Why only 12.6K subscribers. This TH-camr is an overclockers dream. ❤ He explains everything in detail and it all makes sense. Cream of the crop compared other overclockers who make Videos.👍👍👍👍👍👍👍👍
Well explained with some of your OCTVB settings feels good to play at 6ghz , your the underrated king of the overclocking youtube content, At overclocking you are already king :D DDR5 overcloking guides or DDR4/5 in general would be awsome.
Great video and guide, helped me a lot with my 13900ks but on a z690. I was having hight temps out of the box and initially only applied a -0.100 offset. After applying your settings now I have great temps and clocks.
Really appreciate the thoroughness of this guide. Very useful information presented in a manner that makes sense and is easy to understand. Keep up the amazing work!
Wow. I learned so much today. I have previously overclocked just by turning numbers up slowly and praying for the best lol. The Asus Extreme mobo I have now has so much added functionality that I decided to try and learn what everything actually means. This video and your content in general has been astonishingly helpful. Thank you.
I just copied all of these settings to my rig with a 13900KS, ProArt z790 being cooled by a arctic liquid freezer 360. Everything works great! I get roundabout 42,000 points in cinebench, have been using the pc with this overclock for over 1 month now and never had a blue screen. Saw on the comments that people were saying that it doesn't work because he has had luck with his chip, getting a sp predicion of 115 while most people get 107, I GET 105 AND EVERYTHING IS ROCK-SOLID. THANK YOU SO MUCH !!
The selection of the per core limit in 38:20 may only be applicable to your specific CPU and not to others. You also seem to have a highly selected and binned CPU with an SP prediction of 115 whereas the average is about 107. So it is very likely that everyone with a lower binned processor will get an unstable system even when exactly following your video. That's my case with an i9-13900KS at SP 108. Besides this, my comment is not meant to be a rant about your video which gives a fantastic insight into OC and the Rapor Lake Platform. Thank you for this! 👍🙂
Mine is not 100% stable at those settings.. I have a KS 108 (p cores at 117). But CBR20 cannot complete and my browser closes random. I tried 1.48v in all volts that did not help so im gonna try a 1.5v tomorrow. But yes his KS has a VID @ 6 Ghz 1.439v which is pretty insane! My KS has VID @ 6 Ghz 1.469v EDIT: Changeing all adaptive voltages including VRM to 1.5v have resolved all issues! Record for me in CBR23 and CBR20 and CPU runs insane! Its a REAL treat to see the 6300 Mhz when active! ;) EDIT 2: Still not stable even at 1.52v ... What I am currently testing are 2 cores at 6300 Mhz instead of 3 .. I think that did the trick EDIT 3: 6300 Mhz are no matter what and how many cores causing random BSOD and sudden critical crash..This was with tested voltages all up to 1.55v. just to be sure it was not a power issue. Doom Eternal - any master level will SURELY tell you if your OC are stable. That engine are amazing and dont use 1 render thread but all the workers (cores) it can get. So its the ultimate OC test. Trust me! I can pass many test and benchmark but not Doom if the OC are unstable without you know it. I modified the OC profile to have 4 x 6200 Mhz and 8 x 6000 Mhz... THAT works ! this time tested throughly... several Doom Masterlevels passed! ;) Im gonna keep that and not pursue 6300 Mhz for now. Also my CPU now max use 1.46v but are almost all the time at 1.439v. Very very good settings. I think I know what the problem was. LLC 6 are simply not strong enough and have to much vdroop. And the very peak of the limits of the CPU 6300Mhz it does not work on my CPU. The other times I used Scatterbench examples I used LLC8. And that always work but is max LLC. I wanted to get away from that to get better performance overall. LLC6 NEVER worked on my regular K on Scatterbench examples.. I always used LLC8. But this profile are now rock solid with 4 cores at 6.2Ghz! and all core in game 8 x 6 Ghz at LLC6 and its sustainable with great temps etc. Runs amazing and what a treat to see my games and Doom Eternal running all core 6Ghz. Scatterbencher has an extreme bin of an KS. SO thats why he can do the very final step where I cant at least with these power settings.
@@lassekristensen385 hello sir, could you tell me please, was you changing Perf. core specific ratio limit , as athletic did ? want to test your settings….
@SkatterBencher Thanks for the amount of work you have put in. I have one question. I'm using OC Strategy 3 with the sameish settings as you. 5.7Ghz all core under heavy load gets me to 85c. Should I go for 5.8Ghz and increase the adaptive tubo mode voltage?
Thank you for answering my previous question. May I ask some more detailed questions in order to better understand what is happening and possible on my *delidded* 14900KS on a MoRa 420 (so not water temp limited) please, so that everyone can better understand some of your core multiplier settings ? : 1) At 320W default limit under R23 my P Core 7 is 20C hotter than the coolest core, with core 6 rather hot as well. Are these adjacent to the first two groups of E Cores? Reducing P Core 7 by 200MHz drops its temperature by 10C, from 100 to 90. I notice you often reduce two core multipliers and wondered if this was based on observed temperature or v/f curve. 2) Relating to question 1, if I were to plot all possible 13 v/f curves for any specific chip would that indicate the hottest cores by those needing the highest voltage for a multiplier? I know that the SVID requested will be the highest of the values. 3) Finally, is there any general guide to what undervolt is possible? Intel obviously program a margin. Is it something like the 10% that you seem to use? Thank you, and keep up the great overclocking guides. I doubt that anyone outside of Intel has your obvious expertise! P.S. I wonder if within your own overclocking utility it would be possible to extract and save all 13 v/f curves by independently adjusting them and saving values, just as can be done manually, but in a very time consuming manner, using XTU.
Hi Colin, thanks for the kind words. I will try to answer as best as possible: 1a) The reason I limit the ratio of specific cores is not due to temperature or observed V/F curve. It is simply because those cores aren't stable at the highest defined Turbo Ratio Limit for 1/2-active cores. By limiting specific cores, I can still have my other cores hit those peak Turbo ratios. 1b) I'm not sure the reason why they're hotter are due to proximity of E-cores. Some cores might just be slightly leakier due to manufacturing variation. 2) Not necessarily. Some cores run lower voltage but hotter and others vice versa. That can be due to current leakage. V/F curve a good general predictor of core capability but, as we've seen with ASUS' SP values, it's not a defining predictor. 3) Great question. Think about it this way: if there was a general rule that could reliably predict the undervolt margin, then why wouldn't Intel leverage that to build faster or more power efficient chips? The reality is that Intel spends a great amount of resources to find what chips are capable of since higher performance (or perf/watt) means they can charge a higher price. That said, evidently there is margin as we're able to undervolt to a certain degree. But it's not easy to give any general rule other than, for example, "try 50mV and see if it works" just like we'd say to AMD Ryzen overclockers "Try -15 curve optimizer and see if it works." As for the software question: unfortunately extracting the V/F curves automatically is pretty difficult because you need to adjust several settings (like AC LL, VR LLC, TVB, etc) which cannot be changed at runtime. ASUS does a pretty good job providing some information through BIOS, but the only way to get accurate data at the moment is by doing it manually.
@@SkatterBencher Thank you for the very detailed response. I'm attempting to use XTU to get the curves but hitting the problems you mentioned. I'll try again with Bios defaults and no TVB. If not then it's a very long process in the BIOS.
Followed this guide with lowered P cores clocks because 13900kf, works well with Z790 Apex/DDR5 gskill 8000mt sticks, now i need to fine tune those ram sticks to lower the latency a bit hopefully it works
thanks for replying, I wasn't expecting it :D anyway, a very strange thing. Even with the stock settings, if I launch Occt, I have no errors until I connect any USB stick. errors come out but it doesn't crash. if I run other tests no type of error. absurd, every time I connect a USB occt device under stress it gives me errors even with stock values... maybe I have some other problem that causes instability in the overclock and I have no idea what it could depend on.
First to comment! Great video as I'm taking a chance and exchanging the 13900K($579 USD) for the KS($699 USD). I sure hope to get some performance uplift in MSFS a very demanding title no doubt. I'm currently on the 12900K and it's doing ok. Even a few more FPS will make me happy💖
Thanks for these guides, they're amazing at explaining what the settings do. I tried strategy #3 with my 14900KS and it runs extremely cool and stable. I know I'm using the 13900KS numbers and I could go higher, but everything is great and running conservatively. Also running 6000MT/S DIMMS and I used my 6Ghz as the values. However, what is strange is that I am getting only 22K scores in CB23, whereas before I was running 40K + although not stable in games. In your example, you're showing decent scores in Cinebench so I must have done something wrong. Volts are low, temps are low, I show boosting to 6.2 Ghz on 4 cores. Everything seems fine and I am happy with this, but feel like I may be leaving something on the table using such conservative numbers. (Latest Asus BIOS 1202 I think) I also didn't reset everything to defaults prior to starting, and was running Intel Failsafe before. (I know, this is wrong and I ran this for a day) Wonder if there are any insights to my situation.
I share my experience, it may be useful to someone, I solved it by disabling CEP settings and putting six cores at 6.2ghz and 2 cores at 6.1ghz and all ecore at 4.5gh. I had to increase AC/DC to 0.15 now it is stable with occt and cinebench 23. Cinebench done 42628 as point.
What should SVID Behavior be set to? Auto? When mine is set to best case scenario it becomes unstable when trying to play a game. But if I set it to typical it goes up to 1.5v core.
TLDR: Auto Long explanation: SVID Behavior adjusts the AC/DC LL (LoadLine), where best case is AC LL=0.01 (lowest) and worst-case some higher value. If AC LL = 0.01, then the voltage VID requested by the CPU to the voltage regulator will match 1:1 the CPU's factory-fused voltage-frequency curve. If AC LL > 0.01, then the voltage VID requested by the CPU to the voltage regulator will have a small (positive) offset to the voltage request. So the effective voltage will be slightly above the factory-fused voltage. If your VRM loadline is flat, so no Vdroop when the load increases, then setting AC LL=0.01 should be no issue. However, all motherboards have a VRM loadline configured. There are specific reasons for that, primarily related to small spikes in voltage overshoot and undershoot. The effect of VRM loadline is that the voltage regulator will manage the output voltage according to the load. For a given requested voltage, the effective output voltage will be lower as the load increases. The real-world effect of AC LL = 0.01 and a VRM LLC is that the effective voltage under a high load will effectively be lower than the CPU factory-fused voltage-frequency curve. For example, let's say the CPU needs 1.2V for 5.5G for stability. - If AC LL = 0.01, then CPU requests 1.2V from the voltage regulator - If VRM LLC = flat, then VRM will output 1.2V to the CPU (but there will be overshoot) - If VRM LLC = not flat, then the VRM will output 1.2V at no load and perhaps 1.15V at maximum load. That will cause instability of our CPU needs 1.2V. While some CPUs may be able to run stable when undervolted, not all can. If you wish to use SVID best case, you can also change the VRM LLC to a less steep loadline. Hope this helps :)
@@SkatterBencher thank you for the explanation. I’m on the Asus ROG Maximus Z790 APEX with a 13900ks(overall SP 113, P core SP 123 E Core SP 96 MC SP 83). With SVID behavior set to Best Case, I cannot get a stable overclock, with SVID behavior set to Trained(Ai set) the max stable OC I can get is 60x/60x/58x/58x/58x/58x/58x/58x(1-8active cores)And LLC of 6. With SVID Behavior set to “Typical” and a Global SVID voltage negative offset of -.02v,LLC of 4, TVB enabled, TVB voltage optimizations disabled, I can get 61x/61x/60x/59x/59x/59x/59x/59x with e cores synced to 4.3Ghz. Should I even try to raise the E Cores if my main focus is gaming, which is usually medium cpu loads? Thank you! 😊
Hey man great work I can see that you are a master and you know what you talking about ! Do you think I can take your 3rd strategy for the over clocking ? I have the z790 extreme and an z73 nzxt pull/push with kryau extreme thermal past ! My memory same as your ? I think the 3rd strategy will work well ?
I think so, yes. The process followed in OC Strategy #3 is the most basic for manual tuning. Your results may differ of course, but finding out is a large part of the overclocking fun!
Hello, and thank you for your video. Although your video seems perfectly complete, my mastery of the English language does not allow me to understand everything. I have a 13900K on the same motherboard as in your z790 Maximus hero video. CPU and GPU cooled by EK water cooling. I would like to optimize my OC and check that I am using my hardware correctly. Is there a discord or forum or chat so that I can translate.
You can refer to my blog post: skatterbencher.com/2023/01/22/skatterbencher-53-intel-core-i9-13900ks-overclocked-to-6300mhz/. The content is pretty much identical to the video, but in written form :)
I occasionally do paid consulting work, though the terms and fee may not be desirable by everyone skatterbencher.com/contact/. Alternatively you can post your questions in the SkatterBencher discord channel and I'll try to reply when I can :)
Hey @SkatterBencher I got a serious question and I apologize if you have covered this before. With ALL of your overclocks, your temps are atrocious. I realize it does 0 harm to run 13th gen and i9 at 95C-105C, however IT DOES throttle the hell out of your speed and continuously retard clocks back. So why not work in "undevolt" mechanics and reduce some of that heat, so that you arent TJ Max'ing directly into down throttle? Wouldn't you see FAR SUPERIOR results if you managed to overclock your mem+cpu and retain turboboost? I would think your results would be FAR superior if you managed to keep temps hovering between 85C-95C and retained your turbo clock throughput during the entire test bench window (and or indefinitely?). I've seen people run HWINFO64 and run all 8 P-Cores at 6GHZ with less than 300 WATTS and temps all under 95C. With your tests and results, your always averaging 100C with max 100C and your P-Core Clocks all at like 5.5 or 5.7 depending on which OC method you used. This tells me your throttling back significantly? Maybe I am entirely wrong hear and I hope I am, in which you could explain why and how please. Because I just dont see the benefit or gain from ANY of your OC"s. ALL of them look like they yield HORRRRRRRRIBLE results from glancing at your min/max in HWINFO and your scores in general. What am I missing here?!?
I only report in detail on the worst-case stability tests (Prime95 Small FFTs), not other benchmarks. So the temperature in, let's say games, usually wouldn't hit 100C and it wouldn't throttle the CPU frequency.
Great explanation. For Raptor Lake on Asus motherboards the V/F Curve for the P Cores is directly visible, but how do you get the E Core and Ring curves? Any help would be much appreciated.
The only way is manual checking. Set P-core to 8X and then the VccIA voltage is determined by either E-Core or Ring. Also set ring to 8X and then the VccIA voltage is determined by the E-Core.
I use a different strategy now because I think this per core strat. is not good in the long run and can cause a lot of instability. It turned out that my cooling was not sufficient for those high clocks! What I did was I redid all my water blocks and LM and reseated the block on the GPU. Then I bought a Monster Radiator (yes that is its name! :) MO-RA 420 PRO.. Its massive with 2 extra pumps! This enabled me to run 6000 GHz all core with a OCTVB down clip at 70c and 80c. But the cooling is SO efficient that the CPU never gets to 70c. Only in extreme cases like CB R23. Then it clips to x58. But ALL games, even the most CPU intensive its now at 6 Ghz @ 48-58c max. Pretty incredible! (Direct Die) Its runs better than ever with not one single crash so far! (running 14 days!) Before I think what happened was occasionally it hit the ceiling and then made errors and BSOD/ Crash and or instability. So the no 1. thing is make sure the cooling can handle it. You may call it old school but all core is king imo. You don't really get anything useful out of per core imo. Also before I could not run LLC6 on 6 Ghz but its magically working now with new cooling! Please write if you want to know more. So my final settings are LLC 6, VRM Frequency manual 800hz, Powerphase and the other one on Extreme and VRM OFFSET Vcore +0.050v .. that runs the CPU @ 1.439/1.44v 100% stable and now also cool! ;) My SP is 108 and Cooling is 198
After unfortunatly damageing and kill my 13900K , I got my 13900KS today. I must admit although it runs GREAT! That I am very disappointed that its noway near as good as the example in the video! Yours has a VID @ 6 Ghz = 1.428v ... I think mine has like 1.468v or maybe even 1.478v. Thats quite a lot more ! :( Very disappointing now that I though I was a member of the "exclusive" club. I will try this 6.3 Ghz example though. I am runing a modded profile of yours for my previous 13900K but at 6.2 Ghz for my KS. So far no issues! I even run cache at 5 Ghz permanent. I will throw a EKWB Direct Die kit on it in a few weeks. Now the KS is watercooled with a Thermal-something contact frame.. (the other one than the TG) :) Love your content SkatterBencher! ;) EDIT 2: 6300 Mhz are no matter what and how many cores causing random BSOD and sudden critical crash..This was with tested voltages all up to 1.55v. just to be sure it was not a power issue. Doom Eternal - any master level will SURELY tell you if your OC are stable. That engine are amazing and dont use 1 render thread but all the workers (cores) it can get. So its the ultimate OC test. Trust me! I can pass many test and benchmark but not Doom if the OC are unstable without you know it. I modified the OC profile to have 4 x 6200 Mhz and 8 x 6000 Mhz... THAT works ! this time tested throughly... several Doom Masterlevels passed! ;) Im gonna keep that and not pursue 6300 Mhz for now. Also my CPU now max use 1.46v but are almost all the time at 1.439v. Very very good settings. I think I know what the problem was. LLC 6 are simply not strong enough and have to much vdroop. And the very peak of the limits of the CPU 6300Mhz it does not work on my CPU. The other times I used Scatterbench examples I used LLC8. And that always work but is max LLC. I wanted to get away from that to get better performance overall. LLC6 NEVER worked on my regular K on Scatterbench examples.. I always used LLC8. But this profile are now rock solid with 4 cores at 6.2Ghz! and all core in game 8 x 6 Ghz at LLC6 and its sustainable with great temps etc. Runs amazing and what a treat to see my games and Doom Eternal running all core 6Ghz. Scatterbencher has an extreme bin of an KS. SO thats why he can do the very final step where I cant at least with these power settings.
hey im looking to order asus z790 hero and i9 13900ks this week, you had any bios problem or crash problem ? just want to make sure :D thanks for your video btw
Because the onboard embedded controller (EC) needs data on your cooler's capability to handle high loads. Running Cinebench helps the EC understand how fast and at which power consumption the cooler keeps the CPU at a certain temperature. Without the data, the algorithm is flying blind
Following the 6.2ghz settings.. it gives me BSOD.. the only thing different in my system is my ram.. using 2x32gb kingston fury 6000mhz OC to 6800mhz.. Everything else i tried to follow ur guide.. nothing worked.. then i copy and paste ur settings and still the same.. Its not even booting to windows.. so its not even under heavy load. Keep getting different BSOD Clock watchdog timeout system thread exception not handled kmode exception not handled irql not less or equal
5 to max 10 FPS in Call of Duty MW2 (3440x1440) to me isn't worth the CPU running +15°C or 85°-90° C plus noise. Especially when running my RAM to DDR5 6800 XMP (I) has basically the same results even when running two sets vs a 4x set at DDR5 6600 has the same result. CPU-Z j8p8ur
Thank you for this wonderful guide, however the stock values you listed for the 13900KS at around 3:40 is incorrect and misleading. The correct stock values within Intel's specs are as follows: PL1 = 150W PL2 = 253W TAU = unlimited IccMax = 307A (or 400A w/ Extreme Config profile) You can confirm this on Intel's site and datasheets.
Since Alder Lake, for K processors, the default specification for Turbo Boost 2.0 is PL1=PL2. For the 13900KS, Intel also prepared an "extreme" specification which has PL1=PL2=320W. Unfortunately I had already finished my video before the new spec was released.
Amazing video! Can you give any advice on undervolting/minimizing the power draw of my system? I run a 16-thread workload on my E-cores and want to minimize their power usage, but when I increase my P-core speeds, it raises my E-core voltages - and they use as much power as if I'd raised their multipliers! I'm using XTU and the voltage offsets for E-cores don't seem to do anything. I'm currently running them at 26x (I mostly run at 26x lol), but the system starts using far more power if I raise any of my multipliers. I'm trying to make an optimized low-power mode. What settings should I be trying?
in the "simple manual overclock" you say "we get these results because of the aggressive undervolting" but looking at the guide i don't see any undervolting happening? all you do is setting the adaptive turbo mode cpu core voltage to 1.43500 which is higher than the VID tables 1.42800 which does makes sense since you are increasing the max boost, but where is the undervolting? I don't understand
I explain the undervolting part in the segment "VccIA Voltage Rail" th-cam.com/video/dpWw_5KfpIw/w-d-xo.html. The undervolting is not in the traditional sense, where we would set the a negative voltage offset, but using AC loadline and VRM loadline. TLDR: there's undervolting because the CPU voltage is below the fused V/F curve when under load, due to AC and VRM loadline configuration
Truly appreciate your time to answer my stupid question, i guess i somehow missed that part. I did use your guide to apply my overclock but i really wanted to try to understand why it was working... is there anywhere in the video where you disclose how you come to the decision in using 1.43500 as the adaptive voltage? i had to use 1.45600 adaptive voltage to stabilize my 14900k. Your sample seems very strong? @@SkatterBencher
I don't explicitly state it, I think, but it's based on the V/F curve. I just set it slightly higher than the default V/F point. This 13900KS was pretty decent if I recall correctly.
i followed your step by step guide for 6.3ghz in my z790 apex and 13900ks. I have some slight instability, could you suggest me how to set IA AC/DC Load Line to get stability?
Assuming you're on the edge of stability, you can try increasing the AC loadline in steps of 0.05. This will cause the CPU to request higher voltages from the VR controller, thus (hopefully) improving stability. Don't go too high with the AC loadline though ... sometimes you just need to accept that the silicon lottery is not on your side :)
@@SkatterBencher ah damn i thought this was the 13600k video, got the comment box mixed up... Since TBM 3.0 is not available for the 13600k and thus the os will not prioritize the cores best to worst, the 600k will loose some single core performance on average, especially if there is a large gap between best and worst core, right? Now with the 7800x3D out, intel should release a driver update that enables core priorization for the 13600k
what values should you use for the core ratios for P-Cores and E-Cores when overclocking a 13900K? the 13900k guides you released seem to only OC the P-Cores but what if one wanted to OC both the P-cores and E-cores?
I have a really crappy 13900k chip cause my Asus “SP” is 98 with a cooler score of “166” and running this exact default test, I got a score of 38070 in R23 and max VID of 1.425 V. I’ve never been this unlucky with a chip before. Hopefully can get something stable without pushing a crazy VID.
I also had the 13900K and for that CPU I would highly recommed Skatterbenchers "6000 Mhz 13900K" I tried all profiles also and found that best for that particular CPU with that SP and powerdraw- ..6100 or his 6000 Mhz 13900K profiles runs great for that CPU. 6100 Mhz profiles will however require additional powerdraw and cooling. I found the 6000 Mhz profile to be best daily driver. And now I just configured this 6300 Mhz KS profile on my new 13900KS ... runs amazing! A real treat to see the 6300Mhz when active! I did however had to change all voltages to 1.5v to make it stable. My KS are not as high bin as skatterbenchers.
Bro I have SP 96 with my 13900k. That's the average of all 24 cores. Usually the E-Cores are what brings it down. Go to Extreme Tweaker/AI Features to see P-Core SP and E-Core SP. Mine are P-Core SP is 107 and E-Core SP 76. Average is 96. Unfortunately once the KS went into development, all of the higher binned 13900k chips became 13900ks chips so the later you got a 13900k, usually the worst it is. I'm unhappy with mine but it's whatever I guess.
Those asus AI OC features suck from my experience leave that off turn on XMP1 - Set your ram frequency then set disable all limits then boot into windows and use Intel Extreme Tuning app and use its speed optimizer to really boost your system. Just my 2 cents..
Sit at idle for a while, then run a heavy workload. I have used 3 ways to detect these issues: 1) Sit idle, then load Prime95 Small FFTs all-core (non-avx & avx) 2) Run Cinebench R23 with 30min time (between runs the CPU goes to idle briefly before running all cores again) 3) Use AI Benchmark (very varied workload that often catches this instability)
Hello Skatterbencher. So, I used your brilliant guide and 6300 profile on my new KS. I have some issues with Chrome suddenly closing and its unable to complete CBR20. I tried changeing all p cores adaptive voltage and vrm adaptive from 1.475v to 1.48v. Still same in CBR20- crash and usual bug report error. Any suggestions? Or should I just change to 1.5v? EDIT: Changeing all adaptive voltages including VRM to 1.5v have resolved all issues! Record for me in CBR23 and CBR20 and CPU runs insane! Its a REAL treat to see the 6300 Mhz when active! ;) EDIT 2: Still not stable even at 1.52v ... What I am currently testing are 2 cores at 6300 Mhz instead of 3 .. I think that did the trick EDIT 2: 6300 Mhz are no matter what and how many cores causing random BSOD and sudden critical crash..This was with tested voltages all up to 1.55v. just to be sure it was not a power issue. Doom Eternal - any master level will SURELY tell you if your OC are stable. That engine are amazing and dont use 1 render thread but all the workers (cores) it can get. So its the ultimate OC test. Trust me! I can pass many test and benchmark but not Doom if the OC are unstable without you know it. I modified the OC profile to have 4 x 6200 Mhz and 8 x 6000 Mhz... THAT works ! this time tested throughly... several Doom Masterlevels passed! ;) Im gonna keep that and not pursue 6300 Mhz for now. Also my CPU now max use 1.46v but are almost all the time at 1.439v. Very very good settings. I think I know what the problem was. LLC 6 are simply not strong enough and have to much vdroop. And the very peak of the limits of the CPU 6300Mhz it does not work on my CPU. The other times I used Scatterbench examples I used LLC8. And that always work but is max LLC. I wanted to get away from that to get better performance overall. LLC6 NEVER worked on my regular K on Scatterbench examples.. I always used LLC8. But this profile are now rock solid with 4 cores at 6.2Ghz! and all core in game 8 x 6 Ghz at LLC6 and its sustainable with great temps etc. Runs amazing and what a treat to see my games and Doom Eternal running all core 6Ghz. Scatterbencher has an extreme bin of an KS. SO thats why he can do the very final step where I cant at least with these power settings.
Hi friend which strategy are you using ? Are you using with TV on? And in the end what's your adaptive voltage used and what's your CPU so I would love to try your settings since my CPU isn't highly binded
@@khaos555 Hello friend! I use a different strategy now because I think this per core strat. is not good in the long run and can cause a lot of instability. It turned out that my cooling was not sufficient for those high clocks! What I did was I redid all my water blocks and LM and reseated the block on the GPU. Then I bought a Monster Radiator (yes that is its name! :) MO-RA 420 PRO.. Its massive with 2 extra pumps! This enabled me to run 6000 GHz all core with a OCTVB down clip at 70c and 80c. But the cooling is SO efficient that the CPU never gets to 70c. Only in extreme cases like CB R23. Then it clips to x58. But ALL games, even the most CPU intensive its now at 6 Ghz @ 48-58c max. Pretty incredible! (Direct Die) Its runs better than ever with not one single crash so far! (running 14 days!) Before I think what happened was occasionally it hit the ceiling and then made errors and BSOD/ Crash and or instability. So the no 1. thing is make sure the cooling can handle it. You may call it old school but all core is king imo. You don't really get anything useful out of per core imo. Also before I could not run LLC6 on 6 Ghz but its magically working now with new cooling! Please write if you want to know more. So my final settings are LLC 6, VRM Frequency manual 800hz, Powerphase and the other one on Extreme and VRM OFFSET Vcore +0.050v .. that runs the CPU @ 1.439/1.44v 100% stable and now also cool! ;) My SP is 108 and Cooling is 198
@@lassekristensen385 Very nice cooling solution indeed congrats! my sp is pretty close to yours is your 108 the final under prediction, what is your P core SP if I may ask? your cooling solution looks and sounds amazing!, I have mines undervolted to -0.05 with max voltage around 1.39 maybe 1.4 not far from yours my cooling grade is around the 169s range I'm using Thermaltake ultra mx2 not the best but can't complain, I have my P cores limited to the 6000mhz also I can get there if CPU not under load otherwise normally I would get to 5700mhz due to core 7th maybe could try 58x I plan to play around one more time soon and retest as folr the e-cores I have the last remaining to 42-3200mhz, I could get to 4600mhz which I did previously depending if I lower p-core to 56-57g LLC is also set to level 6
@@khaos555 Yes the 108 is final. P-cores are 117 and E--cores are 92. At first I also had E cores running at 4500mhz and cache/ring at 5000 MHz. BUT that could also lead to instability and unnecessary power draw. I then figured out that there really isn't anything to gain running the L3 at 5000 so I have it static at 4500 MHz not AUTO. The E-cores was the same. I have them on auto (x43) this is because it only gives slightly higher score in lets say a benchmark. Are they are supposed to be efficient. But my main reason is that when running "auto" on the vCore or Offset or Adaptive the CPU is hard coded to request voltage with L3 at 4500 and E cores at 4300. Its much much much easier to control and adjust when these two parameters are "auto". I don't like the mobo changing L3 to 50x, 45x or sometimes 46x.... Its super stable with these parameters on static 45x and auto 43x. And believe me I have messed A LOT with this ha-ha! Also because when running lets say a very hard bench or stress for the CPU even with this cooling the watt is 340w (IF I remember correct!) So I really don't want it to further increase! Bottom line is if I have these settings like this it enables me to run the 6 GHz all core in pretty much everything flawless and if I mess with the other parameters its kind of a constant fiddling and adjustment and its better suited for manual voltage which I prefer not to because I want all the power saving features and what not. I am using also a tool called Park Control I can highly recommend. I paid for full version. It disables core parking when I use the PC or game, but when idle in 60sec it goes to POWER SAVE profile which takes all cores including e cores down to x8 or x11. and that works perfectly on the 6 GHz profile. If you then touch mouse or keyboard FULL performance not core parking again. Really a sweet utility. If you mess around with e cores or Manuel voltage some of the power saving features are not working properly. And some OC are not happy to go so low and up in ratio, but it works with this setup. What I am trying to say is P core tuning is the way to go! ha-ha! :) Also only thing that matters in gaming etc. Its way more stable this way at least in very high clocks like 6 GHz. I have had profiles with x57-x58 and ring x50 and E core x45 and that also works. But I find that x60 P and rest stock is best for me and gaming :) And thanks with the cooling it was quite expensive and I decided to go with it because it should last "a lifetime" ha-ha! It can be carried over to PC after PC. The most "cool" thing about it is that I drilled two holes in the wall for WATER IN and OUT and put the RAD in next room! so no heat or noise at all in the main gamer office room. Its pretty cool and nice. I know its a little extreme but why not! :) Before I upgraded my cooling I was running x57 on P... I could if I really stretched it run x58 but it would be almost 100c in CBR23. So I found that x57 was the sweet spot before.
@@lassekristensen385 Oh ok my P sp is 116 one number less than mines pretty sad but based on our conversation it sounds about the same my e-core sp is also 92 so it seems for sure we can both have it higher I also did get higher on benchmarks like cinebench r23 , and yeah i know what you mean when it comes to power saving, on games like elder scrolls I only wasting less than 110w honestly normal browsing less than 100w around in the 70s , while gaming I only get max out around 79 degrees c only for a few short while not under load below 40 degress if ambient temp is low since I live in a place were we get all seasons, like this morning, it's a bit cool(rare for us being (september) otherwise when its humid max I get is around 81-2 degrees but as a short spike, otherwise I expect to be around in the 40-50 degrees while browsing which i can't complain with p cores hitting 5700mhz which it does hit 6000mhz sometimes, but because I do have the cpu with some load I ran over 100 tabs on chrome if that counts as one haha, and some apps running in the background which is why I decided to keep the e-cores(main plan was to disable half 8 cores to keepo the other half) which also works but then i found overclocking it wasn't much higher but I did find another strategy so I will see and then decide if it's worth or not, in the end I will keep this configuration or another I have about 3 profiles saved on my z790 apex motherboard on the bios, currently using my second profile just because under this profile i'm running the e-cores to 4300 on more than 90% while the remaining 4 are running at 4200mhz on my first profile if I set the e-core to auto then 90% also runs at 4300mhz but the remaining 4 are running at 3200mhz
Great video with in depth explanation as usual. I have a question though . Do you have any idea of how i will make my 4x16 G skill ddr5 6400mhz sticks to work with xmp II on z790 maximums hero with 13900k? I bought them as 2x2x16 kits separately and i run 0813 bios and also i cant see the xmp tweaked choice only xmp1 and 2. Thanks
@@maegnificant "nope" doesn't say much, Framechasers is notorious for not doing hardly any stability testing besides opening a game and calling it a day.
If we're going to be precise, it's: - 6.1 GHz all core up to 75C - 5.9 GHz all core between 75C and 85C - 5.8 GHz all core between 85C and 100C - ~5538 MHz all core at 100C (P95 no AVX) - ~5281 MHz all core at 100C (P95 AVX) skatterbencher.com/wp-content/uploads/2023/01/Slide113.png
Looking at single core performance. I'd like to see single core clocks under load like Cine or 3DMark CPU profile. I may have missed that data in the video, can someone help? Skatter what are your impressions of single core performance and speeds? Multi cores looks good for performance improvements but inconclusive for single core? Thanks very much
Man i tried this and I was getting an overvoltage warning error with your advanced methord for my 14900k on my Z690 asus TUF. Was a bit scary. Any idea what i could possibly have done wrong?
For some reason my Z790 Hero when I have everything at default with MultiCore Enhancement disabled my 13900KS is not sticking to it's default PL1 AND PL2 Limit as it is setting them both of the limits to 320.0W for both of them. So I can not get the chip to run at stock could this be a fault with the bios ? I am using Bios ver. 0813
After I finished my video, Intel released a new Extreme Power Delivery specification for the 13900KS processor which has PL1=PL2=320W, not 253W. So your default specifications are correct www.coolaler.com.tw/coolalercbb//INTEL13TH/13900K/ClipKS_1.jpg
Do you think leaving it on Strategy 2 (AI Overclock Method) for longer periods of time will return higher and higher performance (benchmarks) over time as opposed to setting it and testing it right away, as it learns over periods of time and periods of use and adjusts/tweaks the overclocking settings on it's own w/ learning?
I dont think so, it will only make small voltage adjustments based on cooler score. I think its based AI OC is on fused VID, SP quality along with cooler score of corse.
No idea what happened to 0801, will ask around. If you want to try 0801, I've uploaded the file to my blog skatterbencher.com/wp-content/uploads/2023/01/ROG-MAXIMUS-Z790-HERO-ASUS-0801.zip
@@SkatterBencher Interesting, thank you sir! If I use Corsair iCue h170i elite AIO, what's the max overclock should I aim for in your opinion? I got this very same MB, ROG Maximus Hero. Or should I not think of overclocking at with an AIO? You should do a tutorial on over clock the 13900KS with an AIO plz!
I made the 6300 Mhz profile successfull on my KS.. it did however require I upped all voiltages to 1.5v but runs amazing! My previous 13900K could not reach 6200-6300 successfull .. I used Skatterbenchers 6000 and 6100 profiles with success on that CPU.
@@newdawn005 I am using 1.5v on all the adaptive voltages .. runs amazing! This was required for my KS to be stable with this 6300 Mhz profle.. 1.475v was too low unfortunayly for my KS. But 1.5v is fine.. personally I would not go higher.
I stuck with this board for my new build, waiting on PSU but can't work out how to fit my M. 2 5.0 ssd at 5.0 speed and not sharing pcie with GPU. Any one know Im really stuck..
I'm still unstable, I had to put all p-cores at 6.1ghz and AC/DC line 0.20 on both to be stable. Do you have any further advice? I can't go above 0.20 AC/DC because then the temperatures rise too much
mine doesn't have the xmp teak settings only xmp 1 and 2 and with the auto ai my pc keeps restarting?? I got asus maximus hero z790 13900k 4090 I did exactly what you said and when I go run cinebench again after the ai auto tune pc just reboots ??? am on the updated bios and win 11
Sounds like the overclock is not quite stable. You can try either of the following options: 1) Reset to defaults, go in the OS and run Cinebench R23 for 30min to "train" the motherboard. Then try AI OC again. 2) If too low voltage is the root cause of the instability, you can also try adjusting to a less steep VRM LLC
@@SkatterBencher okay tbh idk how to done none of this am a noob idk how to oc my gpu either but ill try that again and see what happens u know why I dont have xmp tweaked? I only have xmp 1 and 2
45200? Ive never seen such a high score.. I just beat my own personal with this OC profile in the video. 43110. I dont quite understand how you can make 2000 additional... there is only a "budget" of heat and power draw available so unless you are using xtreme cooling of some sort it seems incredible!
@Lasse Kristensen do you have Ln2? Now I am using 13900ks 8800CL36 stable Hci 500% and 9000mhz hci 60% unstable it but all bench Geek3 to 6 is passed not good performance bench than 8800Cl36
@@Hwbot-Philip-Park No no no, only normal water cool. I have G Skill 6000 and I overclocked them to 6600 Mhz with 1.46v @ the same timings! 36-36-36-76. I know its Samsung B Die.. But I dont seem to be able to get them higher? Any tricks for B Die? OR is this my limits?
im pretty new to this but for some reason when i remove all limits on this board it gives me around 340 watts in HW monitor and its unstable i believe partly because of the multi core enhancement even on optimized defaults. anyone have any ideas?
Why only 12.6K subscribers. This TH-camr is an overclockers dream. ❤ He explains everything in detail and it all makes sense. Cream of the crop compared other overclockers who make Videos.👍👍👍👍👍👍👍👍
Because his target audience is not made of that many people, but inside the community this guy is a rock start he is widely known and respected.
Best overclocking channel in YT by far. Great content. Thanks.
Well explained with some of your OCTVB settings feels good to play at 6ghz , your the underrated king of the overclocking youtube content, At overclocking you are already king :D
DDR5 overcloking guides or DDR4/5 in general would be awsome.
Thank you! No tuber would do this kind of in depth guide and tutorial like you. It helped average OCers greatly! Appreciate!
Thanks
Fantastic deserves reward very good 🎉
Absolutely fantastic, in-depth video. I very much appreciate the time and effort you put in it. Thank you so much!!
Great video and guide, helped me a lot with my 13900ks but on a z690. I was having hight temps out of the box and initially only applied a -0.100 offset. After applying your settings now I have great temps and clocks.
do i need to set up voltage by my own, if i’m choosing Manual OC strategy. copied everything and blue screen only… (z790 apex)
This was exactly what I was looking for to understand some of the different oc options. Thanks.
Really appreciate the thoroughness of this guide. Very useful information presented in a manner that makes sense and is easy to understand. Keep up the amazing work!
Best teacher for overclocking. Thanks for putting this together
The amount of work you put into this is staggering. Thank you so much for doing this!
Thanks for the kind words! :)
Wow. I learned so much today. I have previously overclocked just by turning numbers up slowly and praying for the best lol.
The Asus Extreme mobo I have now has so much added functionality that I decided to try and learn what everything actually means. This video and your content in general has been astonishingly helpful. Thank you.
I just copied all of these settings to my rig with a 13900KS, ProArt z790 being cooled by a arctic liquid freezer 360. Everything works great! I get roundabout 42,000 points in cinebench, have been using the pc with this overclock for over 1 month now and never had a blue screen. Saw on the comments that people were saying that it doesn't work because he has had luck with his chip, getting a sp predicion of 115 while most people get 107, I GET 105 AND EVERYTHING IS ROCK-SOLID. THANK YOU SO MUCH !!
NEXT LEVEL knowledge share.... keep up the fantastic work!
Great in depth video, this guy deserves way more subscribers!
It works flawlessly, thank you so much!
Great info thanks.
uppdate pleas:) after intel microcode. can i do the same?
I asked the same question hope he has time to answer.
Great channel and super great videos.
this is pure gold. wow. subscribed !
very impressive video's u make! u deserve more! thanx for this great video!
Youre brilliant sir
The selection of the per core limit in 38:20 may only be applicable to your specific CPU and not to others.
You also seem to have a highly selected and binned CPU with an SP prediction of 115 whereas the average is about 107.
So it is very likely that everyone with a lower binned processor will get an unstable system even when exactly following your video.
That's my case with an i9-13900KS at SP 108.
Besides this, my comment is not meant to be a rant about your video which gives a fantastic insight into OC and the Rapor Lake Platform.
Thank you for this! 👍🙂
Mine is not 100% stable at those settings.. I have a KS 108 (p cores at 117). But CBR20 cannot complete and my browser closes random. I tried 1.48v in all volts that did not help so im gonna try a 1.5v tomorrow. But yes his KS has a VID @ 6 Ghz 1.439v which is pretty insane! My KS has VID @ 6 Ghz 1.469v
EDIT: Changeing all adaptive voltages including VRM to 1.5v have resolved all issues! Record for me in CBR23 and CBR20 and CPU runs insane! Its a REAL treat to see the 6300 Mhz when active! ;) EDIT 2: Still not stable even at 1.52v ... What I am currently testing are 2 cores at 6300 Mhz instead of 3 .. I think that did the trick
EDIT 3: 6300 Mhz are no matter what and how many cores causing random BSOD and sudden critical crash..This was with tested voltages all up to 1.55v. just to be sure it was not a power issue. Doom Eternal - any master level will SURELY tell you if your OC are stable. That engine are amazing and dont use 1 render thread but all the workers (cores) it can get. So its the ultimate OC test. Trust me! I can pass many test and benchmark but not Doom if the OC are unstable without you know it. I modified the OC profile to have 4 x 6200 Mhz and 8 x 6000 Mhz... THAT works ! this time tested throughly... several Doom Masterlevels passed! ;) Im gonna keep that and not pursue 6300 Mhz for now. Also my CPU now max use 1.46v but are almost all the time at 1.439v. Very very good settings. I think I know what the problem was. LLC 6 are simply not strong enough and have to much vdroop. And the very peak of the limits of the CPU 6300Mhz it does not work on my CPU. The other times I used Scatterbench examples I used LLC8. And that always work but is max LLC. I wanted to get away from that to get better performance overall. LLC6 NEVER worked on my regular K on Scatterbench examples.. I always used LLC8. But this profile are now rock solid with 4 cores at 6.2Ghz! and all core in game 8 x 6 Ghz at LLC6 and its sustainable with great temps etc. Runs amazing and what a treat to see my games and Doom Eternal running all core 6Ghz. Scatterbencher has an extreme bin of an KS. SO thats why he can do the very final step where I cant at least with these power settings.
@@lassekristensen385 hello sir, could you tell me please, was you changing Perf. core specific ratio limit , as athletic did ? want to test your settings….
Thank you SB!
I have a MO-RA3 420 on a KS but also tempted to get the TEC delta2 as well for it 🤔
Wow amazing video. I was able to get mine stable on AI at 61,61,60,60,58,58,58,56. The last core can’t do 5.8ghz
Very informative video.
@SkatterBencher Thanks for the amount of work you have put in. I have one question. I'm using OC Strategy 3 with the sameish settings as you. 5.7Ghz all core under heavy load gets me to 85c. Should I go for 5.8Ghz and increase the adaptive tubo mode voltage?
As an overclocker, I'd say give it a try and see if you can make 5.8G work :)
very good guide. Sub'd. Perhaps if a 13900KS goes on sale, I'll upgrade and use this guide to optimize the setup.
Thank you for answering my previous question. May I ask some more detailed questions in order to better understand what is happening and possible on my *delidded* 14900KS on a MoRa 420 (so not water temp limited) please, so that everyone can better understand some of your core multiplier settings ? :
1) At 320W default limit under R23 my P Core 7 is 20C hotter than the coolest core, with core 6 rather hot as well. Are these adjacent to the first two groups of E Cores? Reducing P Core 7 by 200MHz drops its temperature by 10C, from 100 to 90. I notice you often reduce two core multipliers and wondered if this was based on observed temperature or v/f curve.
2) Relating to question 1, if I were to plot all possible 13 v/f curves for any specific chip would that indicate the hottest cores by those needing the highest voltage for a multiplier? I know that the SVID requested will be the highest of the values.
3) Finally, is there any general guide to what undervolt is possible? Intel obviously program a margin. Is it something like the 10% that you seem to use?
Thank you, and keep up the great overclocking guides. I doubt that anyone outside of Intel has your obvious expertise!
P.S. I wonder if within your own overclocking utility it would be possible to extract and save all 13 v/f curves by independently adjusting them and saving values, just as can be done manually, but in a very time consuming manner, using XTU.
Hi Colin, thanks for the kind words. I will try to answer as best as possible:
1a) The reason I limit the ratio of specific cores is not due to temperature or observed V/F curve. It is simply because those cores aren't stable at the highest defined Turbo Ratio Limit for 1/2-active cores. By limiting specific cores, I can still have my other cores hit those peak Turbo ratios.
1b) I'm not sure the reason why they're hotter are due to proximity of E-cores. Some cores might just be slightly leakier due to manufacturing variation.
2) Not necessarily. Some cores run lower voltage but hotter and others vice versa. That can be due to current leakage. V/F curve a good general predictor of core capability but, as we've seen with ASUS' SP values, it's not a defining predictor.
3) Great question. Think about it this way: if there was a general rule that could reliably predict the undervolt margin, then why wouldn't Intel leverage that to build faster or more power efficient chips? The reality is that Intel spends a great amount of resources to find what chips are capable of since higher performance (or perf/watt) means they can charge a higher price.
That said, evidently there is margin as we're able to undervolt to a certain degree. But it's not easy to give any general rule other than, for example, "try 50mV and see if it works" just like we'd say to AMD Ryzen overclockers "Try -15 curve optimizer and see if it works."
As for the software question: unfortunately extracting the V/F curves automatically is pretty difficult because you need to adjust several settings (like AC LL, VR LLC, TVB, etc) which cannot be changed at runtime. ASUS does a pretty good job providing some information through BIOS, but the only way to get accurate data at the moment is by doing it manually.
@@SkatterBencher Thank you for the very detailed response. I'm attempting to use XTU to get the curves but hitting the problems you mentioned. I'll try again with Bios defaults and no TVB. If not then it's a very long process in the BIOS.
Would like to see a RAM Oc guide 😊
Thank you
Why doesn't it let me edit the PL1 and 2 limits? It's stuck to Auto
Followed this guide with lowered P cores clocks because 13900kf, works well with Z790 Apex/DDR5 gskill 8000mt sticks, now i need to fine tune those ram sticks to lower the latency a bit hopefully it works
Are you able to give an update to this with intelss new update? Or nothing changes.
You're one good nerd ... Glad I found your channel bro.
Welcome aboard!
at 38:29 how would you be able to arrive at the specific ratio limits per core? Is there a recommended way of getting that information?
nevermind, found it in a reply in the previous 13900k video :)
@@ddd1million Where?
@@ddd1million, give us a link, please
thanks for replying, I wasn't expecting it :D anyway, a very strange thing. Even with the stock settings, if I launch Occt, I have no errors until I connect any USB stick. errors come out but it doesn't crash. if I run other tests no type of error. absurd, every time I connect a USB occt device under stress it gives me errors even with stock values... maybe I have some other problem that causes instability in the overclock and I have no idea what it could depend on.
First to comment! Great video as I'm taking a chance and exchanging the 13900K($579 USD) for the KS($699 USD). I sure hope to get some performance uplift in MSFS a very demanding title no doubt. I'm currently on the 12900K and it's doing ok. Even a few more FPS will make me happy💖
ju r knots XD
You should NOT expect a miracle! 100-200 Mhz increase ingame means basically nothing!
Thanks for these guides, they're amazing at explaining what the settings do. I tried strategy #3 with my 14900KS and it runs extremely cool and stable. I know I'm using the 13900KS numbers and I could go higher, but everything is great and running conservatively. Also running 6000MT/S DIMMS and I used my 6Ghz as the values. However, what is strange is that I am getting only 22K scores in CB23, whereas before I was running 40K + although not stable in games. In your example, you're showing decent scores in Cinebench so I must have done something wrong. Volts are low, temps are low, I show boosting to 6.2 Ghz on 4 cores. Everything seems fine and I am happy with this, but feel like I may be leaving something on the table using such conservative numbers. (Latest Asus BIOS 1202 I think) I also didn't reset everything to defaults prior to starting, and was running Intel Failsafe before. (I know, this is wrong and I ran this for a day) Wonder if there are any insights to my situation.
23K sounds pretty low even for conservative settings.
Can you double check with HWiNFO what are the core effective clocks during a Cinebench run?
Cool video. Hey Skatter, are you going to review the new Zen 4 X3D chip on launch day?
I won't review it but I'll definitely give overclocking a go if I have the opportunity :)
I share my experience, it may be useful to someone, I solved it by disabling CEP settings and putting six cores at 6.2ghz and 2 cores at 6.1ghz and all ecore at 4.5gh. I had to increase AC/DC to 0.15 now it is stable with occt and cinebench 23. Cinebench done 42628 as point.
What should SVID Behavior be set to? Auto? When mine is set to best case scenario it becomes unstable when trying to play a game. But if I set it to typical it goes up to 1.5v core.
TLDR: Auto
Long explanation:
SVID Behavior adjusts the AC/DC LL (LoadLine), where best case is AC LL=0.01 (lowest) and worst-case some higher value.
If AC LL = 0.01, then the voltage VID requested by the CPU to the voltage regulator will match 1:1 the CPU's factory-fused voltage-frequency curve. If AC LL > 0.01, then the voltage VID requested by the CPU to the voltage regulator will have a small (positive) offset to the voltage request. So the effective voltage will be slightly above the factory-fused voltage.
If your VRM loadline is flat, so no Vdroop when the load increases, then setting AC LL=0.01 should be no issue. However, all motherboards have a VRM loadline configured. There are specific reasons for that, primarily related to small spikes in voltage overshoot and undershoot. The effect of VRM loadline is that the voltage regulator will manage the output voltage according to the load. For a given requested voltage, the effective output voltage will be lower as the load increases.
The real-world effect of AC LL = 0.01 and a VRM LLC is that the effective voltage under a high load will effectively be lower than the CPU factory-fused voltage-frequency curve. For example, let's say the CPU needs 1.2V for 5.5G for stability.
- If AC LL = 0.01, then CPU requests 1.2V from the voltage regulator
- If VRM LLC = flat, then VRM will output 1.2V to the CPU (but there will be overshoot)
- If VRM LLC = not flat, then the VRM will output 1.2V at no load and perhaps 1.15V at maximum load. That will cause instability of our CPU needs 1.2V.
While some CPUs may be able to run stable when undervolted, not all can. If you wish to use SVID best case, you can also change the VRM LLC to a less steep loadline.
Hope this helps :)
@@SkatterBencher thank you for the explanation. I’m on the Asus ROG Maximus Z790 APEX with a 13900ks(overall SP 113, P core SP 123 E Core SP 96 MC SP 83). With SVID behavior set to Best Case, I cannot get a stable overclock, with SVID behavior set to Trained(Ai set) the max stable OC I can get is 60x/60x/58x/58x/58x/58x/58x/58x(1-8active cores)And LLC of 6. With SVID Behavior set to “Typical” and a Global SVID voltage negative offset of -.02v,LLC of 4, TVB enabled, TVB voltage optimizations disabled, I can get 61x/61x/60x/59x/59x/59x/59x/59x with e cores synced to 4.3Ghz. Should I even try to raise the E Cores if my main focus is gaming, which is usually medium cpu loads? Thank you! 😊
Hey man great work I can see that you are a master and you know what you talking about ! Do you think I can take your 3rd strategy for the over clocking ? I have the z790 extreme and an z73 nzxt pull/push with kryau extreme thermal past ! My memory same as your ? I think the 3rd strategy will work well ?
I think so, yes. The process followed in OC Strategy #3 is the most basic for manual tuning. Your results may differ of course, but finding out is a large part of the overclocking fun!
@@SkatterBencher thank you for your response ! My SP is 108 ! But I m novice in manual over locking !
I’m wondering if that’s optimal for gaming?
1.19V nice sample
Do you do any of these on MSI Z790 Motherboards? Thx
Can i use this same video to overclock my intel core i9-12900ks with my z790 Hero ????
I was think about buying this motherboard do you think it’s worth it?
My experience was pretty good, but then again I only use it for overclocking :)
@@SkatterBencher i will be using it for overclocking i was wondering if it was much better then the 790e
@@Zamal Good luck finding it in stock.
@@4gbmeans4gb61 its actually one of the few that are in stock rn in canada
@@Zamal Wow lucky, in the US All the stores are sold out, and Some are on Back order for another month. How much are they there? 629 US here.
I can’t change my cpu ghz to above to 4.0 it stays below at 3.0 and I have 149000
Hello, and thank you for your video. Although your video seems perfectly complete, my mastery of the English language does not allow me to understand everything. I have a 13900K on the same motherboard as in your z790 Maximus hero video. CPU and GPU cooled by EK water cooling. I would like to optimize my OC and check that I am using my hardware correctly. Is there a discord or forum or chat so that I can translate.
You can refer to my blog post: skatterbencher.com/2023/01/22/skatterbencher-53-intel-core-i9-13900ks-overclocked-to-6300mhz/. The content is pretty much identical to the video, but in written form :)
@@SkatterBencher Would you be interested in a personalized assistance mission (consulting)? I can give you my discord
I occasionally do paid consulting work, though the terms and fee may not be desirable by everyone skatterbencher.com/contact/.
Alternatively you can post your questions in the SkatterBencher discord channel and I'll try to reply when I can :)
@@SkatterBencher email ok
Hey @SkatterBencher I got a serious question and I apologize if you have covered this before. With ALL of your overclocks, your temps are atrocious. I realize it does 0 harm to run 13th gen and i9 at 95C-105C, however IT DOES throttle the hell out of your speed and continuously retard clocks back. So why not work in "undevolt" mechanics and reduce some of that heat, so that you arent TJ Max'ing directly into down throttle? Wouldn't you see FAR SUPERIOR results if you managed to overclock your mem+cpu and retain turboboost? I would think your results would be FAR superior if you managed to keep temps hovering between 85C-95C and retained your turbo clock throughput during the entire test bench window (and or indefinitely?). I've seen people run HWINFO64 and run all 8 P-Cores at 6GHZ with less than 300 WATTS and temps all under 95C. With your tests and results, your always averaging 100C with max 100C and your P-Core Clocks all at like 5.5 or 5.7 depending on which OC method you used. This tells me your throttling back significantly? Maybe I am entirely wrong hear and I hope I am, in which you could explain why and how please. Because I just dont see the benefit or gain from ANY of your OC"s. ALL of them look like they yield HORRRRRRRRIBLE results from glancing at your min/max in HWINFO and your scores in general. What am I missing here?!?
I only report in detail on the worst-case stability tests (Prime95 Small FFTs), not other benchmarks. So the temperature in, let's say games, usually wouldn't hit 100C and it wouldn't throttle the CPU frequency.
Great explanation. For Raptor Lake on Asus motherboards the V/F Curve for the P Cores is directly visible, but how do you get the E Core and Ring curves? Any help would be much appreciated.
The only way is manual checking. Set P-core to 8X and then the VccIA voltage is determined by either E-Core or Ring. Also set ring to 8X and then the VccIA voltage is determined by the E-Core.
@@SkatterBencher Thanks for the response. I guess I had hoped their was an equivalent map directly available 😢
Why didn’t set a ring/cache ratio in this video? I’m just wondering because I thought you would’ve set that to like 50 mhz.
I use a different strategy now because I think this per core strat. is not good in the long run and can cause a lot of instability. It turned out that my cooling was not sufficient for those high clocks! What I did was I redid all my water blocks and LM and reseated the block on the GPU. Then I bought a Monster Radiator (yes that is its name! :) MO-RA 420 PRO.. Its massive with 2 extra pumps! This enabled me to run 6000 GHz all core with a OCTVB down clip at 70c and 80c. But the cooling is SO efficient that the CPU never gets to 70c. Only in extreme cases like CB R23. Then it clips to x58. But ALL games, even the most CPU intensive its now at 6 Ghz @ 48-58c max. Pretty incredible! (Direct Die) Its runs better than ever with not one single crash so far! (running 14 days!) Before I think what happened was occasionally it hit the ceiling and then made errors and BSOD/ Crash and or instability. So the no 1. thing is make sure the cooling can handle it. You may call it old school but all core is king imo. You don't really get anything useful out of per core imo. Also before I could not run LLC6 on 6 Ghz but its magically working now with new cooling! Please write if you want to know more.
So my final settings are LLC 6, VRM Frequency manual 800hz, Powerphase and the other one on Extreme and VRM OFFSET Vcore +0.050v .. that runs the CPU @ 1.439/1.44v 100% stable and now also cool! ;) My SP is 108 and Cooling is 198
After unfortunatly damageing and kill my 13900K , I got my 13900KS today. I must admit although it runs GREAT! That I am very disappointed that its noway near as good as the example in the video! Yours has a VID @ 6 Ghz = 1.428v ... I think mine has like 1.468v or maybe even 1.478v. Thats quite a lot more ! :( Very disappointing now that I though I was a member of the "exclusive" club. I will try this 6.3 Ghz example though. I am runing a modded profile of yours for my previous 13900K but at 6.2 Ghz for my KS. So far no issues! I even run cache at 5 Ghz permanent. I will throw a EKWB Direct Die kit on it in a few weeks. Now the KS is watercooled with a Thermal-something contact frame.. (the other one than the TG) :) Love your content SkatterBencher! ;)
EDIT 2: 6300 Mhz are no matter what and how many cores causing random BSOD and sudden critical crash..This was with tested voltages all up to 1.55v. just to be sure it was not a power issue. Doom Eternal - any master level will SURELY tell you if your OC are stable. That engine are amazing and dont use 1 render thread but all the workers (cores) it can get. So its the ultimate OC test. Trust me! I can pass many test and benchmark but not Doom if the OC are unstable without you know it. I modified the OC profile to have 4 x 6200 Mhz and 8 x 6000 Mhz... THAT works ! this time tested throughly... several Doom Masterlevels passed! ;) Im gonna keep that and not pursue 6300 Mhz for now. Also my CPU now max use 1.46v but are almost all the time at 1.439v. Very very good settings. I think I know what the problem was. LLC 6 are simply not strong enough and have to much vdroop. And the very peak of the limits of the CPU 6300Mhz it does not work on my CPU. The other times I used Scatterbench examples I used LLC8. And that always work but is max LLC. I wanted to get away from that to get better performance overall. LLC6 NEVER worked on my regular K on Scatterbench examples.. I always used LLC8. But this profile are now rock solid with 4 cores at 6.2Ghz! and all core in game 8 x 6 Ghz at LLC6 and its sustainable with great temps etc. Runs amazing and what a treat to see my games and Doom Eternal running all core 6Ghz. Scatterbencher has an extreme bin of an KS. SO thats why he can do the very final step where I cant at least with these power settings.
hey im looking to order asus z790 hero and i9 13900ks this week, you had any bios problem or crash problem ? just want to make sure :D thanks for your video btw
I didn't experience any issues
Will this guide work with a 14900k on a Z790i Asus mITX mobo, and Corsair Dominator Titanium 7600MHz RAM sticks? 9:03
Yes, but the exactly frequency will depend on the cooling solution you're using.
why i need to run cinebench before enter AI overclock menus?
Because the onboard embedded controller (EC) needs data on your cooler's capability to handle high loads. Running Cinebench helps the EC understand how fast and at which power consumption the cooler keeps the CPU at a certain temperature. Without the data, the algorithm is flying blind
Following the 6.2ghz settings.. it gives me BSOD.. the only thing different in my system is my ram.. using 2x32gb kingston fury 6000mhz OC to 6800mhz..
Everything else i tried to follow ur guide.. nothing worked.. then i copy and paste ur settings and still the same..
Its not even booting to windows.. so its not even under heavy load.
Keep getting different BSOD
Clock watchdog timeout
system thread exception not handled
kmode exception not handled
irql not less or equal
Netbust engineers would be proud of Raptor Lake engineers. Easily 100'C on water
When you set the ACDC loadlines, do you ever get thunderstruck?
No but going down that rabbit hole certainly is a highway to hell
@@SkatterBencher Well played sir, well played 🤣🤣🤣
The Z790 tough boots with the core at 1.65-1.7 v following these settings something wrong with that board when using adaptive voltage
Any suggestions to doing this with a 13700k with the ASUS Z790 motherboard?
The same setup you have I'm building can this be used for workstation motherboard
5 to max 10 FPS in Call of Duty MW2 (3440x1440) to me isn't worth the CPU running +15°C or 85°-90° C plus noise. Especially when running my RAM to DDR5 6800 XMP (I) has basically the same results even when running two sets vs a 4x set at DDR5 6600 has the same result. CPU-Z j8p8ur
If you playing games, cpu undervolting will be better for you.
@@Typhon888 .. I have no noise now, so undervolting will only make it slower.
Thank you for this wonderful guide, however the stock values you listed for the 13900KS at around 3:40 is incorrect and misleading. The correct stock values within Intel's specs are as follows:
PL1 = 150W
PL2 = 253W
TAU = unlimited
IccMax = 307A (or 400A w/ Extreme Config profile)
You can confirm this on Intel's site and datasheets.
Since Alder Lake, for K processors, the default specification for Turbo Boost 2.0 is PL1=PL2. For the 13900KS, Intel also prepared an "extreme" specification which has PL1=PL2=320W. Unfortunately I had already finished my video before the new spec was released.
Amazing video! Can you give any advice on undervolting/minimizing the power draw of my system? I run a 16-thread workload on my E-cores and want to minimize their power usage, but when I increase my P-core speeds, it raises my E-core voltages - and they use as much power as if I'd raised their multipliers! I'm using XTU and the voltage offsets for E-cores don't seem to do anything. I'm currently running them at 26x (I mostly run at 26x lol), but the system starts using far more power if I raise any of my multipliers. I'm trying to make an optimized low-power mode. What settings should I be trying?
Hello , When i copy all ur setup in bios . I got an error saying core voltage error . I dont know why i just copy and paste.
With an rtx 4090 gpu?
in the "simple manual overclock" you say "we get these results because of the aggressive undervolting" but looking at the guide i don't see any undervolting happening? all you do is setting the adaptive turbo mode cpu core voltage to 1.43500 which is higher than the VID tables 1.42800 which does makes sense since you are increasing the max boost, but where is the undervolting? I don't understand
I explain the undervolting part in the segment "VccIA Voltage Rail" th-cam.com/video/dpWw_5KfpIw/w-d-xo.html.
The undervolting is not in the traditional sense, where we would set the a negative voltage offset, but using AC loadline and VRM loadline.
TLDR: there's undervolting because the CPU voltage is below the fused V/F curve when under load, due to AC and VRM loadline configuration
Truly appreciate your time to answer my stupid question, i guess i somehow missed that part. I did use your guide to apply my overclock but i really wanted to try to understand why it was working... is there anywhere in the video where you disclose how you come to the decision in using 1.43500 as the adaptive voltage? i had to use 1.45600 adaptive voltage to stabilize my 14900k. Your sample seems very strong?
@@SkatterBencher
I don't explicitly state it, I think, but it's based on the V/F curve. I just set it slightly higher than the default V/F point.
This 13900KS was pretty decent if I recall correctly.
config i9 13900k ?
i followed your step by step guide for 6.3ghz in my z790 apex and 13900ks. I have some slight instability, could you suggest me how to set IA AC/DC Load Line to get stability?
Assuming you're on the edge of stability, you can try increasing the AC loadline in steps of 0.05. This will cause the CPU to request higher voltages from the VR controller, thus (hopefully) improving stability.
Don't go too high with the AC loadline though ... sometimes you just need to accept that the silicon lottery is not on your side :)
Is there any reason to get the extreme over the hero? What additional use does it have?
For this kind of question, I would suggest you ping JJ from ASUS on Twitter :)
Dont get the extreme... get the apex
@@sagerdood how come?
@@ameerkhan3257 the apex has more functionality jayz 2 cenz has a video about it
@@SkatterBencher if you have good vrm cooling does it matter what your loadline is set at?
Great vid as usual bud!
Thanks Joe!
Isnt OCTVB for 1x and 2x active cores still limited by how high the WORST core on the chip can boost while retaining stabilty?
Not if you also configure the Per Core Ratio Limit correctly. That tool allows you to limit the maximum ratio of specific cores
@@SkatterBencher ah damn i thought this was the 13600k video, got the comment box mixed up...
Since TBM 3.0 is not available for the 13600k and thus the os will not prioritize the cores best to worst, the 600k will loose some single core performance on average, especially if there is a large gap between best and worst core, right?
Now with the 7800x3D out, intel should release a driver update that enables core priorization for the 13600k
what values should you use for the core ratios for P-Cores and E-Cores when overclocking a 13900K? the 13900k guides you released seem to only OC the P-Cores but what if one wanted to OC both the P-cores and E-cores?
I have a really crappy 13900k chip cause my Asus “SP” is 98 with a cooler score of “166” and running this exact default test, I got a score of 38070 in R23 and max VID of 1.425 V. I’ve never been this unlucky with a chip before. Hopefully can get something stable without pushing a crazy VID.
I also had the 13900K and for that CPU I would highly recommed Skatterbenchers "6000 Mhz 13900K" I tried all profiles also and found that best for that particular CPU with that SP and powerdraw- ..6100 or his 6000 Mhz 13900K profiles runs great for that CPU. 6100 Mhz profiles will however require additional powerdraw and cooling. I found the 6000 Mhz profile to be best daily driver. And now I just configured this 6300 Mhz KS profile on my new 13900KS ... runs amazing! A real treat to see the 6300Mhz when active! I did however had to change all voltages to 1.5v to make it stable. My KS are not as high bin as skatterbenchers.
Bro I have SP 96 with my 13900k. That's the average of all 24 cores. Usually the E-Cores are what brings it down. Go to Extreme Tweaker/AI Features to see P-Core SP and E-Core SP. Mine are P-Core SP is 107 and E-Core SP 76. Average is 96. Unfortunately once the KS went into development, all of the higher binned 13900k chips became 13900ks chips so the later you got a 13900k, usually the worst it is. I'm unhappy with mine but it's whatever I guess.
Would you recommend/use the z790 Hero or z790 Master for the 13900k/13900ks?
Hero!
Hero
Thanks.. I got the Hero
6.3ghz just insane
I don't understand how you identified core 1 and 7 as your weak cores.
You can use workloads like CoreCycler to check every core's stability.
Those asus AI OC features suck from my experience leave that off turn on XMP1 - Set your ram frequency then set disable all limits then boot into windows and use Intel Extreme Tuning app and use its speed optimizer to really boost your system. Just my 2 cents..
why? i got more power with asus ai not with intel extreme utilyty
Do you know if the asus strix z790e can oc nonk 12400?
Could you explain how to Check for instabilities During transient loads and single core boost?
Sit at idle for a while, then run a heavy workload. I have used 3 ways to detect these issues:
1) Sit idle, then load Prime95 Small FFTs all-core (non-avx & avx)
2) Run Cinebench R23 with 30min time (between runs the CPU goes to idle briefly before running all cores again)
3) Use AI Benchmark (very varied workload that often catches this instability)
Hello Skatterbencher. So, I used your brilliant guide and 6300 profile on my new KS. I have some issues with Chrome suddenly closing and its unable to complete CBR20. I tried changeing all p cores adaptive voltage and vrm adaptive from 1.475v to 1.48v. Still same in CBR20- crash and usual bug report error. Any suggestions? Or should I just change to 1.5v? EDIT: Changeing all adaptive voltages including VRM to 1.5v have resolved all issues! Record for me in CBR23 and CBR20 and CPU runs insane! Its a REAL treat to see the 6300 Mhz when active! ;) EDIT 2: Still not stable even at 1.52v ... What I am currently testing are 2 cores at 6300 Mhz instead of 3 .. I think that did the trick
EDIT 2: 6300 Mhz are no matter what and how many cores causing random BSOD and sudden critical crash..This was with tested voltages all up to 1.55v. just to be sure it was not a power issue. Doom Eternal - any master level will SURELY tell you if your OC are stable. That engine are amazing and dont use 1 render thread but all the workers (cores) it can get. So its the ultimate OC test. Trust me! I can pass many test and benchmark but not Doom if the OC are unstable without you know it. I modified the OC profile to have 4 x 6200 Mhz and 8 x 6000 Mhz... THAT works ! this time tested throughly... several Doom Masterlevels passed! ;) Im gonna keep that and not pursue 6300 Mhz for now. Also my CPU now max use 1.46v but are almost all the time at 1.439v. Very very good settings. I think I know what the problem was. LLC 6 are simply not strong enough and have to much vdroop. And the very peak of the limits of the CPU 6300Mhz it does not work on my CPU. The other times I used Scatterbench examples I used LLC8. And that always work but is max LLC. I wanted to get away from that to get better performance overall. LLC6 NEVER worked on my regular K on Scatterbench examples.. I always used LLC8. But this profile are now rock solid with 4 cores at 6.2Ghz! and all core in game 8 x 6 Ghz at LLC6 and its sustainable with great temps etc. Runs amazing and what a treat to see my games and Doom Eternal running all core 6Ghz. Scatterbencher has an extreme bin of an KS. SO thats why he can do the very final step where I cant at least with these power settings.
Hi friend which strategy are you using ? Are you using with TV on? And in the end what's your adaptive voltage used and what's your CPU so I would love to try your settings since my CPU isn't highly binded
@@khaos555 Hello friend! I use a different strategy now because I think this per core strat. is not good in the long run and can cause a lot of instability. It turned out that my cooling was not sufficient for those high clocks! What I did was I redid all my water blocks and LM and reseated the block on the GPU. Then I bought a Monster Radiator (yes that is its name! :) MO-RA 420 PRO.. Its massive with 2 extra pumps! This enabled me to run 6000 GHz all core with a OCTVB down clip at 70c and 80c. But the cooling is SO efficient that the CPU never gets to 70c. Only in extreme cases like CB R23. Then it clips to x58. But ALL games, even the most CPU intensive its now at 6 Ghz @ 48-58c max. Pretty incredible! (Direct Die) Its runs better than ever with not one single crash so far! (running 14 days!) Before I think what happened was occasionally it hit the ceiling and then made errors and BSOD/ Crash and or instability. So the no 1. thing is make sure the cooling can handle it. You may call it old school but all core is king imo. You don't really get anything useful out of per core imo. Also before I could not run LLC6 on 6 Ghz but its magically working now with new cooling! Please write if you want to know more.
So my final settings are LLC 6, VRM Frequency manual 800hz, Powerphase and the other one on Extreme and VRM OFFSET Vcore +0.050v .. that runs the CPU @ 1.439/1.44v 100% stable and now also cool! ;) My SP is 108 and Cooling is 198
@@lassekristensen385 Very nice cooling solution indeed congrats! my sp is pretty close to yours is your 108 the final under prediction, what is your P core SP if I may ask? your cooling solution looks and sounds amazing!, I have mines undervolted to -0.05 with max voltage around 1.39 maybe 1.4 not far from yours my cooling grade is around the 169s range I'm using Thermaltake ultra mx2 not the best but can't complain, I have my P cores limited to the 6000mhz also I can get there if CPU not under load otherwise normally I would get to 5700mhz due to core 7th maybe could try 58x I plan to play around one more time soon and retest as folr the e-cores I have the last remaining to 42-3200mhz, I could get to 4600mhz which I did previously depending if I lower p-core to 56-57g LLC is also set to level 6
@@khaos555 Yes the 108 is final. P-cores are 117 and E--cores are 92. At first I also had E cores running at 4500mhz and cache/ring at 5000 MHz. BUT that could also lead to instability and unnecessary power draw. I then figured out that there really isn't anything to gain running the L3 at 5000 so I have it static at 4500 MHz not AUTO. The E-cores was the same. I have them on auto (x43) this is because it only gives slightly higher score in lets say a benchmark. Are they are supposed to be efficient. But my main reason is that when running "auto" on the vCore or Offset or Adaptive the CPU is hard coded to request voltage with L3 at 4500 and E cores at 4300. Its much much much easier to control and adjust when these two parameters are "auto". I don't like the mobo changing L3 to 50x, 45x or sometimes 46x.... Its super stable with these parameters on static 45x and auto 43x. And believe me I have messed A LOT with this ha-ha! Also because when running lets say a very hard bench or stress for the CPU even with this cooling the watt is 340w (IF I remember correct!) So I really don't want it to further increase! Bottom line is if I have these settings like this it enables me to run the 6 GHz all core in pretty much everything flawless and if I mess with the other parameters its kind of a constant fiddling and adjustment and its better suited for manual voltage which I prefer not to because I want all the power saving features and what not. I am using also a tool called Park Control I can highly recommend. I paid for full version. It disables core parking when I use the PC or game, but when idle in 60sec it goes to POWER SAVE profile which takes all cores including e cores down to x8 or x11. and that works perfectly on the 6 GHz profile. If you then touch mouse or keyboard FULL performance not core parking again. Really a sweet utility. If you mess around with e cores or Manuel voltage some of the power saving features are not working properly. And some OC are not happy to go so low and up in ratio, but it works with this setup. What I am trying to say is P core tuning is the way to go! ha-ha! :) Also only thing that matters in gaming etc. Its way more stable this way at least in very high clocks like 6 GHz. I have had profiles with x57-x58 and ring x50 and E core x45 and that also works. But I find that x60 P and rest stock is best for me and gaming :) And thanks with the cooling it was quite expensive and I decided to go with it because it should last "a lifetime" ha-ha! It can be carried over to PC after PC. The most "cool" thing about it is that I drilled two holes in the wall for WATER IN and OUT and put the RAD in next room! so no heat or noise at all in the main gamer office room. Its pretty cool and nice. I know its a little extreme but why not! :) Before I upgraded my cooling I was running x57 on P... I could if I really stretched it run x58 but it would be almost 100c in CBR23. So I found that x57 was the sweet spot before.
@@lassekristensen385 Oh ok my P sp is 116 one number less than mines pretty sad but based on our conversation it sounds about the same my e-core sp is also 92 so it seems for sure we can both have it higher I also did get higher on benchmarks like cinebench r23 , and yeah i know what you mean when it comes to power saving, on games like elder scrolls I only wasting less than 110w honestly normal browsing less than 100w around in the 70s , while gaming I only get max out around 79 degrees c only for a few short while not under load below 40 degress if ambient temp is low since I live in a place were we get all seasons, like this morning, it's a bit cool(rare for us being (september) otherwise when its humid max I get is around 81-2 degrees but as a short spike, otherwise I expect to be around in the 40-50 degrees while browsing which i can't complain with p cores hitting 5700mhz which it does hit 6000mhz sometimes, but because I do have the cpu with some load I ran over 100 tabs on chrome if that counts as one haha, and some apps running in the background which is why I decided to keep the e-cores(main plan was to disable half 8 cores to keepo the other half) which also works but then i found overclocking it wasn't much higher
but I did find another strategy so I will see and then decide if it's worth or not, in the end I will keep this configuration or another I have about 3 profiles saved on my z790 apex motherboard on the bios, currently using my second profile just because under this profile i'm running the e-cores to 4300 on more than 90% while the remaining 4 are running at 4200mhz on my first profile if I set the e-core to auto then 90% also runs at 4300mhz but the remaining 4 are running at 3200mhz
Great video with in depth explanation as usual. I have a question though . Do you have any idea of how i will make my 4x16 G skill ddr5 6400mhz sticks to work with xmp II on z790 maximums hero with 13900k? I bought them as 2x2x16 kits separately and i run 0813 bios and also i cant see the xmp tweaked choice only xmp1 and 2. Thanks
He overclocked the cpu to 5.6 all core. This is the truth that everyone must know
Framechaser did a 6ghz p-core oc with 8200mhz ram
@@maegnificant Yeah and it's probably unstable as hell.
@@maulcs nope. 5.9ghz was also done on 1.35V
@@maegnificant "nope" doesn't say much, Framechasers is notorious for not doing hardly any stability testing besides opening a game and calling it a day.
If we're going to be precise, it's:
- 6.1 GHz all core up to 75C
- 5.9 GHz all core between 75C and 85C
- 5.8 GHz all core between 85C and 100C
- ~5538 MHz all core at 100C (P95 no AVX)
- ~5281 MHz all core at 100C (P95 AVX)
skatterbencher.com/wp-content/uploads/2023/01/Slide113.png
Looking at single core performance. I'd like to see single core clocks under load like Cine or 3DMark CPU profile. I may have missed that data in the video, can someone help? Skatter what are your impressions of single core performance and speeds? Multi cores looks good for performance improvements but inconclusive for single core? Thanks very much
Man i tried this and I was getting an overvoltage warning error with your advanced methord for my 14900k on my Z690 asus TUF. Was a bit scary. Any idea what i could possibly have done wrong?
For some reason my Z790 Hero when I have everything at default with MultiCore Enhancement disabled my 13900KS is not sticking to it's default PL1 AND PL2 Limit as it is setting them both of the limits to 320.0W for both of them. So I can not get the chip to run at stock could this be a fault with the bios ? I am using Bios ver. 0813
After I finished my video, Intel released a new Extreme Power Delivery specification for the 13900KS processor which has PL1=PL2=320W, not 253W. So your default specifications are correct
www.coolaler.com.tw/coolalercbb//INTEL13TH/13900K/ClipKS_1.jpg
Do you think leaving it on Strategy 2 (AI Overclock Method) for longer periods of time will return higher and higher performance (benchmarks) over time as opposed to setting it and testing it right away, as it learns over periods of time and periods of use and adjusts/tweaks the overclocking settings on it's own w/ learning?
I dont think so, it will only make small voltage adjustments based on cooler score. I think its based AI OC is on fused VID, SP quality along with cooler score of corse.
How does your BIOS version say 0801? I didn't find that version anywhere on Asus's website. What do you think of BIOS 0813 for this board ?
No idea what happened to 0801, will ask around.
If you want to try 0801, I've uploaded the file to my blog skatterbencher.com/wp-content/uploads/2023/01/ROG-MAXIMUS-Z790-HERO-ASUS-0801.zip
@@SkatterBencher Interesting, thank you sir! If I use Corsair iCue h170i elite AIO, what's the max overclock should I aim for in your opinion? I got this very same MB, ROG Maximus Hero. Or should I not think of overclocking at with an AIO? You should do a tutorial on over clock the 13900KS with an AIO plz!
this also works with a K (none ks) with ~110sp ? (sp115 is not necessarily high)
The process will be nearly identical, but of course the exact settings will differ.
@@SkatterBencher what is the maximum adaptive voltage you should use with a raptor ?
I made the 6300 Mhz profile successfull on my KS.. it did however require I upped all voiltages to 1.5v but runs amazing! My previous 13900K could not reach 6200-6300 successfull .. I used Skatterbenchers 6000 and 6100 profiles with success on that CPU.
@@newdawn005 I am using 1.5v on all the adaptive voltages .. runs amazing! This was required for my KS to be stable with this 6300 Mhz profle.. 1.475v was too low unfortunayly for my KS. But 1.5v is fine.. personally I would not go higher.
I stuck with this board for my new build, waiting on PSU but can't work out how to fit my M. 2 5.0 ssd at 5.0 speed and not sharing pcie with GPU. Any one know Im really stuck..
Wish my KS was a 115, got a 107 :(
I'm still unstable, I had to put all p-cores at 6.1ghz and AC/DC line 0.20 on both to be stable. Do you have any further advice? I can't go above 0.20 AC/DC because then the temperatures rise too much
mine doesn't have the xmp teak settings only xmp 1 and 2 and with the auto ai my pc keeps restarting??
I got asus maximus hero z790 13900k 4090 I did exactly what you said and when I go run cinebench again after the ai auto tune pc just reboots ??? am on the updated bios and win 11
Sounds like the overclock is not quite stable. You can try either of the following options:
1) Reset to defaults, go in the OS and run Cinebench R23 for 30min to "train" the motherboard. Then try AI OC again.
2) If too low voltage is the root cause of the instability, you can also try adjusting to a less steep VRM LLC
@@SkatterBencher okay tbh idk how to done none of this am a noob idk how to oc my gpu either but ill try that again and see what happens u know why I dont have xmp tweaked? I only have xmp 1 and 2
I try it but not good result only cinebench record is 42100. . . All sinc 61 ecore 48 ucore 53 is more performance 45200pts
45200? Ive never seen such a high score.. I just beat my own personal with this OC profile in the video. 43110. I dont quite understand how you can make 2000 additional... there is only a "budget" of heat and power draw available so unless you are using xtreme cooling of some sort it seems incredible!
@Lasse Kristensen th-cam.com/video/CoIjXSrReAc/w-d-xo.html
Here is my 10 min record with 8533mhz ddr5
@@Hwbot-Philip-Park Awesome! Insane RAM OC! :) I am running a more conservative profile haha! Its a 6000 Kit, I OC to 6600 Mhz :)
@Lasse Kristensen do you have Ln2?
Now I am using 13900ks 8800CL36 stable Hci 500% and 9000mhz hci 60% unstable it but all bench Geek3 to 6 is passed not good performance bench than 8800Cl36
@@Hwbot-Philip-Park No no no, only normal water cool. I have G Skill 6000 and I overclocked them to 6600 Mhz with 1.46v @ the same timings! 36-36-36-76. I know its Samsung B Die.. But I dont seem to be able to get them higher? Any tricks for B Die? OR is this my limits?
im pretty new to this but for some reason when i remove all limits on this board it gives me around 340 watts in HW monitor and its unstable i believe partly because of the multi core enhancement even on optimized defaults. anyone have any ideas?
This just makes my cpu run at max frequency in idle. Anyone know a way to fix that?
Enable C states in UEFI and set Window power plan to balanced