tldr for normies, but please watch the video!!! -Check *CEP & TVB* is enable -Set *Loadline cabration* to extreme & IA AC & DC Loadline to 12 -Put *ICCMax* at 400 if you got a good enough cooling to handle it. 307 or 320 as the sweet spot, below 247 is possible for smaller cooler or for longevity however performance will drop -drop the *E-cores* clock ratios to 42x for an i9 -(Optionally)if you're paranoid, set the temperature limit to 90 Celsius this setting still needs ICCMax as well Please send feedback as a comment on this video not my comment, since buildzoid's work around is experimental, but better than Intel baseline settings.
those settings he entered was not him clocking half the e cores down - they were all limited to 4.2, which is what he said to do because it makes things better for the p core while not negatively affecting performance (it tends to help).
@@xBINARYGODx I know that dropping the e-cores helps the p-cores not lose performance. I written it down in half ratio like "performance in half" like that so normies could understand this work around.
The problem with using the 6Ghz 2 core boost is it was designed to be used with TVB frequency clipping and TVB voltage optimisation enabled. Most bioses have the frequence clipping disabled which is the important one, this prevents the cpu from trying to boost to 6ghz when temp is higher than 70c. Boosting to 6ghz at high temps requires 1.5v+, high voltage combined with the high temps, this is what's probably degrading the cpus. Also as you've already mentioned, 400a ICCMAX is the other very important setting for stability and long life, the P cores have some insane current spikes and vdroop at those currents requires a much more aggressive AC loadline.
G'day Mr Zoid, I watched a lot of videos on this untill i came across your last one, I was unable to run cinebench R23 at all and could not run XMP either untill you fixed it and it fixed the crashing and the temps too. Thankyou very much, all the rest i had seen just said to lower package power over and over again. I have a gigabyte z690 master DDR5 set at a core current limit of 380Amps and max out the package power. With 360mm AIO my 13900k is now stable and max temp of 84c under full load for 1 hour, It is good for 38,000+ cinebench R23 score can go higher but thats good for me. Thankyou sir and have a great day. :)
The crashing appears to be caused by a combination of high core clock speed, hyperthreading, and certain CPU instructions (integer address arithmetic). It is actually less likely to crash under all-core loads, because thermal throttling reduces the max clock speed. I found the most likely condition to cause a crash is when both hyper-threads in the same p-core are loaded, and all the other cores are idle. You can test this using 'set affinity'. The CPU can appear stable under all-core loads and still crash when tested under these conditions. The power and current limits dont really do anything to stop crashing when there are a few p-cores running with both hyper-threads loaded. Thread Director tries to only load one hyperthread at a time, so makes this problem hard to detect under normal operation, hence the need to use Set Affinity to test for it. Thread director is not 100% perfect at avoiding these scenarios, hence you get infrequent random crashes under certain types of load.
Thanks Buildzoid for this video! It has worked pretty good on my system: 14900KS, Aorus Master z790 (non X) and 96 GB of ram at 6800 MT/s. Having similar BIOS & motherboard to yours has simplified a lot of things.
Great observation there regarding running high E-Core frequencies and the impact on voltage on the P-Cores. Helped me squeeze a bit more out of my P-Cores as suggested. Thanks :)
Thanks! I had previously set PL limits and wasn't really thrilled with performance. The first config here got my 14900k to 39.5k in cinebench r23 - which I'll gladly take, and it's much more stable now
Intel in Korea has actually added both Hogwarts Legacy "first launch shader compiling" and specifically Tekken 8 for RMA/stability testing criteria. I don't have HL myself, but I can confirm Tekken 8 is not only an extremely hard shader compile stress test, but it only takes 1-7 seconds, and you'll BSOD, Out of Memory Error, WHEA Correction, or pass. My "it's stable in 6~ back to back TLOU shader compiling + infinite GB6 runs for transient" OCTVB OC had one bsod throughout a weeks usage, and one whea correction during first launch of another dx12 game. Sure, added a smidge voltage, called it good. THEN I started playing Tekken. Here I am 2 days later, and a mix of downclocking 1 bin each @ 90/95c from 5600, and a 290W PL required due to a +25mV(!!!!!!!) Load Voltage required to successfully open Tekken 8 5x in a row after clearing shaders. Those limits are made to prevent degradation. It would easily go above 300W and 95-100c without them. This basically makes running R23 or similar at 56/44/50 completely impossible now on an AIO without it immediately downclocking 2 bins across the board. But hey. It's most definitely fixed the edge case BSODs from earlier, and at least it can fuckin launch Tekken reliably now.
Cheers for posting that. Had no idea about those games and thinking I need tekken now for stability testing and hey comes with a game to play (if my KS can make it last shaders comp)
@@sibzilla They've made it easier to launch in one of recent patches. I noticed I could pass 5 runs by after a reduction of 10mV over the increase (so like 12~15mV net increase versus 20-25~) and an adjustment to Input Voltage/CPU AUX. Yakuza Like A Dragon wasn't happy, and still needed more voltage, but everything besides that game was fine for like a flat month with no issues, idling, etc. There's really no point using stress tests like R23 and similar nowadays - all these modern games hit the cpu's in such drastically different ways even if they have light load.. Better to just find something stable in windows (geekbench 6 stable) then game it out.
well i have a 14700K and I'm using these exact settings but with 253 pl values and it works great. wish intel had a chart for the i7's also. thx BZ for all the hard work. and hopefully we get word on proper IA AC_DC values.
Thanks! Same cpu and mobo here. Copied your settings to the letter - result 39964 in cinebench 23, I did the 10 min throttling test btw. 1 core hit 100 (core 7) lowest was 89 (core 0). 1 core hit 6.0 all the other P cores hit 5.7, none dropped below 5.4 during the test. 398.9W max power draw
How do I get my new Intel CPU working as intended? Firstly, don’t set defaults. Secondly let’s go deep into the BIOS and start changing many settings. It’s how Intel designed it. #plugandpray
Thanks for the content and the video! I'm currently trying to use the same settings, but for an ASUS board. You noted that turning on the CEP would require us to remove the undervolt that comes default with the motherboard. As you noted enabling CEP in BIOS absolutely destroys my CPU performance. Where would you find this undervolt in the BIOS (I am on an ASUS)?
He has it off for testing (I assume) Hard to tell if ram is stable if the cpu is crapping its pants. Super fast ddr 5 on motherboard manufacturers list will also crap its pants. Memtest and gaming might work for days but try running karhu memtest and it will tell you stuff isn't right show an error in 30min or less. I am talking about speeds over 7000mhz. This generation they really lied and hyped way too much.
Hi Buildzoid. I’m basically wondering how you go about correctly adjusting the AC Loadline and DC Loadline according to the Vcore Loadline Calibration (LLC). I’ve watched/listened to your videos a couple of times and tried reading about the subject. But I’m not sure I completely understand it. I have a gigabyte z790 aorus elite x wifi7 with a 13700k. I would like to use the medium LLC but don’t exactly know how to go about calculating the AC/DC values.
I just saw a report from hardware unboxed, that the gigabyte Intel baseline bios was the actual version Intel said that is what it should be out of the box, gigabyte aparrantly had it correct.
I can't believe that 1.7 mOhm of AC_LL ist correct. When I look into the Intel documentation, it says that it should (max) be the same as DC_LL which is max at 1.1 mOhm for 125W and up. And even that gives you some really high idle Voltages. But going even higher can't be right. That whole story is completely insane.
Hey, dunno if you've seen my new comment now but looks like Gigabyte has removed at least my motherboard's latest bios version. Let's see if they're working on something new. This is very odd. I'm now officially on a ghost bios version. lmao (Z790 Aorus Elite DDR4 1.0, non-AX, bios v. F11f)
intel baseline has been option on my asrock board for 5 months atleast, the 125/188w. Never had any stability issues but I never did 6 ghz boost, 5.7 all core with voltage offset undervolt.
Is this worth trying on my asrock z790 on my 13600k? Have a nice undervolt going right now but I believe I don’t have TVB enabled, just 5.1/3.9 standard all core, I know where the AC DC load line values are, im at the medium droop (3) currently
Fix stability issue I did have with older software (and gave up a while ago) on 12900KS MSI Z690 ... assume they were not compatible with big small cores. Clearly the only reliable source on youtube ...
You're an invaluable resource on this subject. I've now held off on upgrading to a 14900K until we have a better resolution on the situation. Just out of curiosity, does this setting help 12600Ks and 13700Ks in any way? I've noticed my 150w limited ixt board does not enable any CEP and sets AC/DC Load line to 0.70 for my 12600K.
Videos like this spooked me, so recently when I upgraded I bought a 13700k with a Gigabyte Aorus Elite X. I run a BIOS version which is 1 version older than the current one and pre-dates all these new "Intel Defaults" settings etc and so far I've just let it run as fast as it wants. Under cinebench it runs at 210 watts, 1.35 volts and reaches around 75c degrees max under a 240aio. Haven't had any stability issues at all.
This is roughly equivalent to level 7 on ASUS. For AC_LL you should try values between 0.10 and 0.20 until the scores look good. You can leave DC_LL on auto since it's automatically matched to the VRM LLC level.
@@ActuallyHardcoreOverclocking Isn't the Intel Baseline Profile on Asus good enough? Or is it still bad? I have mine on Intel Baseline Profile on a z790-F gaming WIFI II and it seems stable so far.
@@ActuallyHardcoreOverclockingis it wrong that i had to adjust DC_LL because vcore was not matching VID? i bumped DC_LL to 1.05 on LLC4 first and VID = Vcore now. then i lowered AC_LL till it was unstable and went up
@@ActuallyHardcoreOverclocking Thanks. I tried these settings on my ASUS ROG Strix Z690-A board and the latest BIOS, but it gave me worrysome results without power limit. Several of the Cores shot up to 100°C during Cinebench R23 and throttled hard (normally it will reach around ~76°C with my custom loop with three 420mm Radiators and D5 pump on full blast which keeps the liquid temperature delta around 3-4°C in a cool room with Auto SVID behavior and CEP disabled, and around ~85°C on Intel Failsafe SVID behavior with CEP enabled), even though I had limited it to 90°C in the BIOS, and the score was way lower than with the Intel Fail Safe settings due to throttling. I was able to somewhat fix it by setting a 125W/253W Power limit with 307A CPU Core/Cache Current limit, but the temps are still higher and scores lower than the Intel Default option in the BIOS this way. Not sure what's going on here, perhaps it's because I'm on a Z690 board and not a Z790, but I think one still has to be careful with these settings on certain boards and should probably set a lower power limit just to be safe.
Is this only an issue for Z790 boards? I recently upgraded an AERO D Z690 board to latest BIOS and put in a 14900KF and its very solid with 2x48GB sticks running at 6200 MT/s. Runs all these benchmarks continuously without lockup/crash. That said, Cinebench R23 still throttles towards the end of a single test. Avg score around 40,500 - 41,000.
I tried it with my z790 gigabyte and 13900k but with these exact settings i have a cinebench23 multicore score of 21000, even lower then with the intel baseline config around 28000. With the unstable default config i get around 33.000. Ive got no more ideas left. And Can i run xmp 1 with this setup? :(
I GOT NEWS LMAO Hello! First, thanks for your videos. I recently got a Gigabyte Aorus board (a Z790 Aorus Elite DDR4 Rev 1.0 non AX) and I've been checking Gigabytes page daily now for updates. Looks like they've removed update F11f (the one I'm currently one). This is the version with all the "improvements" on it. Looks like they're probably working on a new one, hopefully. This is weird. I haven't seen any sort of announcement but yeah. Maybe it's something to keep in mind and something to look forward to. Edit: For those wondering, I recently got a 13600K (upgraded from my 8600K, RIP buddy) and temps are still kind of high despite undervolting and adjusting power settings and doing all I can in a custom watercooled loop. And yes, I'm using an ILM plate. lol But this isn't so you help me troubleshoot or anything. Just letting you know how it's going on my end. Let's see what they've got in store for us next.
BZ, I think you may want to try these settings out. Granted, I have a 13900k, you might see similar results. 13900k; MSI MEG Z790 ACE; LT720 AIO. 55/43/45; HT off; c-states off; TVB all off; PL 1/2 253; Current auto/max; AC LL 10; DC LL 50; Adaptive + Offset auto: -100 mv; LLC Mode 6. HT on, only difference is AC LL 15. I’m completely stable in games and stress tests. OCCT large/extreme/variable, I may hit 80C on a 30 min loop with HT off and 87C with it on. I did a 30 min loop with R23 and hit 40,397 and it never down clocked. Unless I’m doing prime 95 or y cruncher, it won’t down clock. Oh yea, max voltage I hit is 1.259 and drops as low as 1.169 in OCCT. Maybe you can see if that works for you. Obviously some settings will differ depending on the board and CPU quality
I've been running this exact way since Nov 2022. I play Tekken 8 everyday, have compiled shaders in TLOU, No Rest for the Wicked, CoD, Hogwarts, etc without issue. Manual tuning is the entire point of an Intel K sku. I don't know why anyone would buy an i9 K sku cpu to run "out of the box" settings. You've done well.
I am not buying Intel until these sorts of issues are not even possible outside of what's clearly labeled as overclocking/out of recommended spec. I've been an Intel user from the Q6600 to the 8700K and did (and still do) enjoy tinkering, but after a few years of running a rock-stable undervolted 5800X where most of this stuff is handled automatically, I have exactly zero appetite for dealing with power-hungry, hot, unstable, possibly degrading platforms that even BZ has trouble wrangling into safe territory w/o losing massive portion of the performance.
using extreme LLC, wouldnt that be a concern for degradation too? i personally think its just more straightforward and better to apply a static clockspeed (yes its downclock but oh well you gonna lose performance with intel baseline anyway) and static voltage with low LLC.
Can this be applied to an I7-14700K? I had to RMA mine because of the unestability issues, and I want that the doesnt break too. I only see people talking about the I9s, and none about the I7s :(
i put acdc loadline to 12 but my motherboard has the mohm measure so i putted at 12mohm for a sec, did i damage something or am i fine? it went there for small times
Looking good, I also tuning my PC using DC LL / AC LL / TVB. But BZ cpu temp is a bit too high ? what cooler he is using ? (Due to high CPU LL level ?) I got the same cinebench score, 40k-ish from my 14900KS + 240 AIO (30c room temp). CPU P core hovering around just 5.5-5.6 when benching (VOUT 1.18v~1.2v) CPU package ~ 300w peak. (my KS V/F table at 6.2 is a bit trash compare to another KS) 800 0.774 1400 0.774 2400 0.779 3400 0.869 4300 0.989 5100 1.179 5900 1.413 6200 1.513 Now considering should I delid it for fun or not, I like low temp CPU, for lower fan noise in my gaming ITX build.
As he told in vid AIO 360, and AIO is NOT good enough imop. way to many think they can use AIO on these 13-14xxx. Aio is only good enough for gaming with undervolt.
Yes generally people like to have their work deemed successful (validation) hopefully my Asus 13th gen work station/renderfarm will work within this framework
Seems like Intel keeps a very loose spec, and it's been to their benefit (give board makers a lot of leeway and you can target high performance and low price all under one moniker). Historically their chips have had enough headroom that boardmakers, who are always looking to tweak things to get the most out of the chips, could succesfully eek out some extra by playing to the extreme ends of the spec. But these latest chips are pushed to their limits already (especially 14th gen) and so the board makers usual tactics are running into problems this time. TLDR - intel chips usually have enough headroom/leeway that they could afford a very loose spec for board makers and things would mostly be fine. Their latest high end chips have used up that headroom in the name of more performance, which means they can no longer tolerate a loose spec. But Intel being Intel, they don't want to change the spec or admit this. So we get band-aid solutions.
So in your Rambling video you said the only Reason for crashing you can think of is to low vcore. Can you get BIOS Stock settings to run stable in CBR15 with increasing Voltage ? . Just for testing ?
if I set AC_LL to 70 that fixes the crashing (though it doesn't change the all core load voltage it just makes the CPU thermal throttle harder so it runs the same voltage as before but at lower clocks). It also makes the low load voltages even higher.
Hey man i really understand what you are trying to do but the true challenge is to test a possible fix is through compiling Unreal engine shaders. My previous 14900k before the RMA would pass everything and i mean everything,prime 95 small ffts it would run any Cinebench version for hours before i was stopping it and it would fail in a matter of seconds to compile shaders. I went as high as 1.47v under load even on lower frequencies than what the chip was supposed to be and it turned out that in my case anything above 5.2ghz in the p.cores was just throwing errors in compiling even at 1.47v . The whole issue with Intel chips and shader compiling is extremely weird.
@@ActuallyHardcoreOverclocking not a simple way that I know of. I was just using Taken 8, Remnant 2 and The Last of Us shader compiling while I was switching between the newer and previous Nvidia drivers on my 4090 after a DDU driver uninstall to trigger again the shader compiling on these games. The whole process was time consuming and in my experience Taken 8 was the most reliable one to saw shader compiling instability with a close second The Last Of Us which isn't Unreal Engine but it is quite taxing during the compiling and surely takes a lot longer to compile its shaders compared to the other titles.
I had a professional halon fire protection system installed when I upgraded to a 4090 so its good to know that this should also cover my Intel cpu. Imo you're doing it wrong when your not putting your life and house at risk of fire, just say'n...
I was set iccmax on my apex Encore 400A, it still got performance like 511A and about only 2% perf loss, but got better 3 degree temp. I think if cpu dont get crash by default, we can set iccmax to 400A with proper cooling like the galahad ii perf, the perf and temp is better than intel basline.
Tried your ideas on my 14600k best option I found after getting a brief glimpse of the recommended core current, was to remove my turbo power limits and just set max core at 200 then leave everything you recommended at auto. just moved my power to 150w still below 90deg running r23. useful video.
Hmmmm havent tested my system for stability. I have had crashes/ or problems launching only Unreal Games, seems just choosing a different NVIDIA driver fixes the problem. The finals, Darktide, etc. Otherwise no problems but big problems if wrong driver. Hopefully these problems dont continue to get worse. Dont see taking a performance hit is worth not juggling nvidia drivers around. Of course I havent loaded up Grayzone. 14900k Auros Master X w/ a Gigabyte OC 4090 ...hope this gets worked out. Currently hating Gigabyte and Intel for all the time wasted trying to get games to run. I only started really noticing the issues since release of The Finals. I find it strange it is Intels fault but juggling Nvidia drivers seems to make games run fine. Having to keep driver releases named by game on the desktop is pretty crap.
I literally just returned my New Asus mobo and 14700k because the system stopped booting into Windows and if it did load the os was buggy, laggy and nothing would open. I'll try this on the new parts when they get here.
Would appreciate a helping hand 13900kf/deepcool ls720 360mm/ gigabyte z790 gaming x ax F9 bios whatever default settings in there except for XMP 2nd profile (5600) After watching initial video regarding instability and having experienced really infuriating crashes during shader compilation in some games with errors saying I lack video memory (which is not the case) or just app crashes like trying to teleport in D4 (it seems there is a spike in CPU activity, and it crashes the game) just to name a few decided to try checking if I get the same behavior in cinebench and it's the same - R15 crashes, R23 crashed on first run, second run was successful, max vcore around 1.5V, idle 1.384V Score was 39492 points with temp limits hitting 2 P-cores at 99 others were around 91-92 and something called "Package/Ring Thermal Throttling" was also switching to Yes during the run CPU package power was hitting around 350W (i think, don't remember exact number but over 300W) Applied first fix (TVBs on, IA GT CEP on, current limit 400 max 4095 power limits, CPU Vcore loadline calibration set to Extreme, AC DC LLs to 12) Instability in R15 went away, but now score in R23 hits only 29400 tops. Package power draw is down to 287 watts, Vcore max was 1.392 idle 1.270-1370 (not really lower that on default settings) In hwinfo's "performance limit reasons" section I see two things kicking in during the run IA: Max Turbo Limit RING: Max VR Voltage, ICCmax, PL4 but at the same time package/ring thermal throttling doesn't get triggered. Why there is such a drop in performance? Core clocks are the same as previously (5.5 Ghz for p-cores, 4.3 Ghz for e-cores) Might it be that z790 gaming X need different specific AC DC LLs for that board instead of 12/12? (default settings were 40/90) But I am having a hard time figuring out how to even properly understand what these AC DC LL parameters do, so not to kill the cpu with further experimentation with LLs EDIT: went back to default settings by manually switching back everything, R23 was crashing even before showing the first block, on 4th try system locked, went back to bios, rechecked everything, applied "optimized default settings" just for shit and giggles, booted back, R23 worked, score was 38000, max temp 95, "Package/Ring Critical Temperature" did not trip Package power was 305 watts (probably it was this initially i just remembered incorrectly that it was 350) VR VCC Current was 232A Max Vcore 1.428V, idle 1.284-1.370V IA: Max Turbo Limit and RING: Max VR Voltage, ICCmax, PL4 kicked in during the run oh, and i was checking wrong core clock speeds metrics it seems, there are effective clocks, and on "12/12 rails fix" it was 5.3Ghz and 3.8 Ghz p-cores, e-cores
After experimenting and watching last video with intel providing baseline defaults, I switched loadline calibration profile to extraExtreme (the flat line), still was getting ~30k points in R23 with 12 AC/DC LLs, and 1.35v vcore with 1.42v spikes, switched to 13 and it started looking much better, ~38k in R23, and stable 1.38v vcore, but hitting 1.404v in idle lol, decided to see if there will be any difference with setting it to 40/40 and scores even lower 35-36k, but it also uncovered a different issue - there is almost 10 C degrees delta between p-cores 0,1 and the rest of p-cores constantly hitting 99-100 degrees, also it seems i can only run 5.5 Ghz (but it's not stable and crashes on 2 or 3 run in 10 minute test) with default gigabyte settings, with settings above it maxes out at 5.3 ghz. Probably due to thermal throttling, as there is a constant delta of ~10 degrees with 6 cores hitting thermal limits constantly, probably due to ilm bend, I was thinking of installing a contact frame just in hopes to drop temperatures during more intense gaming sessions (like 80-85 degrees), or continue with this settings as probably will not even notice but again, i want mine 5.8 GHz turbo clock ))
It is called experimenting. He was brainstorming what he thought at first would help with stability. Usually bumping up voltages will fix unstability. However he found that it was making the cpu's degrade faster by having higher Temps than previous.
How much do you trust intel to replace the CPU if it breaks? Because IDK if my settings meaningfully improve longevity/reliability but I did try to get the voltages as low as possible while maintaining stability and not compeletely destroying performance.
@@JordanJ01 lol u cant just make point and move on... we are talking about processor that cost a lot and all this explaining is in place... but can he make it better with timekeys? ... YES!!!YES!!!YES!!!
25:29 >> Laptop CPUs run at way lower voltage (and current), and as such.. lower frequency at full load. Smaller chips(dies) can also cope better with very high and fast thermal-cycles. Same reason you generally see more longevity with GTX 970 than 980 Ti ...unless you had a big cooler on the 980 Ti, such as the Gigabyte Xtreme model (keeping temp cycles slower and less dynamic).
There is a new BIOS for my z790 Strix-H Asus board Can you test it and tell your thoughts about it, please? :) Thank you "1. Introduce the ""Performance Preferences"" with options for Intel Default Settings (Performance/Extreme) and ASUS Advanced OC Profile. 2. Redefine the factory defaults based on Intel’s new ""Intel Default Settings"" for various CPU SKUs. 3. Change F5 from ""Load Optimized Defaults"" to ""Reset to Defaults"". 4. Add warnings when users switch from the defaults to other settings.
intel probably got rid of their quality control, they just put the cpu on the market and wait for buildzoid to put out the video, then they make their corrections.
Funnily enough the high frequency high voltage idles doesn't... quite degrade those said frequencies... but it most definitely makes the cpu more sensitive to 92c-100c(+) scenarios. Looking at your cpu, and remembering to my freshly bought one, naturally it redlined 100c at "stock". But it managed at those situations and could sit there. But I've noticed whea and watchdogs induced by temperature. For example R23 would corrected whea often at X config, but after repasting and temps dropping a few degrees, those corrected errors disappeared at the same settings. It's really a struggle to tame these CPUs because certain workloads can just fuck it, now that devs are making games on DX12 with obscene shader compiling stress, and it's different per game. Tekken 8 is a game that makes the news for this, and it literally only takes a few seconds onto launch to bsod or error. I could pass INFINITE GB6 and did like 6 runs of TLOU shader compiling, but it was not, and is still not, having it on Tekken 8 reliably. Interesting phenomenon, because heat induced errors are at play too due to the die setup of 3/5/7 being super hot cores, specifically 7. These video settings are still fucked though. Your LLC pairing with DCLL may be wrong because of such insane wattage and amp reporting, but when you're running r15 and 2b, I'm seeing a max of 295 AMPS over 300W at 100c.. thats still hard degrading
I know people who degraded 13th and 14th gen CPUs by running single threaded memory benchmarks at 6GHz+. Also isn't it kinda strange that heatsink reviewers aren't reporting their 13th and 14th gen test bench CPUs dying but random gamers are complaining about CPUs getting progressively less and less stable? Maybe shoving 100-150W through just a couple cores isn't a good idea.
@@ActuallyHardcoreOverclocking I'm always shocked when seeing review or tech channels, big ones too, just carelessly benching their CPUs at "default" and I see 350W+ 14900K/KS at above 1.25 load volts and sitting 100-105c for the 10-30min loops that they say. That leaves terrible habits for the viewers who later get their chips and do the same thing, after which they've just degraded by a few mV already on all core. I personally haven't experienced high frequency/light load degradation on mine (just observations of much weaker temperature tolerance under load) like an inability to hit clocks. IE; the same settings that could infinite pass GB6 transients still pass it now at the same scores, but that's at a range of 1.46-1.5v ie. R23 SC (proper one, not launching with specific threads set, that is dangerous), SuperPi, Idling/Browsing with a medium strength LLC. A high strength LLC with a 3-5 core active high frequency boost is dangerous because many games will do it, and the voltage won't droop to typical 1.3-1.4 gaming levels and instead that's where you might see 1.45+ at 80+ Amps/100+ Watts because it doesn't droop.
I don't know why Gigabyte and Ausus aren't getting the blame here. This was all they had to do? Have someone like Buildzoid who knows something spend a few days working on their BIOS before they publish and ship it? Why isn't this their responsibility? It seems like they were just completely lazy. And how does he come out with a better BIOS config to fix this than them before they do? I really think people are letting the board manufacturers off the hook. Every big name youtuber blames Intel for this. Even Buildzoid. Is it just more friendly to the algo? I don't get it.
Intel makes the design guidelines. For over a decade now the intel design guide lines have basically been: "do whatever you like just don't touch the core clock ratios" and for years it was fine because intel wasn't shipping CPUs with boost clocks higher than 5.3GHz.
@@ActuallyHardcoreOverclocking So what? Not every board manufacturer did this. Not every model of Gigabyte board even did this, right? It's their job to make sure things work together. And some did. They didn't on this board. But that's literally their function. That's what the motherboard actually does. They write the BIOS for godsake. Don;t you think they should check to see if it works lol? And if it doesn't then update it? Which I guess they still haven't bothered to do or I'm not sure why you're doing this.
@@posmoo9790 board makers won't invest more effort than they need to if they are not forced by intel to make these CPU's long term stable which is why intel should be blamed for this and not board makers look at AMD and how they further tightened their guidelines after 7800X3D's burning up situation while intel ignores issues for 10 years and still refuses to fix those issues
@@posmoo9790 what is really stupid is if you download XTU , that is made by intel, and use the AI optimizer to actually overclock the cpu, XTU sets the icc max to 350 to 400 depending on the the cpu model.
@@xthelord1668 Listen, I buy AMD because I care about not turning my room into a hotbox and I don't have a desire to mess around under the hood and get the most out of what I bought. It's pretty ironic that people whose hobby is tinkering with this stuff are just begging Intel to lock everything down because board makers can't be arsed to make sure their settings are stable - people who have bought brand new Intel CPUs and boards are finding these problems out of the box. That's on the board makers. And careful about getting what you are begging for Buildzoid. I'm gonna laugh my butt off when they lock everything down next gen and you all cry about that because your hobby is gone. And why shouldn't they lock it down? The people who should care most about keeping these specs open are thrashing them when a little due diligence by the board makers would have kept this under control.
I appreciate your attempt at brevity in these trying times.
tldr for normies, but please watch the video!!!
-Check *CEP & TVB* is enable
-Set *Loadline cabration* to extreme & IA AC & DC Loadline to 12
-Put *ICCMax* at 400 if you got a good enough cooling to handle it. 307 or 320 as the sweet spot, below 247 is possible for smaller cooler or for longevity however performance will drop
-drop the *E-cores* clock ratios to 42x for an i9
-(Optionally)if you're paranoid, set the temperature limit to 90 Celsius this setting still needs ICCMax as well
Please send feedback as a comment on this video not my comment, since buildzoid's work around is experimental, but better than Intel baseline settings.
Lets Say i want to try to oc and adjust voltages in future. Can i keep anyway these limits or i Will down clock? Targeting Just a +200 MHz to 5.7
🙂 okedoke, thanks for the tips, i'll copy and paste it
those settings he entered was not him clocking half the e cores down - they were all limited to 4.2, which is what he said to do because it makes things better for the p core while not negatively affecting performance (it tends to help).
@@xBINARYGODx I know that dropping the e-cores helps the p-cores not lose performance. I written it down in half ratio like "performance in half" like that so normies could understand this work around.
Thanks for this - he goes at such a pace that it's hard for us mere mortals to keep up at times!
a shorter version? Who are you and what have you done with the real bz?
Short version is still 30 min. He's alive and well.
LOL
RMA it.
it's not your job to have to rely on someone that experienced like buildzoid just to have a semi functional CPU.
Hey, don't worry, it's him, it's 31 minutes as usual.
The problem with using the 6Ghz 2 core boost is it was designed to be used with TVB frequency clipping and TVB voltage optimisation enabled. Most bioses have the frequence clipping disabled which is the important one, this prevents the cpu from trying to boost to 6ghz when temp is higher than 70c. Boosting to 6ghz at high temps requires 1.5v+, high voltage combined with the high temps, this is what's probably degrading the cpus. Also as you've already mentioned, 400a ICCMAX is the other very important setting for stability and long life, the P cores have some insane current spikes and vdroop at those currents requires a much more aggressive AC loadline.
"Obviously I've not tested it exhaustively... arguably I've tested it more than Gigabyte did" lmfao gottem
G'day Mr Zoid,
I watched a lot of videos on this untill i came across your last one, I was unable to run cinebench R23 at all and could not run XMP either untill you fixed it and it fixed the crashing and the temps too. Thankyou very much, all the rest i had seen just said to lower package power over and over again. I have a gigabyte z690 master DDR5 set at a core current limit of 380Amps and max out the package power. With 360mm AIO my 13900k is now stable and max temp of 84c under full load for 1 hour, It is good for 38,000+ cinebench R23 score can go higher but thats good for me. Thankyou sir and have a great day. :)
Welcome to actually hardcore underclocking
AHOC / AHUC
The crashing appears to be caused by a combination of high core clock speed, hyperthreading, and certain CPU instructions (integer address arithmetic).
It is actually less likely to crash under all-core loads, because thermal throttling reduces the max clock speed.
I found the most likely condition to cause a crash is when both hyper-threads in the same p-core are loaded, and all the other cores are idle.
You can test this using 'set affinity'.
The CPU can appear stable under all-core loads and still crash when tested under these conditions.
The power and current limits dont really do anything to stop crashing when there are a few p-cores running with both hyper-threads loaded.
Thread Director tries to only load one hyperthread at a time, so makes this problem hard to detect under normal operation, hence the need to use Set Affinity to test for it. Thread director is not 100% perfect at avoiding these scenarios, hence you get infrequent random crashes under certain types of load.
Thanks Buildzoid for this video!
It has worked pretty good on my system: 14900KS, Aorus Master z790 (non X) and 96 GB of ram at 6800 MT/s. Having similar BIOS & motherboard to yours has simplified a lot of things.
Great observation there regarding running high E-Core frequencies and the impact on voltage on the P-Cores. Helped me squeeze a bit more out of my P-Cores as suggested. Thanks :)
bz tunes Intel CPUs for stable performance better than Intel and mobo vendors. I appreciate the shorter format.
Thanks! I had previously set PL limits and wasn't really thrilled with performance. The first config here got my 14900k to 39.5k in cinebench r23 - which I'll gladly take, and it's much more stable now
Intel in Korea has actually added both Hogwarts Legacy "first launch shader compiling" and specifically Tekken 8 for RMA/stability testing criteria.
I don't have HL myself, but I can confirm Tekken 8 is not only an extremely hard shader compile stress test, but it only takes 1-7 seconds, and you'll BSOD, Out of Memory Error, WHEA Correction, or pass.
My "it's stable in 6~ back to back TLOU shader compiling + infinite GB6 runs for transient" OCTVB OC had one bsod throughout a weeks usage, and one whea correction during first launch of another dx12 game.
Sure, added a smidge voltage, called it good.
THEN I started playing Tekken.
Here I am 2 days later, and a mix of downclocking 1 bin each @ 90/95c from 5600, and a 290W PL required due to a +25mV(!!!!!!!) Load Voltage required to successfully open Tekken 8 5x in a row after clearing shaders. Those limits are made to prevent degradation. It would easily go above 300W and 95-100c without them.
This basically makes running R23 or similar at 56/44/50 completely impossible now on an AIO without it immediately downclocking 2 bins across the board. But hey. It's most definitely fixed the edge case BSODs from earlier, and at least it can fuckin launch Tekken reliably now.
Cheers for posting that. Had no idea about those games and thinking I need tekken now for stability testing and hey comes with a game to play (if my KS can make it last shaders comp)
@@sibzilla They've made it easier to launch in one of recent patches. I noticed I could pass 5 runs by after a reduction of 10mV over the increase (so like 12~15mV net increase versus 20-25~) and an adjustment to Input Voltage/CPU AUX. Yakuza Like A Dragon wasn't happy, and still needed more voltage, but everything besides that game was fine for like a flat month with no issues, idling, etc. There's really no point using stress tests like R23 and similar nowadays - all these modern games hit the cpu's in such drastically different ways even if they have light load.. Better to just find something stable in windows (geekbench 6 stable) then game it out.
well i have a 14700K and I'm using these exact settings but with 253 pl values and it works great. wish intel had a chart for the i7's also. thx BZ for all the hard work. and hopefully we get word on proper IA AC_DC values.
Thanks!
Same cpu and mobo here. Copied your settings to the letter - result 39964 in cinebench 23, I did the 10 min throttling test btw.
1 core hit 100 (core 7) lowest was 89 (core 0). 1 core hit 6.0 all the other P cores hit 5.7, none dropped below 5.4 during the test. 398.9W max power draw
Waiting for an Asus version of this guide! But it looks like Enable all limits throttle at 90c looks to work like this.
How do I get my new Intel CPU working as intended?
Firstly, don’t set defaults. Secondly let’s go deep into the BIOS and start changing many settings.
It’s how Intel designed it. #plugandpray
Blame the mobo companies, they are the ones that have created most of this situation.
Thanks for the content and the video! I'm currently trying to use the same settings, but for an ASUS board. You noted that turning on the CEP would require us to remove the undervolt that comes default with the motherboard. As you noted enabling CEP in BIOS absolutely destroys my CPU performance.
Where would you find this undervolt in the BIOS (I am on an ASUS)?
16:47 Very interesting. Indeed the voltages do seem more normal now.
Would you anticipate that enabling XMP for the RAM would have any effects (positive or negative) on this method for improving CPU stability?
He has it off for testing (I assume) Hard to tell if ram is stable if the cpu is crapping its pants. Super fast ddr 5 on motherboard manufacturers list will also crap its pants. Memtest and gaming might work for days but try running karhu memtest and it will tell you stuff isn't right show an error in 30min or less. I am talking about speeds over 7000mhz. This generation they really lied and hyped way too much.
Hi Buildzoid. I’m basically wondering how you go about correctly adjusting the AC Loadline and DC Loadline according to the Vcore Loadline Calibration (LLC). I’ve watched/listened to your videos a couple of times and tried reading about the subject. But I’m not sure I completely understand it.
I have a gigabyte z790 aorus elite x wifi7 with a 13700k. I would like to use the medium LLC but don’t exactly know how to go about calculating the AC/DC values.
I haven't had any issues with mine, thankfully ,but I had to try this out,since I also use the same bios it's super simple to copy paste :D
I just saw a report from hardware unboxed, that the gigabyte Intel baseline bios was the actual version Intel said that is what it should be out of the box, gigabyte aparrantly had it correct.
I guess intel plans to replace everyones CPU's one the 1.6V idle does it's thing to the P-cores.
I can't believe that 1.7 mOhm of AC_LL ist correct. When I look into the Intel documentation, it says that it should (max) be the same as DC_LL which is max at 1.1 mOhm for 125W and up. And even that gives you some really high idle Voltages. But going even higher can't be right. That whole story is completely insane.
Hey, dunno if you've seen my new comment now but looks like Gigabyte has removed at least my motherboard's latest bios version. Let's see if they're working on something new. This is very odd. I'm now officially on a ghost bios version. lmao
(Z790 Aorus Elite DDR4 1.0, non-AX, bios v. F11f)
@@Angry_Karakuri The official Intel Statement makes clear that the Gigabyte defined Baseline was wrong.
intel baseline has been option on my asrock board for 5 months atleast, the 125/188w.
Never had any stability issues but I never did 6 ghz boost, 5.7 all core with voltage offset undervolt.
Is this worth trying on my asrock z790 on my 13600k? Have a nice undervolt going right now but I believe I don’t have TVB enabled, just 5.1/3.9 standard all core, I know where the AC DC load line values are, im at the medium droop (3) currently
Would be nice to have this issue addressed again after the latest BIOS updates of the motherboard manufacturers. "Intel Default Settings"
Fix stability issue I did have with older software (and gave up a while ago) on 12900KS MSI Z690 ... assume they were not compatible with big small cores. Clearly the only reliable source on youtube ...
You're an invaluable resource on this subject. I've now held off on upgrading to a 14900K until we have a better resolution on the situation. Just out of curiosity, does this setting help 12600Ks and 13700Ks in any way? I've noticed my 150w limited ixt board does not enable any CEP and sets AC/DC Load line to 0.70 for my 12600K.
12600K I see no reason to do anything. And 13700K should probably be fine on it's own since the stock clocks aren't that high
Videos like this spooked me, so recently when I upgraded I bought a 13700k with a Gigabyte Aorus Elite X. I run a BIOS version which is 1 version older than the current one and pre-dates all these new "Intel Defaults" settings etc and so far I've just let it run as fast as it wants. Under cinebench it runs at 210 watts, 1.35 volts and reaches around 75c degrees max under a 240aio. Haven't had any stability issues at all.
Nice, thanks for the video. What would the manual loadline settings you used translate to on an ASUS board?
This is roughly equivalent to level 7 on ASUS. For AC_LL you should try values between 0.10 and 0.20 until the scores look good. You can leave DC_LL on auto since it's automatically matched to the VRM LLC level.
@@ActuallyHardcoreOverclocking Isn't the Intel Baseline Profile on Asus good enough? Or is it still bad? I have mine on Intel Baseline Profile on a z790-F gaming WIFI II and it seems stable so far.
@@ActuallyHardcoreOverclockingis it wrong that i had to adjust DC_LL because vcore was not matching VID? i bumped DC_LL to 1.05 on LLC4 first and VID = Vcore now. then i lowered AC_LL till it was unstable and went up
@@ActuallyHardcoreOverclocking Thanks. I tried these settings on my ASUS ROG Strix Z690-A board and the latest BIOS, but it gave me worrysome results without power limit. Several of the Cores shot up to 100°C during Cinebench R23 and throttled hard (normally it will reach around ~76°C with my custom loop with three 420mm Radiators and D5 pump on full blast which keeps the liquid temperature delta around 3-4°C in a cool room with Auto SVID behavior and CEP disabled, and around ~85°C on Intel Failsafe SVID behavior with CEP enabled), even though I had limited it to 90°C in the BIOS, and the score was way lower than with the Intel Fail Safe settings due to throttling. I was able to somewhat fix it by setting a 125W/253W Power limit with 307A CPU Core/Cache Current limit, but the temps are still higher and scores lower than the Intel Default option in the BIOS this way. Not sure what's going on here, perhaps it's because I'm on a Z690 board and not a Z790, but I think one still has to be careful with these settings on certain boards and should probably set a lower power limit just to be safe.
Is this only an issue for Z790 boards? I recently upgraded an AERO D Z690 board to latest BIOS and put in a 14900KF and its very solid with 2x48GB sticks running at 6200 MT/s. Runs all these benchmarks continuously without lockup/crash. That said, Cinebench R23 still throttles towards the end of a single test. Avg score around 40,500 - 41,000.
I tried it with my z790 gigabyte and 13900k but with these exact settings i have a cinebench23 multicore score of 21000, even lower then with the intel baseline config around 28000. With the unstable default config i get around 33.000. Ive got no more ideas left. And Can i run xmp 1 with this setup? :(
I GOT NEWS LMAO
Hello! First, thanks for your videos. I recently got a Gigabyte Aorus board (a Z790 Aorus Elite DDR4 Rev 1.0 non AX) and I've been checking Gigabytes page daily now for updates. Looks like they've removed update F11f (the one I'm currently one). This is the version with all the "improvements" on it. Looks like they're probably working on a new one, hopefully. This is weird. I haven't seen any sort of announcement but yeah. Maybe it's something to keep in mind and something to look forward to.
Edit: For those wondering, I recently got a 13600K (upgraded from my 8600K, RIP buddy) and temps are still kind of high despite undervolting and adjusting power settings and doing all I can in a custom watercooled loop. And yes, I'm using an ILM plate. lol
But this isn't so you help me troubleshoot or anything. Just letting you know how it's going on my end. Let's see what they've got in store for us next.
BZ, I think you may want to try these settings out. Granted, I have a 13900k, you might see similar results.
13900k; MSI MEG Z790 ACE; LT720 AIO.
55/43/45; HT off; c-states off; TVB all off; PL 1/2 253; Current auto/max; AC LL 10; DC LL 50; Adaptive + Offset auto: -100 mv; LLC Mode 6.
HT on, only difference is AC LL 15.
I’m completely stable in games and stress tests. OCCT large/extreme/variable, I may hit 80C on a 30 min loop with HT off and 87C with it on. I did a 30 min loop with R23 and hit 40,397 and it never down clocked. Unless I’m doing prime 95 or y cruncher, it won’t down clock.
Oh yea, max voltage I hit is 1.259 and drops as low as 1.169 in OCCT. Maybe you can see if that works for you. Obviously some settings will differ depending on the board and CPU quality
I imagine none of this is necessary if I've manually overclocked on water, right? I'm rock solid at 5.8/4.5/5.0 p/e/ring with 1.37v LLC mode 6.
I've been running this exact way since Nov 2022. I play Tekken 8 everyday, have compiled shaders in TLOU, No Rest for the Wicked, CoD, Hogwarts, etc without issue. Manual tuning is the entire point of an Intel K sku. I don't know why anyone would buy an i9 K sku cpu to run "out of the box" settings. You've done well.
I am not buying Intel until these sorts of issues are not even possible outside of what's clearly labeled as overclocking/out of recommended spec. I've been an Intel user from the Q6600 to the 8700K and did (and still do) enjoy tinkering, but after a few years of running a rock-stable undervolted 5800X where most of this stuff is handled automatically, I have exactly zero appetite for dealing with power-hungry, hot, unstable, possibly degrading platforms that even BZ has trouble wrangling into safe territory w/o losing massive portion of the performance.
I dont have a "GT CEP" option on my motherboard. I have IA CEP and SA CEP. Would the SA be approximately the same thing as the GT?
using extreme LLC, wouldnt that be a concern for degradation too?
i personally think its just more straightforward and better to apply a static clockspeed (yes its downclock but oh well you gonna lose performance with intel baseline anyway) and static voltage with low LLC.
What were the settings again? Could you please post the profile adjustments - I'd like to try this
Can this be applied to an I7-14700K? I had to RMA mine because of the unestability issues, and I want that the doesnt break too. I only see people talking about the I9s, and none about the I7s :(
does "C-State Control" set to auto on gigabyte xtreme x z790 mean enabled?
I know this is for 13/14th gen stability issues but would these tips be beneficial on a 12th gen also?
i put acdc loadline to 12 but my motherboard has the mohm measure so i putted at 12mohm for a sec, did i damage something or am i fine? it went there for small times
I know it's off topic, but bz have you heard about the new Sapphire B650 ITX motherboard they just released?
Looking good, I also tuning my PC using DC LL / AC LL / TVB.
But BZ cpu temp is a bit too high ? what cooler he is using ?
(Due to high CPU LL level ?)
I got the same cinebench score, 40k-ish from my 14900KS + 240 AIO (30c room temp).
CPU P core hovering around just 5.5-5.6 when benching (VOUT 1.18v~1.2v)
CPU package ~ 300w peak.
(my KS V/F table at 6.2 is a bit trash compare to another KS)
800 0.774
1400 0.774
2400 0.779
3400 0.869
4300 0.989
5100 1.179
5900 1.413
6200 1.513
Now considering should I delid it for fun or not, I like low temp CPU,
for lower fan noise in my gaming ITX build.
As he told in vid AIO 360, and AIO is NOT good enough imop. way to many think they can use AIO on these 13-14xxx. Aio is only good enough for gaming with undervolt.
@@perhansen3959 What do you suggest? My case wont support anything bigger than a 360.
Hi nice video. I did this settings, and the games are still crushing. any other solutions? I am using new bios F14a for Z790 aorus master
Yes generally people like to have their work deemed successful (validation) hopefully my Asus 13th gen work station/renderfarm will work within this framework
Any kind soul willing to translate these adjustments to an ASUS bios? Specifically I cannot find "Voltage Reduction Initiated TVB" and "GT CEP"
Seems like Intel keeps a very loose spec, and it's been to their benefit (give board makers a lot of leeway and you can target high performance and low price all under one moniker). Historically their chips have had enough headroom that boardmakers, who are always looking to tweak things to get the most out of the chips, could succesfully eek out some extra by playing to the extreme ends of the spec. But these latest chips are pushed to their limits already (especially 14th gen) and so the board makers usual tactics are running into problems this time. TLDR - intel chips usually have enough headroom/leeway that they could afford a very loose spec for board makers and things would mostly be fine. Their latest high end chips have used up that headroom in the name of more performance, which means they can no longer tolerate a loose spec. But Intel being Intel, they don't want to change the spec or admit this. So we get band-aid solutions.
on asus we have ia cep and sa cep, is sa the same of gt cep and should we touch it? pls answer
Great work... Now get to the APUs!
Thanx! You are awesome.
So in your Rambling video you said the only Reason for crashing you can think of is to low vcore. Can you get BIOS Stock settings to run stable in CBR15 with increasing Voltage ? . Just for testing ?
if I set AC_LL to 70 that fixes the crashing (though it doesn't change the all core load voltage it just makes the CPU thermal throttle harder so it runs the same voltage as before but at lower clocks). It also makes the low load voltages even higher.
@@ActuallyHardcoreOverclocking Thank you very much for the good work and sharing with us 😁💪🏻💪🏻💪🏻.
Its unstable for me with these settings but a offset voltage of +.050V stopped giving me some errors
I can't believe I'm watching this version after watching the full length movie
Anyone know why HWINFO reports 0.79v additional with these settings? Bug? Looks like it’s as simple as ‘not using Intel defaults’ will do it.
I'm curious if your 14900k crashes on Unreal Oodle as well as R15. Have you tried it?
IDK how to use Oodle.
Hey man i really understand what you are trying to do but the true challenge is to test a possible fix is through compiling Unreal engine shaders. My previous 14900k before the RMA would pass everything and i mean everything,prime 95 small ffts it would run any Cinebench version for hours before i was stopping it and it would fail in a matter of seconds to compile shaders. I went as high as 1.47v under load even on lower frequencies than what the chip was supposed to be and it turned out that in my case anything above 5.2ghz in the p.cores was just throwing errors in compiling even at 1.47v . The whole issue with Intel chips and shader compiling is extremely weird.
Is there a simple way to test compiling unreal? because I've literally never used it.
@@ActuallyHardcoreOverclocking not a simple way that I know of. I was just using Taken 8, Remnant 2 and The Last of Us shader compiling while I was switching between the newer and previous Nvidia drivers on my 4090 after a DDU driver uninstall to trigger again the shader compiling on these games. The whole process was time consuming and in my experience Taken 8 was the most reliable one to saw shader compiling instability with a close second The Last Of Us which isn't Unreal Engine but it is quite taxing during the compiling and surely takes a lot longer to compile its shaders compared to the other titles.
BIOS date 2044? Man's living in the future !
I had a professional halon fire protection system installed when I upgraded to a 4090 so its good to know that this should also cover my Intel cpu. Imo you're doing it wrong when your not putting your life and house at risk of fire, just say'n...
Hahahahahahahah
don't even own an intel but still watching this genius fix a multi-billion dollar companies issues lmao
Intel probably laid off anybody with actual engineering skills long ago
@@XiaOmegaX Nah, they just cut marketing and influencer budgets 😂.
Any downside for LLC Extreme?
I was set iccmax on my apex Encore 400A, it still got performance like 511A and about only 2% perf loss, but got better 3 degree temp.
I think if cpu dont get crash by default, we can set iccmax to 400A with proper cooling like the galahad ii perf, the perf and temp is better than intel basline.
Great job.once again brotha
"shorter version"
"30 minutes"
never change dude :)
Tried your ideas on my 14600k best option I found after getting a brief glimpse of the recommended core current, was to remove my turbo power limits and just set max core at 200 then leave everything you recommended at auto. just moved my power to 150w still below 90deg running r23. useful video.
the intel doc I'm using is public so you can just look through that yourself.
Hmmmm havent tested my system for stability. I have had crashes/ or problems launching only Unreal Games, seems just choosing a different NVIDIA driver fixes the problem. The finals, Darktide, etc. Otherwise no problems but big problems if wrong driver. Hopefully these problems dont continue to get worse. Dont see taking a performance hit is worth not juggling nvidia drivers around. Of course I havent loaded up Grayzone. 14900k Auros Master X w/ a Gigabyte OC 4090 ...hope this gets worked out. Currently hating Gigabyte and Intel for all the time wasted trying to get games to run. I only started really noticing the issues since release of The Finals. I find it strange it is Intels fault but juggling Nvidia drivers seems to make games run fine. Having to keep driver releases named by game on the desktop is pretty crap.
for me it’s unstable with these settings.. can u give us loadline value for asus
What if I change LLC to extreme without activating VR config and changing AC DC LL ?
you're probably gonna end up with some extremely high voltages.
In this video bz sounds a lot like the kind of guy that will be no longer receiving free samples from Gigabyte :D
I literally just returned my New Asus mobo and 14700k because the system stopped booting into Windows and if it did load the os was buggy, laggy and nothing would open. I'll try this on the new parts when they get here.
Are you getting another make of board?
@@Jacko_486 Yeah, I returned the Asus gaming WiFi and replaced it with a ROG Maximus Dark Hero.
These settings should work good on an ROG strix board as well?
Probably but I don't have any strix boards to check.
@@ActuallyHardcoreOverclocking thanks for replying, ill save my settings and try these tweaks and see how it performs.
Hey. Can you do a quick video on asus motherboard?
Similar settings for ASUS?
I got the best temps with this guide, but score on r23 dropped to 28.500.
I9 14900KF.
Exausted already.
Is setting extreme llc 100% safe in this configuration?
should be way safer than the 1.6V that the intel baseline profile hits.
Would appreciate a helping hand
13900kf/deepcool ls720 360mm/ gigabyte z790 gaming x ax F9 bios whatever default settings in there except for XMP 2nd profile (5600)
After watching initial video regarding instability and having experienced really infuriating crashes during shader compilation in some games with errors saying I lack video memory (which is not the case) or just app crashes like trying to teleport in D4 (it seems there is a spike in CPU activity, and it crashes the game) just to name a few decided to try checking if I get the same behavior in cinebench
and it's the same - R15 crashes, R23 crashed on first run, second run was successful, max vcore around 1.5V, idle 1.384V
Score was 39492 points with temp limits hitting 2 P-cores at 99 others were around 91-92
and something called "Package/Ring Thermal Throttling" was also switching to Yes during the run
CPU package power was hitting around 350W (i think, don't remember exact number but over 300W)
Applied first fix (TVBs on, IA GT CEP on, current limit 400 max 4095 power limits, CPU Vcore loadline calibration set to Extreme, AC DC LLs to 12)
Instability in R15 went away, but now score in R23 hits only 29400 tops.
Package power draw is down to 287 watts, Vcore max was 1.392 idle 1.270-1370 (not really lower that on default settings)
In hwinfo's "performance limit reasons" section I see two things kicking in during the run
IA: Max Turbo Limit
RING: Max VR Voltage, ICCmax, PL4
but at the same time package/ring thermal throttling doesn't get triggered.
Why there is such a drop in performance? Core clocks are the same as previously (5.5 Ghz for p-cores, 4.3 Ghz for e-cores)
Might it be that z790 gaming X need different specific AC DC LLs for that board instead of 12/12? (default settings were 40/90)
But I am having a hard time figuring out how to even properly understand what these AC DC LL parameters do, so not to kill the cpu with further experimentation with LLs
EDIT: went back to default settings by manually switching back everything, R23 was crashing even before showing the first block, on 4th try system locked, went back to bios, rechecked everything, applied "optimized default settings" just for shit and giggles, booted back, R23 worked, score was 38000, max temp 95, "Package/Ring Critical Temperature" did not trip
Package power was 305 watts (probably it was this initially i just remembered incorrectly that it was 350)
VR VCC Current was 232A
Max Vcore 1.428V, idle 1.284-1.370V
IA: Max Turbo Limit and RING: Max VR Voltage, ICCmax, PL4 kicked in during the run
oh, and i was checking wrong core clock speeds metrics it seems, there are effective clocks, and on "12/12 rails fix" it was 5.3Ghz and 3.8 Ghz p-cores, e-cores
After experimenting and watching last video with intel providing baseline defaults, I switched loadline calibration profile to extraExtreme (the flat line), still was getting ~30k points in R23 with 12 AC/DC LLs, and 1.35v vcore with 1.42v spikes, switched to 13 and it started looking much better, ~38k in R23, and stable 1.38v vcore, but hitting 1.404v in idle lol, decided to see if there will be any difference with setting it to 40/40 and scores even lower 35-36k, but it also uncovered a different issue - there is almost 10 C degrees delta between p-cores 0,1 and the rest of p-cores constantly hitting 99-100 degrees, also it seems i can only run 5.5 Ghz (but it's not stable and crashes on 2 or 3 run in 10 minute test) with default gigabyte settings, with settings above it maxes out at 5.3 ghz. Probably due to thermal throttling, as there is a constant delta of ~10 degrees with 6 cores hitting thermal limits constantly, probably due to ilm bend, I was thinking of installing a contact frame just in hopes to drop temperatures during more intense gaming sessions (like 80-85 degrees), or continue with this settings as probably will not even notice but again, i want mine 5.8 GHz turbo clock ))
Looking for help finding an Asrock video like this, please leave a reply if you know a channel that has arock mobo info for 13/14th gen.
One question, how do you addresa all previous videos where you pump all these stuff on BIOS and now limiting the stuff????
It is called experimenting. He was brainstorming what he thought at first would help with stability. Usually bumping up voltages will fix unstability. However he found that it was making the cpu's degrade faster by having higher Temps than previous.
SP 100 14900K testing on the same board actually can pass R15 on BIOS defaults, does that mean no need for adjustment?
How much do you trust intel to replace the CPU if it breaks? Because IDK if my settings meaningfully improve longevity/reliability but I did try to get the voltages as low as possible while maintaining stability and not compeletely destroying performance.
If the extreme loadline is not stable what else can I do?
you're still crashing with IA_CEP turned ON?
@@ActuallyHardcoreOverclocking no crashing just no power 13900 kf clocks normally at 5.4 mine with your setting does 4.2
@@onlinegaming4826 try set AC and DC_LL to 20.
@@ActuallyHardcoreOverclocking Nop not working
thnx!😀
MSI's latest BIOS actually disables CEP for 14th Gen
It doesn't disable it. It just gives you the option to disable it.
Hand back BZ you creepy space aliens! There is no way this is the real BZ. Short videos?! Never! 😁🤷♂️
Why don’t you have access to their Data? They should provide it to you
Thanks, you were rambling in the original video and was waiting for point.
we're here for the rambling
@@spacechannelfiver Yeah, well I'm not. Make your point and move on.
@@JordanJ01 lol u cant just make point and move on... we are talking about processor that cost a lot and all this explaining is in place... but can he make it better with timekeys? ... YES!!!YES!!!YES!!!
25:29 >> Laptop CPUs run at way lower voltage (and current), and as such.. lower frequency at full load. Smaller chips(dies) can also cope better with very high and fast thermal-cycles.
Same reason you generally see more longevity with GTX 970 than 980 Ti ...unless you had a big cooler on the 980 Ti, such as the Gigabyte Xtreme model (keeping temp cycles slower and less dynamic).
980ti mainly die from vrm failures. The gigabyte cards tend to have the memory vrm blow up.
Will we get a vid with an asus motherboard 🥹
🤔 does this mean we'll be able to pickup i9's for pennies on the dollar? 🐢🐢🐢🤔
There is a new BIOS for my z790 Strix-H Asus board Can you test it and tell your thoughts about it, please? :) Thank you
"1. Introduce the ""Performance Preferences"" with options for Intel Default Settings (Performance/Extreme) and ASUS Advanced OC Profile.
2. Redefine the factory defaults based on Intel’s new ""Intel Default Settings"" for various CPU SKUs.
3. Change F5 from ""Load Optimized Defaults"" to ""Reset to Defaults"".
4. Add warnings when users switch from the defaults to other settings.
That would be very nice! Gigabyte also release a new bios update.
intel probably got rid of their quality control, they just put the cpu on the market and wait for buildzoid to put out the video, then they make their corrections.
Can't be arsed edition
Funnily enough the high frequency high voltage idles doesn't... quite degrade those said frequencies... but it most definitely makes the cpu more sensitive to 92c-100c(+) scenarios. Looking at your cpu, and remembering to my freshly bought one, naturally it redlined 100c at "stock". But it managed at those situations and could sit there.
But I've noticed whea and watchdogs induced by temperature. For example R23 would corrected whea often at X config, but after repasting and temps dropping a few degrees, those corrected errors disappeared at the same settings.
It's really a struggle to tame these CPUs because certain workloads can just fuck it, now that devs are making games on DX12 with obscene shader compiling stress, and it's different per game.
Tekken 8 is a game that makes the news for this, and it literally only takes a few seconds onto launch to bsod or error. I could pass INFINITE GB6 and did like 6 runs of TLOU shader compiling, but it was not, and is still not, having it on Tekken 8 reliably. Interesting phenomenon, because heat induced errors are at play too due to the die setup of 3/5/7 being super hot cores, specifically 7.
These video settings are still fucked though. Your LLC pairing with DCLL may be wrong because of such insane wattage and amp reporting, but when you're running r15 and 2b, I'm seeing a max of 295 AMPS over 300W at 100c.. thats still hard degrading
I know people who degraded 13th and 14th gen CPUs by running single threaded memory benchmarks at 6GHz+. Also isn't it kinda strange that heatsink reviewers aren't reporting their 13th and 14th gen test bench CPUs dying but random gamers are complaining about CPUs getting progressively less and less stable? Maybe shoving 100-150W through just a couple cores isn't a good idea.
@@ActuallyHardcoreOverclocking I'm always shocked when seeing review or tech channels, big ones too, just carelessly benching their CPUs at "default" and I see 350W+ 14900K/KS at above 1.25 load volts and sitting 100-105c for the 10-30min loops that they say. That leaves terrible habits for the viewers who later get their chips and do the same thing, after which they've just degraded by a few mV already on all core.
I personally haven't experienced high frequency/light load degradation on mine (just observations of much weaker temperature tolerance under load) like an inability to hit clocks. IE; the same settings that could infinite pass GB6 transients still pass it now at the same scores, but that's at a range of 1.46-1.5v ie. R23 SC (proper one, not launching with specific threads set, that is dangerous), SuperPi, Idling/Browsing with a medium strength LLC.
A high strength LLC with a 3-5 core active high frequency boost is dangerous because many games will do it, and the voltage won't droop to typical 1.3-1.4 gaming levels and instead that's where you might see 1.45+ at 80+ Amps/100+ Watts because it doesn't droop.
05:30 Why are you living in the year 2044?
What kind of instability? Never had this problem...
The shorter version is 30 minutes 😅
Engagement
31 minute is short? :DDDDDDDDDDDDDD
I think its the fault from intel and the MB vendors. They didn´t test these CPUs accurately. They are lazy.
I've tried your settings on a 14700KF, i got much better score than the power limit of 253, but higher temps also...
well yeah these settings will pull more power than a 253W power limit if there's cooling headroom
I don't know why Gigabyte and Ausus aren't getting the blame here. This was all they had to do? Have someone like Buildzoid who knows something spend a few days working on their BIOS before they publish and ship it? Why isn't this their responsibility? It seems like they were just completely lazy. And how does he come out with a better BIOS config to fix this than them before they do? I really think people are letting the board manufacturers off the hook. Every big name youtuber blames Intel for this. Even Buildzoid. Is it just more friendly to the algo? I don't get it.
Intel makes the design guidelines. For over a decade now the intel design guide lines have basically been: "do whatever you like just don't touch the core clock ratios" and for years it was fine because intel wasn't shipping CPUs with boost clocks higher than 5.3GHz.
@@ActuallyHardcoreOverclocking So what? Not every board manufacturer did this. Not every model of Gigabyte board even did this, right? It's their job to make sure things work together. And some did. They didn't on this board. But that's literally their function. That's what the motherboard actually does. They write the BIOS for godsake. Don;t you think they should check to see if it works lol? And if it doesn't then update it? Which I guess they still haven't bothered to do or I'm not sure why you're doing this.
@@posmoo9790 board makers won't invest more effort than they need to if they are not forced by intel to make these CPU's long term stable which is why intel should be blamed for this and not board makers
look at AMD and how they further tightened their guidelines after 7800X3D's burning up situation while intel ignores issues for 10 years and still refuses to fix those issues
@@posmoo9790 what is really stupid is if you download XTU , that is made by intel, and use the AI optimizer to actually overclock the cpu, XTU sets the icc max to 350 to 400 depending on the the cpu model.
@@xthelord1668 Listen, I buy AMD because I care about not turning my room into a hotbox and I don't have a desire to mess around under the hood and get the most out of what I bought. It's pretty ironic that people whose hobby is tinkering with this stuff are just begging Intel to lock everything down because board makers can't be arsed to make sure their settings are stable - people who have bought brand new Intel CPUs and boards are finding these problems out of the box. That's on the board makers. And careful about getting what you are begging for Buildzoid. I'm gonna laugh my butt off when they lock everything down next gen and you all cry about that because your hobby is gone. And why shouldn't they lock it down? The people who should care most about keeping these specs open are thrashing them when a little due diligence by the board makers would have kept this under control.
I'd love to watch this, but it keeps stalling 3 seconds in