my voltage is static does it actually reach 1.4V ????
@@ActuallyHardcoreOverclocking Thank you for the reply, 😉 Normally I do this to all my computers before this but because of this new "bug" I was not sure if even doing this might have problems with this new "bug"
I wonder how this applys to msi Z790ACE , the msi options are a bit diferent . I understand what needs to be done and and will try to aply it on my ACE +14900K
How is my Corsair AIO 360mm cooler not able* to control temps (thermal throttles during r23) for 13700k but your 240mm AIO is keeping everything on a 14900k under 90C? Even on default settings on the same microcode??? Edit: Typo
The 700Ks are both a little bit worse binned and no lower power than the 900s. Basically the added E cores don't make them harder to cool, but they do take a bit of power budget away from the P cores, meaning a 900 ends up with lower temps.
Ecores were a mistake. Take as-much or more power, for less performance, and they hog power away from p cores. Glad this ring bus fudge up is making it clear how poorly planned Intel's last 2 generations have been.
@@RotaryJunkieOh I see, thank you for the quick response. I can't even seem to get a 5.3 Ghz overclock let alone a 5.0 Ghz to stay under thermal throttles. Not sure if my chip just needs to be RMA or a new cooler is needed. Temps are 40s-50c on no load, except for the first load of the day.
@@ActuallyHardcoreOverclockingThank you for your videos Buildzoid! The fans and pump are maxed. Let me know if you'd like any other info regarding the setup just in case it's something resolvable outside of RMA. I've got an Asus Z790-p prime wifi motherboard, a 13700k, Corsair H150i 360mm AIO water-cooler. Using a 0.1 adaptive offset as opposed to the 0.11 used in the video. Edit: Added specs for setup.
With all the various brakes now being applied by Intel and/or the mobo makers, I wonder if/how much the average user will notice in real life? I mean yes they might see say 170FPS instead of 180FPS for instance, or an app taking 0.1 secs longer, but in actual use, will they notice? Yes, we should have had the product as described, but we didn't. I'll settle for settings that give me stability and longivity, over a 5 - 10% gain in performance that is doubtful I could realistically notice.
Hey guys. Are we always using balanced power plan in windows for 13700k i7? Is there any advantage using high performance or e.g. Bitsum Highest performance? I am only gaming on my pc. I already did all the stuff (A/C load line settings, newest bios) to adress the degrading issue. With balanced plan all these settings I am getting 30500 cinebench 23 score. And is there an advantage using parkcontrol to unpark cores? Or just enable c-states and let windows and cpu handle it? Ty for answers
10:56 BZ says that the Vmax voltage is higher now with LLC 5 vs LLC 3 (1.100 vs 0.730mohms.) I may be misunderstanding this but I presumed that was one of the problems in the first place that the AC/DC loadlines were too high? Can anyone explain that in a simple way that a baby could understand? 🤣
I've been watching your previous LLC videos and I'll admit I'm still a little confused. I'm sorry I'm really dumb, but yeah could use another schooling on this if you've time.
For what its worth your videos still make WAY more sense than other ones. Someone on TH-cam with an LLC video was trying to say XMP doubles the access lanes and started rambling incoherently. ..... Urgh, that sounds like a back-handed complement.... what I mean to say is your genius and efforts to actually understand are much more meaningful than a lot of other videos out there. Thanks so much.
What exactly don't you understand? LLC = Load Line Calibration, a higher level (on Asus) leads to less vdroop. vdroop is the difference between the voltage at light load vs. full load. So instead of 1.4v single core vs. 1.15v multi core @ LLC 1, the voltage will be 1.3v single core vs. 1.25v multi core @ LLC 6 (made up numbers, additional undervolting recommended with higher LLC level)
@@cmiex Sure, but what do the other LLC settings do in combination with this one, and what other settings does this influence and why does my chip still hit a power throttle somewhere despite lifting all the power limits? I'm still missing something, and regardless it took me watching for days before realizing I got it backwards and switched LLC we're talking about from 3 to 6; much better. But there's this different LLC setting I'm wondering about, and I'm hitting some power limit despite having all kinds of power headroom. So yes I'm dumb, but this is a bit beyond setting the numbers on the LLC. I can change the numbers and watch the results, but I still don't understand how it all goes together.
I think the “worst” thing is the idle voltages. Difference between 0.7v idle or 1.xxv idle depends on the voltages manually set. So manual override mode would draw more power when idle.
Do you think Cinebench R15 is too sensitive to be used as a measure of system instability? I can't run R15 on my i7 14700k without it crashing, however I've been using the system like this for 3 weeks without any issues.
no AFAIK R15 isn't even as sensitive as some UE5 games(based on other people's comments). Though I've also noticed that which software causes an intel CPU to crash is a bit random. Some people have trouble running prime95 SFFT(which works fine for me) Other have issues with installing Nvidia drivers. Some have issues with compression/decompression software. As far as I'm concerned a system that is working correctly shouldn't have any issues running Cinebench R15.
My i7-14700k passes R15 everytime with adaptive undervolt, at least when I stay under -150mV, a few whea's on that voltage, thanks buildzoid's earlier vid on the Gigabyte z790, there's an unreal engine demo on a new game that you can run to check how it runs
4:08 I was about to complain about my chip on 480x65mm rad and AlphaCool Aurora Core 1 hitting 88 max running R23 on "Default" but then I remembered it's a14900KS and it has power limit of 320W and 400A
May be the firmware depart and those guys,um… support team? Doubt of announcement and take this responsibility something? Sometime you may know they cannot do it like: ASUS announces the firmware fixed intel issue! Hey , coz of this intel business , they won’t do this Or intel dun want ? 😂
Same test as Apex encore and 14900ks x2 alternative 14900K x1 at proart z790 WiFi Etc….. I can lock all core at 5.9 @ at 1.48v max all core with KS LLC 6 Since this is one of my computers For gaming only without any productivity needed. But I most favourite to set 5.7 at vcore @ 1.331v LLC 6 It may lose about 3% FPS Some ppls may blame this setting But I compare with same setting I didn’t test many game with unreal engine. But my user experience is IO VRM rail getter better ,especially when connect 240 monitor No sparkling and wired issue For some buggy game low frequency at lower voltage setting lower temperature , all thing is ok and smooth. Just want to share : Compare with R23 , I honestly test to R15 and time spy is more reliable.
Can you help a noob here? I got a 14600K around 1 month before all this came in to the spotlight and when i first installed it on my Gygabite Aorus Pro DDR4 B660m, it would run at full load at the boost speed of 5.3 Ghz. But now with the microcode update it just wont boost at full load beyond 4.88. I'm a newbie at modern overclock (before i used to do it on a AMD 1100t Phenom II x6) and i know i'm not really overclocking for real on this board. But i really want the all core 5.3 Ghz on full load. What settings on VRM, Voltage and CPU settings should i try?
B660 and B760 are bad less powerful motherboards for K series processors & they have no settings for SVID voltage and CPU Core voltage only can undervolt and change CPU LLC. Z690 or Z790 allows changing settings for voltage, svid etc etc like every other options. I would suggest run intel default settings, change core clock of p-core to 5.0 ghz, change PL1 and PL2 settings and 200-210A etc and see if it boots and clock works fine. I would suggest if possible upgrade motherboard but don;'t get Asus prime, MSI Pro or MAG series they are not ideal also.
@@sunteraze Thanks yea this board i got back when i had my 12400F, i got this new CPU thinking that it could atleast boost at full load on auto. I'm thinking in a Tuf Z790 as my system is still running DDR4 and DDR5 cost a kidney here. Thank you for the advises i'll try to set it up. Thanks @ion4290 too Edit: It actually boot with 210A and the frequency limit of 4.8 on full load @Cinebench R23 raised a little with a variation from 4.9 to 4.8 on the full pass. Thank you i feel i will indeed need to invest in a new motherboard and when this board had a 5.3 frequency lock on full load at R23 it was before 129 and the specs to reach it was unsafe i bet. I may breath and try 220 amps or i may quit before get the new board ;)
@@sunteraze I have a z790, using Intel Performance I'm also capped at 4.8. Using Intel Extreme, capped at 5.1. 70 and 80 degrees respectively, so not thermal throttling.
I find it funny that my 14700KF on a NZXT N5 Z690 (x129) can take -160 mV and is still stable in Cinebench R23 on Windows but Linux crashes instantly after booting the Kernel and is only stable with -125 mV.
That's because Cinebench, R23 especially, is no stability test. It's a benchmark. It's better to test with YCruncher and/or multiple different programs
@@cmiex I've also used Intel's own Processor Diagnostic Tool on Windows that runs SSE and AVX2 stability tests which pass. I've also played several games for hours on Windows under the mentioned conditions which also run fine. That is good enough stability for me. But I still don't get it why Linux just wouldn't tolerate it.
@@emini6 I didn't even get that far. But the undervolt was for sure the reason for that crash. As to why the system behaves differently, I have no idea. With a former BIOS, I knew that the microcode loading was to blame, as that newer microcode disabled undervolting or had issues with it. I could disable that on the command line which brought back some stability on Linux.
![[LIVE] : ONE ลุมพินี 88 | คู่เอก "ป้อมเพชร vs อัสลามจอน"](http://i.ytimg.com/vi/1ZqTVVbMgbo/mqdefault.jpg)
29:35 optimization settings
55:11 when to RMA CPU
35:55 undervolting settings - recommended for performance and even safer voltages but optional bc of instability risk: 38:00; also listen to 42:40 regarding undervolting!
Like he said, every CPU is different. The process of finding out how much you can undervolt is the following:
- Download Hardware Info 64 and BenchMate.
- Make sure you set the settings BZ listed (29:35, LLC6 is the most important even if you change other stuff).
- Set a negative VID offset, start with -0.07 f.a. (45:05).
- Boot, start BenchMate, start Cinebench, set a custom number of threads: top left corner click on "Preferences" to change this.
- Run multiple benchmark runs with different numbers of threads active (f.a. 1, 4, 8, 12, 16,...)
-> this ensures that different frequencies are stable. If fewer threads are active, the cores that are active have a greater "power budget" before the power-limit is reached. Therefore they can clock higher. You want to make sure that the highest clock is stable at the lowest voltage.
- If the CPU is stable, increase the negative offset value. I'd recommend increasing by 0.02 increments. Safes time. If you crash, lower the negative offset value by 0.01 and test again. After you found your stable offset (f.a. -0.11 like BZ has in this video), I'd personally lower it by at least 0.01 again (to -0.100) as an additional safety margin. Especially because cinebench is not a real stability test. YCruncher would be better f.a.
... or just google "how to undervolt"
how do u tune the lcc tho i dont get it rlly
@@Dnzn LCC?
Do you mean LLC? Load Line Calibration?
yes mb it was a typo i meant LLC.
Load Line Calibration
Thanks for the timestamps
Im on a Z690-A and my LLC only goes up to 7, his goes up to 8, should I still use 6? My AC/DC in HWInfo with his Optimized settings is 1.7/1.7, his ends up being 0.49/0.49
Edit turns out the "Intel Fail Safe" SVID which is default on F5, caused it, using auto solved it.
However my CPU only goes up to 4.8p, 3,7e ghz now.
Unrelated:
For the Global SVID do you not have to lower Cache aswell to match? Everyone else seem to say they should be equal.
With the undervolt, VID is about 0.05 lower than Vcore, is this correct? I get 1.288 allcore R23, vid is requesting around 1.23, this is with global adaptive -0.07
Actually changing global svid to -0.09 doesnt seem to do anything at all its still the same 1.288 vcore, changing to -0.1 also makes no difference to vcore.
Apex user here, thank you for doing a video on it.
Right? I killed my 13900KF with a bad delid just as all of this stuff about VID blew up. bought, delidded, and installed a 14900KF a couple of weeks ago, and am only now getting around to tuning. Good guy Buildzoid has a vid for me to watch about just that thing :P
Why do I get so excited when I see a new Buildzoid video that's an hour long and that pertains to a motherboard brand that I don't even own? 😂
Lol 😂 because nerd haha
Because at least he has a comprehensive knowledge and deserves the views.
Because even though he does ramble , he gets annoyed about the same stuff we do.
Actually shows configurations and what they do and how he got there.
Does not hold back when criticizing something just like how i usually think about these things that annoy me.
Awesome. Another reason to update-I’ve seen several 1.65+ VID requests on pre-0x125 with a 14700k on undervolted setups. This is when the ETVB formulas were messed up badly. If not constantly monitoring for it, I never would have caught it. 0x129 also includes that fix as well.
I think the stick figure drawing explained it perfectly.
Thank you! Great info. A lot more informative than other videos on this subject.
Black Myth Wukong Benchmark Tool is free on steam if you really want to stress via the UE5 shader compilation. Personally, I think that’s the best way to tell if you have instability or degradation beyond R15. UE5 after all was the first indicator that 13th/14th gen had problems to begin with.
thanks for the info
Buildzoid explains unlike most who just say do this. 💯
Thanks buildzoid. I finally took the courage to put intel default back on and play with load-lines after this. Got LLC 6 stable like you with an undervolt and ac loadline set to 0.4 with CEP enabled. Performance now near same within margin of 5-10% of asus oc profile
OK. First of all thanks for including TL;DW in the description. But second - I still do need to see this ;)
Thanks for the video buildzoid.
Just wanted to report that ASUS has done a poor job supporting laptops from as late as 2023.
My laptop with a 13980HX (essentially a 13900K with a wattage cap) latest bios revision from this past month is using microcode 0x123, and can easily voltage spike between 1.55V - 1.6V
Thank you for this video. I learned a ton.
Buildzoid more like goatzoid
“Actually Hardcore Undervolting”
Thanks for all the information
With an ASUS Tuf Z790 and 13700k I get slightly better numbers under ASUS OC than yours under intel settings. With ASUS they use LLC 3 with 0.4/1.1 ACDC. Under this setting I get around 29600 R23 score and 85 C max temps. But with Intel settings at LLC 6 0.49/0.49 ACDC and a - 0.100 v offset I was getting 30000 R23 score but temps were 95 C. Both options were pulling around the same wattage but the ASUS one has very slightly higher vcore voltages. Using IA VR limit of 1450 in both options too.
Not sure which would be the better option to go with but for now I'm sticking with the lower temps.
Same, 10c higher temps
same
Intel fixed the overvolting issue, so the difference between stock (LLC 3 + 1.1 Ohms) and LLC 6 + 0.49 ohms is very minor now. Of course if you do unmatched LLC and AC/DC you have to disable CEP which may reduce stability. Otherwise you can still do the SVID override (adaptive voltage offset) with stock LLC and AC/DC
Question to clear up some confusion: Most Asus "guides" online (for AC undervolting) suggest set LLC = 4, AC = 0.40 and reduce AC until unstable, if you can reach 0.01 AC, reduce LLC = 3, repeat with AC til unstable.
This obviously triggers CEP in most configurations since LLC 3 is 1.1 mohms but having an AC that is way lower, makes CEP unhappy.
So lets say the system is set to LLC = 3 / AC = 0.20 do you just set DC to also 0.20 or just go with a higher LLC (like 6 as shown in the video, (just as an example)).
I think that is where (for Asus) lots of the confusion is coming from for undervolting because everyone says a more droopy setup is good and don't use a high LLC (meaning above 4).
This is a good question and something I've also encountered.
That is why I am confused about Buildzoid's chip. He says his isn't stable at the mobo default settings, but pretty much everyone I see have chips that not only are stable at the 0.4ohm, usually they can go much lower. My 13700K can go all the way down to 0.01ohm on the AC even and still be stable. I wouldn't think all the major motherboard makers would set their default settings to voltages that weren't at least stable, in the past if anything they have errored on the side of giving it too much voltage. I really just wonder if Buildzoid has a broken CPU from the factory.
I tried to get my same undervolt with the Intel Default Settings just using a voltage offset, but at the certain point, it just doesn't go any lower unless I start messing with the cache offset even though my uncore VID is already way lower than it was with my LLC undervolt.
@@ZoneXV I thought my 14900k was stable just AC LL undervolting down to 0.1, passed everything til I ran OCCT, Insta core crashes/WHEA errors. I plan to test out his settings here. All I really care about is gaming performance and temps during that gaming.
The OP does have a good question though
Janitorus reading all this comments to update that reddit post "13/14th gen "Intel baseline" can still degrade CPU"
@@ZoneXV Regarding your last paragraph: The reason it won't go any lower is because not only the cache vid but also the e-core vid can limit the VCORE at a certain point. The VID offset is only for the p-cores.
And regarding your first paragraph: luckily BZ chip is one of the lowest end chips on the curve. That way we get first hand experience and guidance of the worst case! His chip isn't broken, just very bad.
Hi Mr,
please create new version according to new bios update released 2 days ago.
I think there will be 2-3% lower performance :(
I got same CPU and MOBO as on this video, and before new bios 1703 I had 39.6k points with similar settings as on movie, but after update I set same settings and have "only" 39k points...
New BIOS version have new microcode 0x12B and looks like it lower performance even more...
damn i was here to see how SVID behavior effect cpu , there is a "intel fail safe" vs "typical" scenario war rn but still got some other ideas . thanks for the video
the SVID beahvior setting just messes with the AC/DC LL. don't use it.
@@ActuallyHardcoreOverclocking should we leave SVID behavior on auto?
@@unhappymeal5652 yes just leave it on auto
@@unhappymeal5652 I just tested this on my Z690 Board, and if i leave it on "Intel Fail Safe" the AC/DC LL is set to 1.100 mOhm not matter what i set my LLC Level to (AC/DC LL sync is enabled). So the SVID Behavior on ASUS Boards does indeed influence what the other SVID/Load-Line related settings actually do. If i set it to "Auto" and the LLC to Level 6, i also get 0.49 mOhm AC/DC LL
Fail safe = killing safety to RMA
Thanks. I am just upgrading from 13900k to 14900ks and switching from Z790 Aorus Master to Z790 Apex Encore for that 7800 MHz stable RAM and 1% FPS boost :)
(no cap, really, just for fun)
Great informative video, as usual. Cheers!
lol
IA VR Voltage limits are for peace of mind, although bere in mind if your CPU is a potato setting 1.4v will prevent your CPU even maintaining 5.7Ghz even in gaming. You’ll suddenly be fluctuating between 5.3Ghz to 5.7Ghz. An undervolt addresses this since it moves you up the VID boost table. Although I’m still unsure about level 6 LLC, since you’re then moving into AIB AC/DC ranges, and technically speaking the existence of CEP prevents the instability you would otherwise get. Think I’m getting that right….
Very useful videos👍👍thanks BZ
10:00 that's not true. It is also the ICC_MAX for the 13900K and 14900K in the Extreme profile, just like the 13900KS and 14900KS. The only difference between the K and KS in Intel Default Settings spec is that KS has PL2 and PL4 of 320W and 550W respectively, while the K has PL2 and PL4 of 253W and 380W. This is all just assuming that you use the Extreme profile (which is "Recommended" while the "Performance" profile is "Default" according to Intel Default Settings spec, on all of the mentioned CPUs). So you should be using the Extreme profile on the 14900K, which would set the ICC_MAX to 400A.
Buildzoid, could you investigate microcode update 10E? Tech Yes City put out a video in August that just seems like nonsense to me but suggests that he had lower voltages on that patch and that newer microcodes have been cranking them up, particularly VDD and IMC values that he believes should be at 1.3V. Pretty sure these are just memory-related though...
당신의 정보는 매우 유용합니다. 감사합니다
I was working on finalizing my undervolt and after watching this video decided to try to run P95 Small FFT with only 8 cores and no HT to really push the P-cores and check for instablity. I received the out of memory error and noticed that MSI AB crashed. This does not happen with 24-cores.
Current settings when MSI AB crashed: Asus Z790 Dark Hero MB
Vcore to 1.4v (IA VR LIMIT Bios setting), set 253w/400A limits, use LLC#6 with ACLL@0.49 and DCLL@0.54(calibrated), enable CEP, use V/F curve 6,7,8,10 undervolt -0.13.
I reduced my undervolt in increments of -0.05 and each step MSI AB would still crash within Test 2 of the Small FFT run. When i finally hit -0.11v MSI AB no longer crashes and I am able to get through many more runs of P95(3 cycles of 6 tests each). Does the out of memory error only happen when the chip has degraded or can it also happen with a too severe undervolt? In other words, is the out of memory error simplly an i/o issue with too little Vcore, or does this only happen once the chip has degraded and more voltage allows you to hide it for a while? Prior to this I was R23/R15 stable and have not received any any WHEA errors (weeks), but had not yet tried P95. I am curious if i need to be concerned.
Telling an internet user to read = the most personal attack you could possibly make
So to sum up, i want to change load line calibration to 6 and then change global svid to offset - and set for example as you did 0.11
And thats it?
The IA VR Voltage Limit i do not know what works best for my CPU maybe just try 1400 as youre example and see how it goes?
With further inspection i did set load line calibration to 6, set global svid to adaptive and - but the voltage offset was greyed out. Couldnt be changed. So i guess
there is a setting that isnt turned on for it to be changed? I did not have "offset" as setting in drop down. Only auto, adaptive and manual. So our bioses for our motherboards look different.
Gained 400 points in cinebench r23 with those settings. What im i missing and what should i try?
Using Z790, 14700K and 1503 bios with 0x129 microcode, intel default settings, performance profile. XMP I enabled.
Without the undervolt and that tweaks i get about 33.000 points in cinebench r23
Awesome video!
I've been overclocking/undervolting for about 6 weeks now and this video just really made so many things make sense for me.
I've watched it now 4x's and some things just clicked.
One thing that still confuses me is Intel's pl1/pl2. Is this intended for people with minimal cooling abilities wether lack of fans or small AIO? Or is this a hard limit no matter the system?
Anyways, thanks so much for this video.
Excellent video bz! Thank you kindly for doing this on an asus board. Appreciate your hard work and i always learn something useful
I've been waiting for this! Thanks for this!
Ohhhhhhhhh baby Finnally the real deal the grand daddy my Board and the best z790 board out!
How much can I apply this to the Asus B760-A? First thing I noticed is that "sync ACDC loadline to VRM" doesn't apply, it stays at 1.1 mOms regardless of VRM Loadline.
Secondly, there is no level 8 for VRM LL, does this mean Level 6 may not be 0.49 mOhms ACDC for my board?
Lastly, I can't do adaptive SVID offset, it's either Auto or Manual. V/F points are also fixed. The only thing I can touch is actual VRM core voltage, which means I'm unable to undervolt without CEP kicking in, and that is forced enabled on my B series board. So I'm totally out of any way to undervolt?
Z690-G board here, m y Sync ACDC LL feature doesn't work either.
This is great as I run a 13900KS on this board.
Just wondering if your thought process would do anything differently for the 13900KS off the new intel defaults than what you show/describe here.
Does turn off CEP disable the 1.55v VID limit?
At least Asus shows the VID table when you want to set a VF point offset. MSI only shows the fields to enter the offset, with no VID values whatsoever. 😒
I haven't found a way yet to display the VID table on an MSI motherboard, the info seems to be so low-level that it's only available in the BIOS and not accessible by regular software.
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
ASUS OC
Thank you so Much for this BZ! Keep up the great work!
You Rock!
🤘
If i limited my Vcore to 1.4v (IA VR LIMIT Bios setting), set 253w/400A limits, use LLC#6 with ACLL@0.49 and DCLL@0.54(calibrated), enable CEP, use V/F curve 6,7,8,10 undervolt -0.13. Is there any danger to using ASUS OC Profile over the Intel one? I feel like the VRM limit being 1.4v supersedes the Intel 1.55v limit from the microcode. is this correct?
Will you be doing Asrock motherboard too?
I second the request 🙋🏻♂️🙏
I have also discovered another odd thing that using the "Intel Default Settings" profile does that I'm not sure is specific to ASUS or not. Using the Intel settings, it forces a really aggressive level of C-States. Aggressive enough that I'm having certain p-cores drop down to 800mhz during gaming. That is either the balanced or high performance power plans and c-states set to auto in the BIOS.
I saw this too, Asus board and Intel Extreme or Performance, PCores dropping to 800mhz
Just turning "Enhanced C States" off should fix this?
@@wardeagle1015 I've got a note to check this, thanks. Why would c-states limit the clock speed on the Intel Profiles?
@@Intel13thGenBurns it shouldn't limit the clock speeds but it does allow them to downclock that low is my semi-understabding
Just go to bios and look for C1E and set to disabled. Leave regular C states enabled. If you want clockspeeds to drop during balanced power plan, set speedshift to Disabled and EIST (Enhanced Intel Speedstep) to Enabled. This way balanced power plan will function properly.
0x129 enables C1E by default, this causes the random 800mhz drops in my experience.
I've never seen over 36555.. bummer.
Edit: by the end of the video, 38767! Happy fuqing camper, thanks dude.
Hi, what do you think about my undervolting on a 14900kf. I have LLC4 with Intel Default Extreme with -0.07V on Global Svid Voltage on an Asus Rog Strix Z90-F Gaming Wifi.
I also have TVB (on) and TVB Voltage (auto = on). In most games I have a Vcore of 1.38V (+- 0.03V) with 5.7ghz on the P-Cores and 4.4 on the E-Cores.
In Cinebench R23 I have 38,500 points but in Time Spy a CPU score of 20,500 points (possibly for G-Sync. The temperatures are 50-65 degrees depending on the game. Microcode is 0x12B
I'm on a I7 14700k do you recommend LLC6 for that as well? I'm currently running LLC4 at the moment
I think lower the better if its stable.
Low as possible while stable
I'd use level 6 for all CPUs.
@@ActuallyHardcoreOverclockingLLC6 it is. Enough said when coming from BZ!
@@ActuallyHardcoreOverclocking I wish an explanation for that please why some says like that and some will say the less the better?
Hey @buildzoid,
I'm facing an issue with constant current/EDP throttling ever since my CPU arrived last month. No matter what I do, the throttling persists. I tried lowering my PL1/PL2 limit below 220W and increasing ICCMax to 307A, even up to 370A or 400A, but it's still throttling.
I followed your 0x129 microcode video and set LLC to level 5, which lowered the resistance to 0.730/0.730 mOhm from 1/1 mOhm at level 4. However, after applying these settings, the throttling continues, even when idle and while playing CPU-heavy games.
I feel stuck and frustrated. As you said "i want to use my pc not fix it". I've read tons of articles, but most of it goes over my head (not very tech-savvy).
Here's my setup:
i7 13700K (Cooler Master 360 Atmos AIO)
Asus ROG Strix Z790-H Gaming WiFi
ROG Strix 750W PSU
Any help or insight into why this is happening would be really appreciated. Thanks for your time, and sorry for the long message!
Are you setting your higher iccmax in xtu or the bios? I had a weird issue with xtu showing my pl1/2 and iccmax were raised but would never go above default valuesm I had to go to the bios and set them in the bios to actually get different values for some reason.
@zachpaw i changed it from bios, still getting the the same issue.
@@xvjaqen how hard are you undervolting? Not sure how much it applies to computer cpu's but basic electrical theory says when doing the same amount of work, amperage increases if voltage decreased and vice versa. So if you have a giant undervolt it needs more amps to do the same amount of work. If you don't have a big undervolt then it's something I'll have to keep thinking about or wait for someone smarter to comment.
Thx for ASUS option
Thanks!
Whats your opinion on Sync all Cores to prevent the 2 p-cores from running away from you? My problem with the new microcode is that my idling voltages and temps are stupid high.
I just set a max ratio to 57 on my 14900k, as 6Ghz is the culprit for silly spikes to begin with. And let’s be honest, it’s pretty useless and you’re not going to miss it. There was no need for 6Ghz boosts apart from Intel claiming the “fastest gaming cpu crown”. Then set an IA VR Voltage to 1.42, and a 0.070 undervolt and forget about it imo.
Can anyone go into more detail on the accuracy of the oscilloscope readings. I have V Latch and VR VOUT monitoring on this board the Asus Apex Encore and wondered how close this figure was to the reading from the scope? I believe BZ mentioned that the reading could be off by several millivolts under full load and close to actual at idle. Not sure if that would mean reading more or less volts. Any comments? Thanks!
What I did was a delid with an EK Nucleus Direct Die aio, my chip is not the greatest. The latest bios/microcode locked my two preferred cores to 57, then to get the 5.7ghz consistent on all P cores I set my IA VR Maximum Voltage to 1420, Vcore hovers around 1.35v and my VID request do not exceed 1.35-138Xv per core.
Now if I want to fully micromanage my system, I cannot use the new baseline bios with microcode, I have to use the bios before the first baseline and an older microcode version, otherwise tuning/fully micromanaging the cpu becomes a painful nightmare.
6ghz+ requires too much voltage they should have just limited their chips to 5.7-5.8ghz maybe made a 12-16P core no E core chip and called it a day...
23:05 "It does get over 40k points though.. it's is a real shame that it isn't, ya know, stable"
Is it just me or loadline undervolting works better than this? I have a 13900k on an Asus Prime Z790P and an Arctic Liquid Freezer II 360. If I undervolt in this way I still get thermal throttling in Cinebench R23. If I undervolt disabling CEP and setting LLC level 6, AC 0.19 and DC 0.5 I no longer thermal throttle. EDIT: I use LLC5 not 6
Are you unstable below 0.19? Are you testing with anything aside from R23?
So you're saying by using his settings including a - 110mv offset you thermal throttle in R23?
I would also say try his exact settings but disable IA CEP and see what happens also, be curious to know. Maybe even try lowering the offset slightly more?
@@wardeagle1015 Yes I'm unstable at AC 0.17, cannot pass R15, so I set it at 0.19 to have a buffer. I test with R23, R15 and R15 extreme. Using his settings including the offset I thermal throttle, yes. An offset higher than -120mV is not stable for me, and I still thermal throttle even at -120mV. So unfortunately his way of setting things don't work for my CPU, at least if I want to get the most out of the undervolt.
@@IlCode85 That's so surprising actually I would think it should yield similar results at worst. I wonder if there's something else changing that's not been made aware of. Not insinuating there is but still just surprises me.
Whats your SP and Cooler score? Just out of curiosity
Also have you tried something more droopy just out of curiosity like LLC4 and then just drop AC LL as low as you can, turn CEP off and test that. Be interested to see if the results differ
@@wardeagle1015 Could it be that undervolting with AC and DC loadlines has a different effect depending on the load compared to undervolting with an offset? Just my thought, I'm not an expert at all. Unfortunately I cannot check my SP and cooler score as my motherboard is just a Prime Z790P, not one of those high end ones.
Will Asus release another BIOS update applying the 1.55v limit to their custom profile?
Wouldn’t it be worth doing an all core setting? Your single core boosts is what causes the extremely high voltages
i straight up love ur vids i always put them on loop and just respect the work that u put into each single vid
Hey, great videos ~! thank you! do you recommend HT turned on for gaming on the 14900k? I allways turned it of or let it on only 2 cores
Depending on which that I am answer,like gta many npcs backgrounds should be improved but also according the programming designed
More scenarios I just turn off may improve the fps. And some game will suxk of HT
Thx for doing it on Asus motherboard
Im on a Z690-A and my LLC only goes up to 7, his goes up to 8, should I still use 6? My AC/DC in HWInfo with his Optimized settings is 1.7/1.7, his ends up being 0.49/0.49
Edit turns out the "Intel Fail Safe" SVID which is default on F5, caused it, using auto solved it.
However my CPU only goes up to 4.8p, 3,7e ghz now.
Unrelated:
For the Global SVID do you not have to lower Cache aswell to match? Everyone else seem to say they should be equal.
With the undervolt, VID is about 0.05 lower than Vcore, is this correct? I get 1.288 allcore R23, vid is requesting around 1.23, this is with global adaptive -0.07
Actually changing global svid to -0.09 doesnt seem to do anything at all its still the same 1.288 vcore.
I’m still trying to figure out why sync all cores etc options seem prohibited to use since x129 and bombards you with power limit warnings instead.
Probably been asked in the past but @buildzoid I love your wallpaper where can I find it? =)
I make all the wallpapers I use and I haven't uplodaed this one anywhere.
@@ActuallyHardcoreOverclocking I had a feeling that was going to be my answer. Well, I love it. Would buy 😎😁
Asus board is very easy to optimise on but unfortunately in the intel profile mode we cant cap our p cores only in the asus oc profile mode we can. I was doing that for eztra safety to not run at higher clock frequency for an automatic undervolt on top on the slight adaptive undervolt for the svids at -0.01
Can you make a step by step guide locking all cores and tuning voltage on Apex Encore/14900k Thank you for this video
there's no reason to lock the core clocks just set a voltage limit and then undervolt.
@@ActuallyHardcoreOverclockingB-but Intel dip
Please tell me: how to make an undervolt when you have the 13th generation and a ASUS B760 motherboard has no function "Adaptive Mode" in Global Core SVID Voltage setting?
you need Z board
we’ll get far By MY 🔥calculations
Thank you very much for this great explanation, I'm glad to see that my board handles the new microcode correctly. I was still getting apps crashing with 0x0000005 and RMA'd my 14900KS, the replacement will arrive tomorrow. The board is already updated and I'm thinking about if I should undervolt using the method shown in the Video or if I just leave everything on factory defaults - I've chosen the 14900KS because I don't wanted to bother with overclocking at all but still wanted get the best possible performance out of the box, but now I'm thinking; The undervolt seems to be even safer than the Intel defaults AND having a higher score. Are you using a contact frame or the Intel ILM? Question to all: Which contact frame would you recommend?
If I'm not mistaken, undervolting can only damage your sanity, if your app or computer crashes when you don't want it to. Other than that, it's trading a bit of time in setting it up and testing that it's stable, for better everything. That is, the crashing from undervolting, from what I know, don't degrade the CPU. So in a way, it would make it even safer.
That being said, I'm not an overclocker (well, probably I'd be, if I wasn't on a laptop) so, do check with others too.
so for new intel user, just update the bios and reset the bios to default and youll be fine?
I manually set all the default power limits to intel extreme on my 14900ks with the Asus OC profile but with some safety nets turned back on . On a z790 Maximus Dark Hero.
Set Svid behavior to auto, CPU LLC Lvl at 5 and matched the internal CPU power management to intel settings.
Core usage to Auto. I have various other small settings changed for stability and gaming performance, like turning off PCIE Aspm and other random stuff.
But the main thing is I use to fine tune the voltage curve is actually their AI features tab and I set my cooler training to Manual and use that to fine tune the voltage curve.
My peak performance is at setting cooler score to 178 and leaving it alone.
what are your temps? I have same specs but thermal throttle all the time
@@schneezockt about 32c Idle (same as liquid temps). About 55-60c running something cpu heavy like Read Dead 2 or Spiderman Mile Morales, 40c running Black Myth Wukong. I'm using an Asus Ryujin III 360 Aio with some Noctua 3000rpm IPPC fans with some pro-active fan curves as well.
@@TheAJKid did you apply thermal paste on your own? Because I play the same gamed and have the same AIO :(
@@schneezockt Yes, I have Thermal Grizzly on there as well as a contact frame.
Can u share your bios settings somehow that I can see it and copy everything to my bios?
35:57 Should “undervolt protection” be set to disabled?
My 14900K (and Asus Prime Z790) is only running at 5000 as it's hitting around 253W and therefore will not send more voltage to the CPU (my guess). With CPU under full load, I'm hitting an average of 1.14V, with an -0.01V offset. When I try to increase PL1 to 300W in Bios, it's still showing PL1 max of 253W. When I then set it in Intel ETU, it is changing the PL1 max to 300W, but only drawing up to 264W. Are there any other bottleneck in settings I have to change?
Gigabyte finally released new bios time to test it on GA board as well :D
Will you be also testing an ASRock board?
He doesn't have any ASRock boards
What's the difference between Intel Default Settings and Asus Advanced OC Profile with all the settings manually set the same as Intel Defaults?
Is this guide applying on a asus z790 f gaming?
Running heavy computation on 14900k can tell that these settings are unstable, not even default constantly get segmentation fault after updating to 129 microcode, doing heavy computation on Intel become a nightmare (which wasn’t before update, still can’t find stable settings or CPU degraded to the point where it can’t be stable anymore)
The only thing helped with segmentation fault errors is disabling CEP
they have 0X12B，any update？
0x12B was just the past two Microcodes rolled into one, basically. I'm sure there were a few more minor underlying changes but all in all shouldn't be much different than 129. All of my benchmarks and undervolts still work and bench the exact same. Just my personal experience, though
Idk i have a 14700 and a b760 i cant set an offset on vid only on actual vrm voltage, so i have to turn cep off to get my cores above 4.6ghz what do you think?
how would I go about determining the proper LLC level for other Asus motherboards? say Z790-E, Z790-H
So my cpu was "fine" except that it had some mayor problems with training tertiary ram timings on all my different ram dies, after the update my screen goes black in certain games, game crashes and 5 beeps, any advise?
58:00
I have a question for you, is this bellow safe ?
I'm on the version before that but I see no reason to update, the difference is I do not use the offset, I use static voltage and I have it set to 1.25V, my load line is 0.500 and my load calibration is on level 4, my real vid is 1.220V on idle.
My vid request on load are max 1.340V and the "real" vid drops to 1.190V ( the request go to 1.340V ).
My CPU (13700KF) stays at 5.4Ghz all P cores, board is Asus Z790-P WIFI, tested with OCCT extreme stability test, also test with Ezbench (unreal 5) and keeps the CPU ratio of 54 fully stable (1.195V/1.197V), temp at max 72C, idle temps are around 34C to 36C
I actually mess with the V/F and the max I allow is 1.35V for 54, have my IA VR set to 140, also I have the Intel default settings set.
Now if I understand the video right you also have a max request around 1.35V ish , but I see on the C1 vid max 1.45V ish , now if my voltage is static does it actually reach 1.4V ????
what do you mean by static voltage. You mean you set the "Actual Vcore VRM voltage" to "manual" and then set the voltage to 1.25V?
@@ActuallyHardcoreOverclocking thank you for the reply, that is correct, I set the voltage to manual and the board respect that voltage
@@custume yeah that's safe
@@ActuallyHardcoreOverclocking Thank you for the reply, 😉
Normally I do this to all my computers before this but because of this new "bug" I was not sure if even doing this might have problems with this new "bug"
Can I use the same settings for KS as well? I have thermal throttle issues
hey man if you get time can you test an onsemi gigabyte board would be very much appreciated
I wonder how this applys to msi Z790ACE , the msi options are a bit diferent . I understand what needs to be done and and will try to aply it on my ACE +14900K
He has a video of these same steps basically on an MSI MOBO. He released it right before this one
Thanks , I saw it but it's different chipset Z690
is there a hedt roadmap beyond amd threadripper 9000 ? and for after intel emerald rapids hedt ?
How is my Corsair AIO 360mm cooler not able* to control temps (thermal throttles during r23) for 13700k but your 240mm AIO is keeping everything on a 14900k under 90C? Even on default settings on the same microcode???
Edit: Typo
The 700Ks are both a little bit worse binned and no lower power than the 900s. Basically the added E cores don't make them harder to cool, but they do take a bit of power budget away from the P cores, meaning a 900 ends up with lower temps.
Ecores were a mistake. Take as-much or more power, for less performance, and they hog power away from p cores. Glad this ring bus fudge up is making it clear how poorly planned Intel's last 2 generations have been.
@@RotaryJunkieOh I see, thank you for the quick response. I can't even seem to get a 5.3 Ghz overclock let alone a 5.0 Ghz to stay under thermal throttles. Not sure if my chip just needs to be RMA or a new cooler is needed. Temps are 40s-50c on no load, except for the first load of the day.
make sure your pump speed is maxed?
@@ActuallyHardcoreOverclockingThank you for your videos Buildzoid! The fans and pump are maxed. Let me know if you'd like any other info regarding the setup just in case it's something resolvable outside of RMA. I've got an Asus Z790-p prime wifi motherboard, a 13700k, Corsair H150i 360mm AIO water-cooler. Using a 0.1 adaptive offset as opposed to the 0.11 used in the video.
Edit: Added specs for setup.
With all the various brakes now being applied by Intel and/or the mobo makers, I wonder if/how much the average user will notice in real life?
I mean yes they might see say 170FPS instead of 180FPS for instance, or an app taking 0.1 secs longer, but in actual use, will they notice?
Yes, we should have had the product as described, but we didn't.
I'll settle for settings that give me stability and longivity, over a 5 - 10% gain in performance that is doubtful I could realistically notice.
Hey guys. Are we always using balanced power plan in windows for 13700k i7? Is there any advantage using high performance or e.g. Bitsum Highest performance? I am only gaming on my pc. I already did all the stuff (A/C load line settings, newest bios) to adress the degrading issue. With balanced plan all these settings I am getting 30500 cinebench 23 score. And is there an advantage using parkcontrol to unpark cores? Or just enable c-states and let windows and cpu handle it? Ty for answers
For gaming I reckon balanced plan may be OK, performance is pretty much required for low latency stuff, like live audio synthesis
10:56 BZ says that the Vmax voltage is higher now with LLC 5 vs LLC 3 (1.100 vs 0.730mohms.) I may be misunderstanding this but I presumed that was one of the problems in the first place that the AC/DC loadlines were too high? Can anyone explain that in a simple way that a baby could understand? 🤣
I've been watching your previous LLC videos and I'll admit I'm still a little confused. I'm sorry I'm really dumb, but yeah could use another schooling on this if you've time.
For what its worth your videos still make WAY more sense than other ones. Someone on TH-cam with an LLC video was trying to say XMP doubles the access lanes and started rambling incoherently. ..... Urgh, that sounds like a back-handed complement.... what I mean to say is your genius and efforts to actually understand are much more meaningful than a lot of other videos out there. Thanks so much.
What exactly don't you understand?
LLC = Load Line Calibration, a higher level (on Asus) leads to less vdroop.
vdroop is the difference between the voltage at light load vs. full load.
So instead of 1.4v single core vs. 1.15v multi core @ LLC 1, the voltage will be 1.3v single core vs. 1.25v multi core @ LLC 6 (made up numbers, additional undervolting recommended with higher LLC level)
@@cmiex Sure, but what do the other LLC settings do in combination with this one, and what other settings does this influence and why does my chip still hit a power throttle somewhere despite lifting all the power limits? I'm still missing something, and regardless it took me watching for days before realizing I got it backwards and switched LLC we're talking about from 3 to 6; much better. But there's this different LLC setting I'm wondering about, and I'm hitting some power limit despite having all kinds of power headroom.
So yes I'm dumb, but this is a bit beyond setting the numbers on the LLC. I can change the numbers and watch the results, but I still don't understand how it all goes together.
I also though you were all crazy telling me I had to have my TFAW at 16; didn't realize I had to add about 0.08v to account for that. 0_0
Hey, 36:11 so what happens if you undervolt via Actual VRM Core Voltage and not Global SVID Voltage? What could go wrong?
I think the “worst” thing is the idle voltages. Difference between 0.7v idle or 1.xxv idle depends on the voltages manually set. So manual override mode would draw more power when idle.
Not using dark theme on HW Info is a war crime, I think.
How about 14700k optimum settings?
Do you think Cinebench R15 is too sensitive to be used as a measure of system instability?
I can't run R15 on my i7 14700k without it crashing, however I've been using the system like this for 3 weeks without any issues.
no AFAIK R15 isn't even as sensitive as some UE5 games(based on other people's comments). Though I've also noticed that which software causes an intel CPU to crash is a bit random. Some people have trouble running prime95 SFFT(which works fine for me) Other have issues with installing Nvidia drivers. Some have issues with compression/decompression software. As far as I'm concerned a system that is working correctly shouldn't have any issues running Cinebench R15.
My i7-14700k passes R15 everytime with adaptive undervolt, at least when I stay under -150mV, a few whea's on that voltage, thanks buildzoid's earlier vid on the Gigabyte z790, there's an unreal engine demo on a new game that you can run to check how it runs
Black Myth: Wukong benchmark
4:08 I was about to complain about my chip on 480x65mm rad and AlphaCool Aurora Core 1 hitting 88 max running R23 on "Default" but then I remembered it's a14900KS and it has power limit of 320W and 400A
Yes my KS F5 also
Go 400A 320 W
420mm x2 46mm rad Pwm DDC D5 pump
1503 F5 for KS and K isn’t extreme 320?? 3chips is same
What ASUS do? Why the video is 253?
Funny to find out it was user error. So asus boards aren't having more issues with it. It's just users didn't read. Lol
May be the firmware depart and those guys,um… support team? Doubt of announcement and take this responsibility something?
Sometime you may know they cannot do it like:
ASUS announces the firmware fixed intel issue!
Hey , coz of this intel business , they won’t do this
Or intel dun want ? 😂
0x129 running flawlessly on my rog strix z790i with 13700k.
Same test as Apex encore and 14900ks x2 alternative
14900K x1 at proart z790 WiFi
Etc…..
I can lock all core at 5.9 @ at 1.48v max all core with KS LLC 6
Since this is one of my computers
For gaming only without any productivity needed.
But I most favourite to set 5.7 at vcore @ 1.331v LLC 6
It may lose about 3% FPS
Some ppls may blame this setting
But I compare with same setting
I didn’t test many game with unreal engine.
But my user experience is IO VRM rail getter better ,especially when connect 240 monitor
No sparkling and wired issue
For some buggy game low frequency at lower voltage setting lower temperature , all thing is ok and smooth.
Just want to share :
Compare with R23 ,
I honestly test to R15 and time spy is more reliable.
Forget to thanks this video 😂
Can you help a noob here?
I got a 14600K around 1 month before all this came in to the spotlight and when i first installed it on my Gygabite Aorus Pro DDR4 B660m, it would run at full load at the boost speed of 5.3 Ghz.
But now with the microcode update it just wont boost at full load beyond 4.88.
I'm a newbie at modern overclock (before i used to do it on a AMD 1100t Phenom II x6) and i know i'm not really overclocking for real on this board. But i really want the all core 5.3 Ghz on full load.
What settings on VRM, Voltage and CPU settings should i try?
1) last bios default settings 200A 181W
2) undervolting (adaptive offset)
3) use it
B660 and B760 are bad less powerful motherboards for K series processors & they have no settings for SVID voltage and CPU Core voltage only can undervolt and change CPU LLC. Z690 or Z790 allows changing settings for voltage, svid etc etc like every other options. I would suggest run intel default settings, change core clock of p-core to 5.0 ghz, change PL1 and PL2 settings and 200-210A etc and see if it boots and clock works fine. I would suggest if possible upgrade motherboard but don;'t get Asus prime, MSI Pro or MAG series they are not ideal also.
@@sunteraze Thanks yea this board i got back when i had my 12400F, i got this new CPU thinking that it could atleast boost at full load on auto.
I'm thinking in a Tuf Z790 as my system is still running DDR4 and DDR5 cost a kidney here.
Thank you for the advises i'll try to set it up.
Thanks @ion4290 too
Edit: It actually boot with 210A and the frequency limit of 4.8 on full load @Cinebench R23 raised a little with a variation from 4.9 to 4.8 on the full pass. Thank you i feel i will indeed need to invest in a new motherboard and when this board had a 5.3 frequency lock on full load at R23 it was before 129 and the specs to reach it was unsafe i bet. I may breath and try 220 amps or i may quit before get the new board ;)
@@sunteraze I have a z790, using Intel Performance I'm also capped at 4.8. Using Intel Extreme, capped at 5.1. 70 and 80 degrees respectively, so not thermal throttling.
I don't own an Intel CPU but I am really enjoying these videos testing the voltage 😂
I find it funny that my 14700KF on a NZXT N5 Z690 (x129) can take -160 mV and is still stable in Cinebench R23 on Windows but Linux crashes instantly after booting the Kernel and is only stable with -125 mV.
That's because Cinebench, R23 especially, is no stability test. It's a benchmark. It's better to test with YCruncher and/or multiple different programs
@@cmiex I've also used Intel's own Processor Diagnostic Tool on Windows that runs SSE and AVX2 stability tests which pass. I've also played several games for hours on Windows under the mentioned conditions which also run fine. That is good enough stability for me. But I still don't get it why Linux just wouldn't tolerate it.
Is it a kernel panic or a MCE message reported on dmesg?
@@emini6 I didn't even get that far. But the undervolt was for sure the reason for that crash. As to why the system behaves differently, I have no idea. With a former BIOS, I knew that the microcode loading was to blame, as that newer microcode disabled undervolting or had issues with it. I could disable that on the command line which brought back some stability on Linux.