Rambling about the new Intel 13th/14th gen Intel recommended default settings

My Patreon: / buildzoid
Teespring: teespring.com/stores/actually...
Bandcamp: machineforscreams.bandcamp.co...
Twitch: / buildzoid
Twitter: / buildzoid1
Facebook: / actuallyhardcoreovercl...
#overclocking #intel #14900k #13900k

[@Atilolzz]

This is actually an interesting topic tbh, gladly listening to the ramblings here

[@cemsengul16]

I feel stupid for buying a 14900K.

[@vincentvega3093]

Live and lern... ​@@cemsengul16

[@RobBCactive]

Intel just seem to have made this worse. You'd think they'd know what their CPUs can operate at for an optimum performance, after all they are designing, fabbing, testing and binning the damn things to meet some spec. But are the mobos even reporting the numbers correctly?

[@deadlymecury]

@@cemsengul16 I feel stupid for buying a 14900ks to solve instability issues with a 13900ks... Double stupid because I also stepped on a landmine called "EKWB Direct Die". Current build is heck of a ride. "Should have bought 7950x3d instead of 13900ks initially" kinda.

[@cemsengul16]

@@deadlymecury Yeah I am going AMD next time for sure when I build a newer rig in a few years.

[@jasonbuckner5858]

I actually like listening to your rambling. I'll turn on a random ram tuning video and listen to you on the drive to work.

[@antondovydaitis2261]

Also beats "Rainstorm on a Train" or "Learn Japanese While You Sleep."

[@punishthemeatpocket]

same his voice is soothing

[@TheMagicCatQ]

@@antondovydaitis2261 you're a frakin' genius... this is perfect to fall asleep... it's like listening the endless pointless powerpoint meetings from work...

[@TheMagicCatQ]

don't listen to this guy while driving... you might fall asleep...

[@antondovydaitis2261]

Thank you. I learn a lot from you, even though I may not ever apply most of it.

[@mortlet5180]

24:22 ; Yes you can verify if the settings match the physical hardware implementation if you have an oscilloscope. You just look at the same magnitude Icc step response, at a constant steady state operating point, average multiple captured events to improve SNR, and measure the % overshoot/undershoot, the damping time constant and the ringing period. You can even directly measure what the hardware values are without the VRTT if you really wanted to. That would involve disabling *all* power saving, DVFS, etc. settings and configuring the CPU to run at a constant and conservatively low all-core frequency and voltage (e.g. 5GHz @ 1.3V) such that the CPU's Tc remains at a consistent value (somewhere around 50-65°C) under full load, with a significant (100-200mV) Vcc in excess of what is required for stability. Then, while running a very consistent load with no observable transient loading behavior, you inject a square wave AC current with an amplitude roughly 10-20% of the current Vcc amperage and a frequency of 1-10Hz, into the Vcc rail at the socket and record the resultant voltage fluctuations with an oscilloscope. The ratio Vout/Iin = Rout thus giving you the DC output impedance, while taking an FFT of the transient behaviour at the switching edges will give you an idea of the closed-loop Bode plot of the VRM for rising and falling transients respectively.

[@HOMERUN1379]

Was running into the issue of fortnite refusing to launch with dx12 and saying it was out of video memory when I have a 3090. I found your video where you said that it was related to instability with my 14900k. So I updated my asus bios and applied Intel base line and it fixed my error. Thanks!

[@LaKillingFrenzyy]

I'm so confused. I thought that baseline is outdated.

[@trixniisama]

​@@LaKillingFrenzyy It can still fix instabilities though

[@mxgreg]

thanks for making this series of videos!

[@someusername121]

In non-public documentation: the Vrel for Raptor Lake is 1.45V up to 70C and then it drops to about 1.3V at 100C. And the whole deal with eTVB (because that's got shitty documentation) is that it basically jacks up the voltage closer to Vrel to reach higher frequencies, and you'd hit 5.8Ghz no problem up to 100C and 1.3V (in theory) but then it's like "hey we need 6Ghz to be competitive so let's crank to 1.4V". "So....yeahhhhhh" best to make sure the thing isn't slamming more than 1.45V in for extended periods. I wonder if Intel will do an Nvidia and lock you to Vrel as a "PerfCap". In other fun facts, the 14900K was 'hoped' to be 6.2Ghz and the KS at 6.5Ghz.

[@vac59]

Casual 100c for retail customers 🤣🤣🤣

[@jayfrog5446]

I would like to ask some question please. I use recommended low latency settings like, disabling hyper-threading, disabling all E-Cores so I just use some raw 8 core 14700k. On top of that, I also only use 5.4 GHz, maybe 5.5 GHz on all cores later. Do you think I need to worry about something because I never saw 1.4 v or anything like that under load? Thanks!

[@someusername121]

@@jayfrog5446 that setup sounds good to go. Maintaining 5.5Ghz requires very little voltage…like under 1.3V.

[@jayfrog5446]

@@someusername121 Thank you! I don't even hit 1.3 so I will stick with 5.4ghz because it's fast enough. Best regards!

[@LuftManu]

Hey man! Just said this a few months ago in one of your streamings. Thanks for checking this! I knew something was up. Kind regards

[@bauerns5er]

Thank you a lot for this Video and your explanations

[@bthjf12003]

Thanks

[@coldsleepingcreature]

Thanks for rambling, bought some merch!

[@MhillPlays]

I'm using the Asus Bios 1102, i don't experience any crashing myself. I use Enforce all Limits with PL1 = 253 PL2 = 253 ICCMAX= 400A, when i enabled IA CEP/SA CEP it seriously destroyed my scores in both R15 and R23. I usually hover around the 37K mark with my power profile in R23, with CEP's enabled it was just 24k! Using an Asus Maximus Formula Z790 + 14900K AC 0.550 DC 1.100mohm seems to be what this board does on Auto. Update: Using the latest Bios with Intel Profile, has odd behaviour. Temps loading games are very high, much higher than Cinebench runs. Yet actually playing the temps are somewhat in the normal range. The only difference is AC/DC are now both 1.100mohm.

[@SahajPS]

Looking forward to your test video

[@kiwivda]

I totally agree. I'm telling this since 10th gen. They are unaware of how these cpu works, and Load Line is not calibrated.

[@outlet6989]

I'm giving your video a thumbs-up. It's an excellent video to watch if you're into overclocking, which I'm not. I understood some of what you were saying. I've watched some videos recently that tell me some MB manufacturers have released new BIOS updates to solve the problems discussed in your video.

[@Analogtodigital]

For what it is worth, on my EVGA Z790 Classified with the -50% droop setting places the AC/DC load line at 0.280/0.280 per hwinfo

[@ActuallyHardcoreOverclocking]

Oh I guess I'll check my DARK.

[@Oomlie]

​​@@ActuallyHardcoreOverclocking let us know what you find, I still am on a 790 Dark as well and EVGA emailed me back saying they havent implemented any intel changes - with the 14th gen beta bios

[@Analogtodigital]

It just sucks that EVGA is dead for all intents and purposes. I REALLY like my Classified Z790

[@HartFalleon]

That's my whole question with this, why are there so many customized settings that are left up to the motherboard vendors?. If you have a spec for your CPU, clearly enforce that so that everyone who gets your product gets the performance that is advertised when the board is set to its defaults. It shouldn't be a gamble to the consumer based off of what motherboard they go with but that's what has been allowed for many years. There can be overclocking and tweaking but it should be clearly separate and not an "auto" setting that board vendors don't even give an accurate tooltip or description for. If AMD can enforce locked voltages and power targets for certain processors, I see no excuse why Intel can't do the same but we all can guess that excuse. Is so that their product beats their competitor.

[@CHA0SHACKER]

The answer is that Intel is mainly an OEM company that likes that their direct customers have the ability to „differentiate“ themselves from their competitors.

[@deidian635]

All CPU(and hardware) are configured the same in all cases: values are written in an arranged location in RAM during boot sequence then the hardware reads the location. Changes in OS runtime do change those RAM locations. Essentially locked CPU have all this knobs that you know from Intel K CPU and all AMD have their own knobs, whether they can be changed in UEFI or not. Enforcing operating values is via contracts with board manufacturers the same as whether those values can be exposed to the consumer or not. Basically motherboards vendors have to have access to the values and many are necessary: imagine laptops running in cooling constrained scenarios, they need to change PL1 and PL2 to fit their needs, it just doesn't cut it the CPU manufacturer setting defaults. The chip maker can enforce baseline values(Intel until now recommended but seems to be heading to enforcing them) but there's always going to be better performance: look at the GPU market, not all have the exact same performance, just the guarantee that they meet the baseline and the design has the approval of the chip maker. Almost everything is not shown to the consumer aside power range and frequency offsets for core and memory. I personally don't like that because I'm a tinkerer, but I understand the stance of such chip makers towards their consumers and assemblers: consumers may like control but then they will not take accountability of how they use the hardware(will try to lie and hide under warranty/grey areas) and assemblers have sometimes put low quality assemblies that have given issues due to competition. In an ideal world people would be able to live with Intel's current system if everyone wasn't trying to weasel out of consequences when they break it.

[@Torbjorn.Lindgren]

Some of it is because the "correct" AC/DC loadline heavily depends on the VRM setup - which is why Intel provides tools to measure it, it's just that all the MB manufacturer ignore those because that's not how you "win" motherboard comparison and a lot of Tech-sites will heavily promote your MB because it's 0.5% faster. Newsflash - it's only possible if you violate the Intel guidelines, there's a genuine argument that many tech sites are part of the problem (under the "aiding, enabling, promoting and abetting" clauses). There's even valid reasons for AC and DC loadlines to differ *slightly* some of the time but MB manufacturers have unfortunately shown time and time again they can't be trusted with those kind of controls and it's small enough that it probably should just be one setting. Honestly, I think Intel might be best off going with something like "use these standard value, or values from an independent Intel approved lab" approach for the future (and really do revoke the lab certification if those cheat). And some of the other settings are really intended for extreme overclocking, you DON'T want TVB, CEP or ICCMax on LN2! So for those Intel may need to just mandate that they can't be enabled by default. Which they should have done from the start, Intel is far from blameless in this.

[@jjlw2378]

Locking down vendors is what everyone complains that Nvidia does. I like that Intel gives vendors freedom. Sometimes issues happen but in the grand scheme of things, recommendations with freedom is far better than strictly enforced limits.

[@Torbjorn.Lindgren]

@@jjlw2378 Yes, the problem here is that the DEFAULT settings are wildly inappropriate - not that they exist.

[@cemsengul16]

Since there is such a wild silicon lottery I really wish that motherboards had function that the first time you insert a brand new CPU into the socket the bios does tuning and matches everything perfectly to your sample. Imagine you build a new rig and on first boot it needs to take 1-3 hours running tests and then it works at perfect efficiency for your chip. We are supposed to read a textbook to tune an i9 CPU these days.

[@xBINARYGODx]

there wasnt a wild lottery - most cpus, such as the 14900k can just 5600 just fine - but you have to set the settings yourself and not rely on a whole load of auto settings

[@RobBCactive]

So you do that, then 6 months later Tekken 9 comes out and your magic calibration now is crashing. The whole point of enforced specs and binning models that meet the grade of frequency/power consumption is predictability.

[@tommihommi1]

that's called VID table

[@timothygibney159]

@@xBINARYGODxIt’s tough to do without damaging your cpu. My 13900k was hell with under volting and .3 llc at llc level 4 to get it 5.6 with being in spec. CB still down locks to 5.0 GHz

[@LDWilliams]

Your sarcasm doesn't just drip through, more like open the flood gates 😂😂 Love it!

[@drgoodness7]

"how do i keep doing this" hahhahahaa imagine me just watching..... thanks man for the info, im learning a lot because of you.

[@JJFX-]

So basically Intel 'allows' board partners to juice performance figures beyond what they could reliably do. Reviewers put out their numbers then once it becomes a problem they simply point the finger and effectively tell people to run slower settings. Well done lads, you've killed default settings being the one thing we could count on right before Zen 5 launches.

[@cemsengul16]

Which is why I will switch to Zen 5 when it comes out.

[@xBINARYGODx]

@@cemsengul16 your not even a rounding error and no one cares - also, are you like the people who are forever threatening to dump windows for linux? If you really cared, you would have switched already.

[@RobBCactive]

Review kits send out selected CPUs with boards and RAM. Any silicon lottery loser will be weeded out.

[@kosmosyche]

@@xBINARYGODx Oh, he is a rounding error, all right, but do you think that people in general don't pay attention and make conclusions from this shitshow of a situation, including about their future CPU purchases? 😆 This irresponsible and frankly infantile behavior by Intel in regards to their specs is a hit to their reputation and right at the time when their competition is knocking it out of the park with their products. I have no doubt, these chickens will come home to roost.

[@cemsengul16]

@@kosmosyche Yeah I meant what I said because my previous processor was a really old 4770k and I didn't deal with any bullshit so I will switch to AMD next time. I have no loyalty to Intel, all my previous purchases from them were perfect but I was burned now so I will switch.

[@adamsleath]

depending on your sku/silicon lottery sample. am i right to assume that different samples would have different power limits? im an i5 buyer and im guessing the power limits ill be setting will be lower than the ones in the table ("intels recommendations: 'intel default settings'") in this video. ?? i'd prefer to control thermals with power limits...or current limits whatever the hell it is. ...

[@mattlafy]

Maybe intel needs to start providing VRTT tools (and education) to some reviewers if they don't want to test motherboards themselves. If intel refuses to be involved at all, reviewers might be able to get manufacturers to show them the output of their VRTT tests.

[@someusername121]

The tool is very expensive because it's basically an oscilloscope that plugs into the cpu socket. It costs $10,000 and each socket is another $1000 for the interposer thingie, and another $1200 for a yearly license. (Intel does sell them if you're a registered designer)

[@ActuallyHardcoreOverclocking]

@@someusername121 that would be within the budget of say LTT...

[@YGadgETech]

@@ActuallyHardcoreOverclockingis it ok to just set a low value for ac/dc and if crashing start to raise it slowly? Like starting from 0.7 / 0.7

[@TheMagicCatQ]

can't imagine anyone rambling almost one hour about this... this is just brilliant stuff

[@kevinzhu5591]

The problem of making AC LL = DC LL = LLC means that we need a huge adaptive offset to undervolt. I tried to set DC LL as close as AC LL with IA CEP enabled to somewhat undervolt better (for example on my motherboard, 45 AC/65 DC LL, LLC Low), but IA CEP kicked in and I got a lower score on Cinebench R23. IA CEP lets you run your system without crashing but lowers performance. Thus I turned off IA CEP and just limit the Power and ICCMAX.

[@EddieGD21]

BuildZoid, could you relate what you said about motherboard default settings to the MSI world. Does MSI have a setting for IccMAX bit or do you just have to enter the Max Amps you want without further input? And what about AC and DC syncing? I've seen past reports that Gigabyte often significantly underreported Voltage and MSI typically was slightly over. Is this a true statement, still? And if so, is this because of how the AC and DC LL are configured from the factory?

[@AndrewHines42]

Seeing speculation that part of the issue is that since z490, acll has been acting like there’s a full current at all times (307a?) and that’s why there’s so much idle boost, instead of just being a load scaling compensation for droop.

[@CaptainShiny5000]

Tbh, this generation of Intel CPU's don't seem to be created as actual viable products but rather as placeholder CPU's to have at least *something* on the market.

[@razoo911]

13rd gen and 14rd gen are the same cpu/architetture silicon, with minor difference in the nunmber of e-cores and clock speed so yes you're right

[@megadooooom6582]

one question does AC/DC load line not configured correctly over time might make your board catch fire? or the CPU die faster?

[@andytroo]

25:11 there is a "minimum value" column there - replace the "--" with "SAME AS DC_LL" -- that is what the documentation should say - most people would argue the exact min/max spec sheet is more authoritative than a technical preamble text ...

[@JynxedKoma]

These clowns need to be in court facing a class action lawsuit.

[@cemsengul16]

Yes! I am not buying Intel for my next rig that's for sure.

[@SolarianStrike]

Even if the court rules them guilty, it will be the US tax payers that should the bill. Intel can just get the US gov to bail them out, like they did already.

[@adjoho1]

I'll say. I flushed my loop, and on rebuilding, had a bad contact on the cpu block. powered up, and 3 minutes later the system hard shut down. Huh weird I thought. Hit the power button again, and it's error 00. Inspected, and found the bad contact. Because the ASUS bios settings had removed the temp limits, rather than shutting down at 90C, the system ran until the cpu died. So now I have a dead 13900KF - thanks to fucking ASUS and their bullshit bios settings.

[@POVwithRC]

​@@adjoho1😬

[@vincentvega3093]

RMA and sell the new chip ​@@adjoho1

[@ChrisGR93_TxS]

my question is : does Intel know if their cpus are capable of doing their temp/voltage/frequency table? Seems to me that if you have a rly good cooler and your cpu is not capable of doing 6ghz at all, its gonna crash no matter the voltage or temperature. I might be wrong but from my experience cpus have rly steep frequency drop and without extreme cooling and voltage you face a wall pretty fast

[@NINEWALKING]

Short answer, no they don't. At one side of the story they are selling practically pre over clocked products that they can't guarantee the levels of the actual OC. So clearly Intels own fault. At the other side, with current power level and variety of mainboards, cooling solutions, cases and other heat sources in the case how can you guarantee that some speed will be stable for everyone at that power levels? You simply can't. So they rely that everyone else will do their part good. Not easy scenario at all.

[@someusername121]

Oh god yes. There's literal teams of people whose sole job is to understand VF curves and guardbands and all that shit with the voltage.

[@betag24cn]

intel is in panic mode pushing nonsense and too many watts and allowing motherboards to push even more they knew but they kept their mouth shut in case this didnt happened inteo is really in trouble rigth now and next gen no idea how bad can be, amd is really running over them nonstop

[@qT_p13]

It's a madhouse!

[@williammurdock3028]

On my ASUS ROG Strix X299-E under CPU power management the Current Limit is PL4 when displayed in HwInfo. Funny the bios labels it as amps, so does XTU, but HwInfo shows Watts.

[@kpm4620]

I could understand matching AC/DC LL for a baseline/stock setting. But I wouldn’t say it only makes sense to do that. If you UV, it definitely makes sense to lower your AC LL. I do believe your DC LL should always match your LLC. I’’m running, by far, the best UV I’ve ever been able to achieve on my 13900k with a .1/.5 LL and mode 6 LLC (.6) with a -.100 UV

[@kpm4620]

Sorry, that’s with HT off. .15/.5 with HT on

[@ActuallyHardcoreOverclocking]

sure if you want to UV the CPU your self then set the AC_LL to whatever you like. Whatever instability you (do or don't) create by doing that is your problem. But out of the box the motherboard shouldn't be setting an undervolt that may or may not work on all CPUs.

[@kpm4620]

@@ActuallyHardcoreOverclocking agreed. I took it as you saying you should never have a mismatch…that it didn’t make sense to.

[@cemsengul16]

Man I don't understand any of this shit and I shouldn't have to. The chip should run properly the day I assemble my new rig. Intel is smarter than this but in their pursuit of AMD they have done shoddy practices.

[@kpm4620]

@@cemsengul16 100% correct. You should be able to plug and play. It should only have an overclock and/or undervolt if the user chooses to.

[@Greez1337]

Just checked my MSI Z-690 A-Pro (Bios ver 1.B0), and on Hwinfo, MB set: 0.010/0.010 mOhm This is with a 13700K, though. Although I did hear people having some instability issues on them.

[@ufukpolat3480]

Intel's fault. They had a fully integrated voltage controller up until Haswell/Broadwell which did wonders for low load power consumption scenarios. They wanted to pass the cost to the motherboard vendors and the vendors decided they can justify eating this cost if it means they can get more performance and marketing out of it.

[@andrey7268]

FIVR had problems of its own. In particular, it's another heat source right there on the CPU package. And Haswell already had thermal problems due to poor TIM.

[@kelvinjinxd]

AC/DC and LLC are independent settings, but to tackle the same problem. AC/DC changes the SVID values based on the load, while LLC changes the VR out behavior by given SVID and current. The final voltage is the compounded result of these two settings. You can run AC/DC very high such as 1.1mOhm, and run 100% vdrop LLC(the droopiest). You can also run AC/DC very low such as 0.01mOhm and run very strong LLC like 15% vdrop LLC. However you should not run high AC/DC values with strong LLC. That will lead to massive overvolt.

[@deidian635]

AC_LL and DC_LL are informative values the CPU needs to know because they speak of the board power plane physical impedance, which the CPU doesn't know about. Then the CPU does it's power and SVID estimations based on those values. LLC is a board setting to adjust VRM impedance aimed to reduce voltage transients(undershoot and overshoot): since the VRM is part of the power plane in the board obviously changing LLC means changes in AC_LL and DC_LL. You changed the power plane impedance, you need to tell the CPU about it. They are related but don't tackle the same problem. If everyone acts in good faith the thing works, if the board lies to the CPU when reporting AC_LL and DC_LL values get thrown off.

[@tamamoko9725]

​@@deidian635 Hi, If I applied the Intel suggested settings from the table in the video (other than the AC/DC LL cuz obviously we don't know what the 'correct' value should be at this point), what do you think I should set the LLC to? Or does it matter at all?

[@olega5141]

Actually Hardcore Stabilizing!

[@NINEWALKING]

BTW I have installed Apex just to be sure does it work or not. With my normal setup it worked. Cinebench 15 was crashing sometimes. All other game worked accept the issue with the HZD. HFW worked in any setup and HZD was never working. Then I have made many alternative setups and some would resolve even the occasional Cinebench 15 crashes. Now I made new setup with all from the Intel document and power draw is smaller and everything works but the HZD. I am not sure that it is the issue people have. Currently on 330Watt power limit, no under volt, CEP on, 50/50 AC/DC LL, load line calibration at mode 6 on MSI Z790 Tomahawk with the latest bios. Still testing with all the application I use. To be quite honest I do not see much difference with anything I use. I did not care if the single version of the benchmark occasionally get stuck. Now it does not and still does not make any change. 14900K Funny thing is that with my old setup CPU Was running all P cores at the 5,7 GHZ in Prime95 and with this new setup it seems to downclock with the 24 threads small FFT. It was stable back then as it is now but it does not boost as high?? Like now it goes anything from 5,1 to 5,3 GHz on P cores but mostly at the 5,2 GHz. E cores are anywhere from 4,1 to 4,3 GHz and were pegged at 4,4 GHz. And now I see in other Benchmarks it is nerfed as well with this setup. Seems like way more testing is needed. I think I am going back to my old Bios profile.

[@ActuallyHardcoreOverclocking]

mode 6 on MSI is probably a lot more Vdroop than 0.5mOhms which is probably why HZD is still crashing.

[@NINEWALKING]

@ActuallyHardcoreOverclocking I am just playing around finding what works for me. Played with or without CEP, with or without undervolt. Played with different power and current limits. It's like changing 6 things that all together make load line droop curve and the power draw figures. Limiting power never gives accurate results. Limiting current often downclocks the CPU. I have found a strange setting where I get it to work 6GHz in two cores and 5.7GHz in all core loads, and it runs both Prime95 small FFT'S 24 thread and all versions of the Cinebench. It even runned HZD once, but I was getting progressive audio issues, and after exiting, it didn't exit, really. Normally, to get Prime95 working with no down clock It draws 360 up to 404 Watt. This stupid setting where it works the best draws only 310 Watt and doesn't crash. Apex works as well as all other games I play. HZD is stuck again while HFW is running. We'll to be honest F them all. Intel and mainboard companies and all developers and their code. You give this ridiculous amount of money, and you can't run an age-old game?

[@andrey7268]

24:59 The spec does indeed say "at *maximum*, AC_LL should be the same as DC_LL". The newly released blue table describes "Intel Default Settings", which is basically the most conservative settings within spec that should result in a stable CPU. So, "Intel Default Settings" chooses AC_LL equal to DC_LL to maximize stability (at the cost of increased voltage), but AC_LL values below DC_LL are still within spec.

[@deidian635]

AC_LL and DC_LL must be equal because if they're different it's basically cheating the power limit: if AC_LL > DC_LL the CPU package power estimation will be lower than reality(aka cheating PL1/PL2/IccMax). DC_LL is used to estimate power draw, while AC_LL is used to estimate SVID requests to the VRM: both are to compensate Vdroop between the VRM and the CPU die. AC_LL and DC_LL are dependent on the board but heavily related to LLC: that's because LLC it's configuring electrical resistors in the board VRM which changes the impedance between VRM and CPU die. Or shortly: since LLC changes Vdroop AC_LL and DC_LL need to be modified accordingly. Board makers should make this: especially because LLC is usually spoken of levels rather than just saying the mOhm the resistors in the VRM will be configured to and because to do so is necessary to know total board impedance from VRM to CPU die, not just the VRM resistors impedance.

[@PeterMarszalkowski]

I have the ac dc load line active on, I have the clock speed too high, 1966 at 4x14 is the limit, after that it goes crazy when you restart, you also have to reset the xmp timely, you can switch the switches off quickly by turning them on three times with the image on. The F8 f6 command brings the xpm back to standard and it's nonsense that I can turn off the auto set, yes they do that to me, you have to increase the board voltage from 1.24 v1.34 volts to 2x to 1.48 volts, then the xmp will run too, just not in the car after the restart. And RAMs on 1.35 forget directly from 1.54volt up to 17.4 the memory must be high 39.38x 4x12 RAM gpu =8x16 lvk normal until you get only up to 7/24 r2 = 43x is overvoltng must board on 185x37m36x x12 on x 39.38.5 187volz 1.2pll from 85 degrees made bucket at speech 19/ 32x max tnlimings 43x6x450ll 12 from 214 +- 2.4volt RAM. Timing 6. 4.!voktn. And the boats. Went inmax bye 197

[@celeriumlerium8266]

Okay I have a 13900k and Z790 Aorus Master, both are about 1 year old and I have been running it with a 320 watt power limit (aside from the first week or so) since that's what my Cooler Master ML360 illusion is capable of cooling. I do not have any issues with crashing or stability that I know of. What should I do? Do I set these new limits, wait for a new BIOS or just leave it as is? The hardest thing I throw at it would be VR + FFB simulator wheel, depending on the game I have seen up to 90% CPU usage (assetto corsa with traffic mods) but aside from that and MSFS2020 I'm never using all the cores anyway.

[@razoo911]

usually crashing start with worse cpu binning then the mixed binning then the high binning cpu, and from what i understood it's not a temperature problem but a fast silicon degradation problem, so to avoid any future problem use intel base setting

[@user-tc4tz8ww1z]

Buildzoid, I have a very important question for you. For those of us that have carefully undervolted our chips, given the need to equalize the DC and AC LL values.... Should we just reset the CMOS and start all over? Or can we just enable IACEP with the LL settings as they are?

[@veganporkkebab389]

noob question but is it beneficial for performance to pick 2x24 DDR5 8000 CL38 on Z790 and 14700k over 6000 CL30. Or is there any limitations.

[@Gindi4711]

I think the way Intel is doing works perfectly fine with OEMs: They are selling them CPUs, shipping the needed documentation and tools and then the OEM validates their board and depending on the customers plans more or less headroom. And if the OEM decides to throw all recommendations out of the window they will be facing tons of warranty claims from customers so they will probably not do that in doubt configure everything to be on the safe side. Now in the DIY market this does not work because the average user is unable to check all that stuff. The mainboard vendors are doing all that stupidnstuff to be first in reviews or be able to cheap out of components at the same performance. And then if the system is not stable they are telling their customer that the CPU is defective and if the board isnrunning fine with another CPU everyone is going to believe them. I think that for the class of DIY boards (or smaller system builders that dont make their own boards) Intel needs to stop giving "recommendations" and give "specs" only.

[@rei3512]

so for the gigabyte motherboard the ICCMAX unlimited bit is disable when i put 400A on the ICCMAX value ?

[@deadlymecury]

yes, it is unlimited if current limit is on auto.

[@kevinzhu5591]

Yep, when I set a limit on my Gigabyte Z690 UD AX DDR4 it enforces the limit, for example, if I set it to 307A, the CPU only consumes 210W at Cinebench R23 at motherboard default settings.

[@evan-du3vk]

Have question. I have 13900k in asus motherboard running p1 and p2 set to 253 watts and 400 amps. I don't have crushes or any other problem. I'm running 80 degrees in cine and 60 in games. Should I bee worried and try to change it?? Or can I go with those settings??

[@Keaweahe]

Pretty sure its fine as long as vcore isn't too high.

[@Rinnzlerr]

i know it’s amd but what power plan do you recommend for ryzen cpus ?

[@RuiN4265]

Buildzoid if i apply 12/12 for loadline my voltage shoots up to 1.705v and i get a warning from my bios immediately. So obviously i didnt keep those settings, with no values for loadline and CEP enabled my multicore score on r23 is 22k. What am i doing wrong? 14900KS delidded w/liquid metal Z790 Apex Encore

[@timothygibney159]

Downgrade your bios. Asus is buggy AF. My hero version of your board got weird where it would exhibit wrong GHz and strange cb scores until I went back and manually entered my low ac lA and under volting settings and it’s now within limits. Asus vrms and llc settings are so out of spec and wack that they can’t handle Intel base lines

[@RuiN4265]

@@timothygibney159 been thinking about doing this anyways. Going back to 0801, I had the best luck with ram stability on this bios anyways. If anyone has an encore and has suggestions for a good bios version for ram stability i'm all ears and would appreciate it!

[@Mirage_Unknown]

Going forward, I have a feeling motherboard manufacturers might still cheat ac dc loadline by intentionally setting dc loadline lower than where its mostly accurate so they can set ac loadline equal to a lower than would be stock dc loadline.

[@sirmonkey1985]

Once this all blows over in the media definitely. Bet next platform Intel follows what AMD does with their bios so only the user can modify settings.

[@davidsmith4186]

Any video on setting up 13900K with overclocking settings? While not being able to boil water at the same time! Asus Maximus Z790 motherboard.

[@ActuallyHardcoreOverclocking]

in my exp it's not worth overclocking the i9s because there's no headroom.

[@NINEWALKING]

What Bildzoid said. All overclocking, accept for serious sub-zero cooling will not yield more performance. That's true from like i9 9900K at least. Disregarding on how good is your cooling system, at room temperature it will not be able to transfer the heat fast enough from the silicone. Surface is to small for that power level to be cooled effective. Plus in stock form IHS and "soldering" and thermal compound just lower the efficiency of the thermal energy transfer. In lot if cases even if it gets stable at higher speed it doesn't guarantee better results. It might clock stretch even. The manual Overclocking with good gains is practically dead. All you can do to "gain" the performance or better said reach the promised performance is to provide good cooling and setup bios so that you gain the performance by making all those dynamic boosts actually work and keep it at the acceptable temperature. That's enough of the challenge.

[@longboardfella5306]

I’m a newby - I think there should be an option for starting with a guaranteed long life and stability setting - that you can then VARY from to up your performance - at your discretion. My Gigabyte Z790 with i914900K is currently NOT stable using Optimise unless I use the beta “Intel Baseline” BIOS - which I am worried about for long term CPU survival and which is now withdrawn. I will watch and rewatch this video until I understand how to safely detune to something with long term stability and safety. Next time AMD!

[@TheMongolHordes]

I thought that the voltage is set in the VID table and temperature has no bearing on it.

[@cemsengul16]

Just checked my 14900K that is on a Z790 Dark Hero motherboard AC LL value with Hwinfo like your video shows. Mine says 0.550/1.100 yikes. Do you know how I could correct this on an Asus bios for a Z790 Dark Hero?

[@xBINARYGODx]

yes, change the relative settings - there is no way in hell the ASUS bios doesn't have settings for the LLC and AC LL and DC LL if Gigabyte has it.

[@IlCode85]

I have a similar situation with my Asus Prime Z790P, mine says 0.4/1.1. The problem is that we have no clue on what are the correct numbers for our motherboards. I guess we just have to wait for a new BIOS that sets these number properly, hoping that Asus will do things the right way this time.

[@cemsengul16]

@@IlCode85 I am hoping Intel forces them to issue a correct bios once and for all but I wouldn't hold my breath since Intel is chasing synthetic benchmark numbers.

[@IlCode85]

@@cemsengul16 My fear is that they will issue a correct bios but then since they cannot push the CPU as far as they have been doing we will suddenly have CPUs that perform much worse with the new bios.

[@tamamoko9725]

This is crazy.. 14900k and z790 Dark Hero here but my AC/DC LL is 0.55/0.49.. Btw have you disabled Asus MCE and use Intel turbo boost instead (just wondering if this caused the difference)?

[@nenedab9999]

On MSI they just ask you what cooler you're running and base PL1 from there, I don't even see any AC or DC load line in the setting, CEP and TVB is default off

[@NINEWALKING]

I have them on my MSI board. Though you need to dig them up. First, go to advanced bios setup. Then, to overclocking, and then you still do not see it until you go in all VRM settings. Just open everything you can, and it should be there. Sure, it can depend on board to board, but it should be there. I have Tomahawk Z790 and it's there.

[@nenedab9999]

@@NINEWALKING I don't see it in my z690 DDR4

[@Mike500]

@@nenedab9999 On MSI it's called "CPU Lite Load". And if you switch it from "Normal" to "Advanced" you can set different values for AC and DC.

[@OfficialWhip]

So if I have a 13900k and a z790 e gaming WiFi with MCE off and don’t do anything over 150w what’s should I do?

[@volodumurkalunyak4651]

I think CEP will take care of wrong LL settings. CEP goes absolutely crazy on clock streching IF CPU thinks is doesn't get enought voltage. Buildzoid, your fix is probably absolutely right (set DC LL to match CPU Package Power and set AC LL=DC LL and set CEP ON).

[@deadlymecury]

Unfortunately CEP doesn't prevent applications crashes. It kinda can be used as indicator: if your performance is crippled - CEP is kicking in and something is wrong with the voltage.

[@volodumurkalunyak4651]

@@deadlymecury but CEP really crippling performance is a very bad sign for moutherboard manufacturer. It means they need to fix their moutherboards. Moutherboard manufacturers desire to massage the AC_LL (and probably DC_LL) into bencmark enhancing way is completely crippled with CEP on and now they have to tune those 2 to reflect real VRM capabilities and settings

[@deadlymecury]

@@volodumurkalunyak4651 you mean that if Intel will enforce CEP they will have to make proper LLC? Hm, maybe. I meant that enabling CEP doesn't solve the stability issues itself, apps are still crashing.

[@volodumurkalunyak4651]

@@deadlymecury yes. Moutherboard manufacturers dont tune VRM in the wrong way or tune AC_LL or DC_LL setting wrong just to be pretty. That is mostly done to win bencmarks. Intel enforcing CEP doesnt allow that anymore so they are forced into correct VRM tuning with correct AC_LL and DC_LL to win benchmarks.

[@deadlymecury]

@@volodumurkalunyak4651 what if they cheat again and simply shove more voltage and then blame your cooler?

[@shaneeslick]

G'day Buildzoid, 🤔I guess since he is not touring anymore Cliff Williams could go around checking everyone's AC⚡DC Baselines

[@Neverwinterx]

I wonder if the largest cause of all that instability and "out of video memory" errors is the max current (iccmax) that the motherboards are pushing so high. I'm sure those other settings also influence it, but the most dangerous one seems the current. It's the only one in the chart where Intel clearly felt the need to put an additional explicit "Never exceed 400A". I think for the same reason they only pushed the max power higher for the KS compared to the K (considering they're the same silicon, just binned better), they knew going with higher current would be dangerous. And yet lots of motherboards push it to more than 500A. Also the max current / icc is one of the main things that Asus changed in their recent bios update to deal with the instability/crashing complaints. I have an MSI motherboard and for once it seems like MSI is the reasonable one. For sure it isn't perfect and also labeled ambiguous in their bios sometimes, but at least it isn't as bad as with Asus (and gigabyte also apparently?), it's fairly easy to set it to reasonable settings even without bios updates.

[@Chevalier_knight]

I have a i13900k on a z790p a d when turn enforce limits on it makes my lp1/2 253w does that mean i should use the extreme collem int he blue chart and set it to iccmax 400a instead of 307a

[@GodKitty677]

For my 10900k the AC/DC Loadline for the overclock is 1.1/1.1mOhm.

[@ActuallyHardcoreOverclocking]

10th gen's LL settings work very differently compared to LGA1700 AFAIK

[@GodKitty677]

@@ActuallyHardcoreOverclocking Yeah but both the DC/AC are the same. 10th gen S-Processor Line DT - 8/10-Core (65W, 80W XeonW, 125W, XeonW 125W) 1.1mOhm maxiumum. 13/14th gen S-Processor Line S/S Refresh - Processor Line (65W,125W) maximum 1.1 mOhms.

[@IlCode85]

I have a Asus Prime Z790P and the new BIOS comes with the Intel Performance and Extreme profiles but they both set AC and DC loadline settings at 0.5 and 1.0, so with CEP enabled the Cinebench score goes from around 38.000 to around 27.000. Disabling CEP brings the score back to more than 37.000. So this is still not what Intel recommends.

[@ShermSpinner]

maybe i'm too nooby when it comes to this kinda in-depth stuff so I'm missing something, but... arent we basically back where we started now? Intel has given a new, albeit somewhat less vague spec, but in the name notion again essentially says "oh but you dont need to adhere to this if your board is good" and we all know how mb vendors market their boards *looking at you, trash-tier B660 boards that are marktet like they have god tier VRMs despite struggling to run a 65w CPU* So... what has actually changed? Besides Intel making up a retroactive, theoretical spec to deny warrantiy because "just run baseline LOL"

[@Chevalier_knight]

I have a asus z790p is that a bad board? I using a i13900k 4090 and 64gbs ram.

[@Jarrahfe]

I still can't fix my vcore and the ac/dc loadline even with latest bios for intel fix

[@ChrisJackson-js8rd]

pmbus should allow the board manufacturers to reprogram the vrm controllers via bios IF they implement that functionality - pmbus implementations seem to vary a lot, and sometimes it's not much more than a bus to write error codes to also the way they bin cpu's IS incredibly weird for instance the binning performance characteristics of the dies taken from the vicinity of a given die can make valid predictions about the long term performance characteristics of that die over time that CANNOT be determined from testing that die at time of manufacture. and the clearest example of this is CPUs randomly dying before the end of the warranty period, which is best predicted by the failure rate in that region of the wafer at time of manufacture ultimately i dont think intel has very much ability to dictate bios settings to manufacturers because there's not really an effective mechanism to punish manufacturers for not listening to intel. the only thing that polices anything is the risk of rma.

[@megamix]

When are BIOS updates with these new default settings coming out so I don't have to set everything myself and get something wrong?

[@onlinegaming4826]

Never xD

[@NINEWALKING]

Actually it is even worse now. When they will give us enough information so we can set it right without losing performance we paid dearly? Already setting this CPU'S to be happy thermally is the balance act if it's own even with the super duper cooling system. Even then allowing it free rains will start pulling 50 to 100 Watt more and will not improve the results at all. So, even with great cooling system you need to optimize it or it will heat up un necessary and pull un necessary power.

[@YGadgETech]

@@NINEWALKINGI really don’t know who to listen to right now it’s a one big mess Intel did here , should I listen to their statement defaults settings?

[@NINEWALKING]

@@YGadgETech I would be happiest man alive if I would have one liner that could help anyone in this situation to be honest. The complexity is in the fact that even with the same mainboard with the same bios we all have different CPU that has same name. You have 3 variables in one mathematical problem plus it is load dependent. Not only how much load there is but what type of load it is as well. At this moment it seems that most of so called Bios versions with the Intel default LL values will solve the crashes. But they do not do same job at the same level. Some just reduce the performance way too much. My honest advice is that if you have made the setup that is thermally sound (you are not running to hot in real life use scenario) and your games and applications are running fine just forget that it can't run Cinebech 15 or some game that you do not own or play/want to play. I you have real issue only thing is to play with setting until you find your own compromise between stability you need and the performance. Even that is a huge challenge now. Anyhow those both solutions are for the time until the Intel and mainboard companies come with the better solution. Good luck.

[@XiaOmegaX]

Intel should send you a VRTT test tool...

[@pdub420]

So, what is actually happening when the AC is set lower than the DC? I have my AC at 0.3 and I remember researching this before and reading they were supposed to stay the same. I was getting WHEA errors at 0.3, so I set DC to auto and now hwinfo64 reports the DC at 0.9, and I am not getting WHEA errors...yet.

[@Augenhose]

If AC is too low, you reduce your operating voltage Vcore. You are essentially undervolting the CPU to a point where it starts to clock stretch, leading to failures due to insufficient voltages. At 0.9 AC you allow higher Vcore voltages increasing stability, compared to 0.3 AC. However, if you have a descent silicon quality, the truth lies somewhere in between, probably 0.4 to 0.6 AC. Since 0.9 AC is likely providing overvoltage and senselessly increasing thermals and thus, wasting power.

[@pdub420]

@@Augenhose I'm not at 0.9 AC, I changed the DC to 0.9 and am suddenly stable then when I was at 0.3 for both AC and DC.

[@photobeast]

I was wondering if somebody could help me. I have the Intel 9 13k chipset. I put it on the Asus strix Z690E motherboard. I was wondering if it is normal when doing cinebench r23 to have one core hotter than the others? During my stress test most of my pee course will be about 85 degrees Celsius but number 5 will reach 99 and 100 ° C and number 7 will hit 98 degrees Celsius. All the other ones are around 85 to 90° at their Max these two run hotter than the rest of the cores. I try to undervolt them but I'm not getting any results.. is this normal by chance

[@MrBillzebud]

Yes its normal.

[@AamirM87]

You might have bad contact with the cooler.

[@NINEWALKING]

There is literally 0 processors that have the same temperature on all cores. It's impossible at so manny levels. Starting from the load it self. Operating System thread handler/scheduler can't load them all the same plus when easier part is rendered on one core it has less load. It's dynamic thing. Plus silicone might not be as thick over the all cores. Then welding to the IHS is not as thick over the cores. Lastly your cooler might not have same contact with all surface of the IHS. So some difference is perfectly normal. But if the difference, during synthetic load, is to big like 12 degrees or more, than you need to check were it is. Best case scenario is bad mount of the cooler. Worst case scenario, Lemmon CPU with internal issue.

[@photobeast]

I kind of was wondering that myself due to the fact that I actually just redid the thermal paste about 2 months ago when I was cleaning out my system and changing the ram out. I was getting blue screen of death constantly and attributed it to the ram even though the meme test said it wasn't all the computer guys I know said it's most likely the culprit. After replacing the RAM and getting blue screen of death again I did every command prompt without any good results to the point of reinstalling Windows for like the fourth time. It was at this point that I started doing KZbin research and ran across the Aces and Intel specifications are different ​@@AamirM87

[@AamirM87]

@@NINEWALKING You're right but the delta between cores should be within 5C and NOT 15C as OP is having (85C vs 100C).

[@VonDelano]

Sat here with AC_LL 0.550 and DC_LL 1.0 what are you doin ASUS

[@CaveWaverider]

My ASUS Z690-A board has this set to 0.550/1.1 on my 14900K when the SVID Behavior is set to Auto.

[@elita2cents]

Gigabyte's default bios settings have the C-states disabled, always. I turn those back on, every time.

[@Ryxxi_makes]

Can you make a guilde on how to change these specific settings on Asus please.

[@NINEWALKING]

He did it. At least with the information he had like week ago. You can use it still and check values with this video. There is no better solution sadly because Intel them self is not able to solve it. It's case to case thing anyway.

[@deadlymecury]

yeah, and the second problem is that we have no idea what are the correct values on asus or any motherboard. so still waiting that Intel report on their findings.

[@IIHydraII]

I know you don't have a tool to measure and test yourself, but do you have any suggestions for AC/DC LL and LLC for Asus? Does .9 .9 LLC 3 sound reasonable?

[@Augenhose]

What Mainboard model, and what BIOS version are we talking about?

[@IIHydraII]

@@Augenhose Z790 Strix E & latest bios (2202 I think it's called)

[@Augenhose]

@@IIHydraII OK do you have Cinebench R23 and HWinfo64 installed? Because you would need to make a single run using your 0.9 / 0.9 ohms AC / DC settings + Intel limits (pl1=pl2=253W, Current Max=307A), make sure MCE is off (Disabled "Enforce All limits"). Then note Vcore, CoreVIDs, Vrout and pakage power average readings over the CB R23 run. If they all match, and Vcore under full 253W load is not too high (not above 1.3 V), it is fine to have your current settings. Also note highest temperatures of p-cores down. If they are close to thermal limit (above 95C), you are pushing too much voltage over Vcore (AC_LL is too high).

[@IIHydraII]

@@Augenhose thanks for the info, I can check everything tomorrow. :)

[@Augenhose]

@@IIHydraII Sure thing.

[@0Wayland]

Waiting for board partners to release Intel baseline profile (the actual one)... 🙄

[@NINEWALKING]

Will be there in 6 months or so. Plus performance will never be the same.

[@R32639]

My Asus does 0.02/1.0 at auto with P-Cores only and sync all cores 55 (13900K)

[@Ohlukei]

the MSI Pro Z790-A WIFI board by default is totally set up wrong. Even on Auto settings for PL1 and PL2 it set 4096 W and even worse 512 A for ICCMax.

[@eliotrulez]

So, the engineers of ASUS & Gigabyte don't understand the Intel spec. Thanks for explaining :)

[@betag24cn]

no, they do, but to sell more motherboards they will push harder and harder so people believe their products are better and the competition sucks it is the same thing that amd had not long ago woth burned cpus because motherboard pushed too much volts intel is lucky no units ended in flames like amd did the blame is on both sides, motherboard manufacturers do not respect their own products or the customers, cpu manufacturers do not protect cpus in the way they should

[@eliotrulez]

@@betag24cnfrom the point of liability that was a very bad idea. If there boards degraded the CPUs, they are liabale for damages. Intel does not need to replace the CPUs, but ASUS & Gigabyte have to. (And we know how good ASUS warranty service is).

[@betag24cn]

@@eliotrulez im not 100% sure abot that, sounds logic but since intel makes the chipset and puts all limits, they are responsible too, if is true that intel cpus used to have security measures on the cpu and were removed, it is shared fault

[@eliotrulez]

@@betag24cnThey are responsible because they let the vendors do what they wanted. But it can lead to a class action especially in the US where the vendors will have to pay for damages.

[@betag24cn]

@@eliotrulez yes, i would not remove that blame, what i wanted to say is that is a shared blame, in intel casr where they control everything, is their fault first, but they are in panic mode, the fumes of amd glue made tyem inject all the watts possible and here we are, trying to blame someone and with no real solution when the amd burned cpus happened last year, everybody blamed amd, but i think it wad tye same thing, in that case it was more on the motherboard manufacturers but shared

[@ATestamentToTech]

I think you have the best thumb up thumb down ratio on YT

[@liaminwales]

I wonder if intel encouraged them to juice the CPU or if there just not checking consumer for problems?

[@sirmonkey1985]

Of course, this is all Intel trying to save face by blaming board partners.

[@cemsengul16]

This is exactly what happened. They could have forced motherboard partners to follow their official healthy spec but they needed to beat AMD in Cinebench scores at all costs.

[@tommihommi1]

well in cinebench the e cores make them win, games is what the crazy power is for ​@@cemsengul16

[@deadlymecury]

This whole situation with crashes under gaming load, 0xc000005 errors in every possible app etc are known for more than a year. But they started to move only when Nvidia (not customers on their own fucking forum!!!) told publicly "go to Intel". And first thing they did is they say "oh it's because of power limits" - while gaming loads are below that. Or single core loads below that. Or regular windows apps crashing in the background below that.

[@cemsengul16]

@@deadlymecury I get some kind of error when I click shut down on Windows but I don't know if it's related.

[@ncohafmuta]

I think all mobos should use the Performance profile out of the box and the KS one should be 150/253W. Otherwise, now CPU performance is dictated by the mobo. A $200 mobo will now perform worse than a $600 one, even if the $200 one is capable. So now CPU review scores are going to be based on the mobo of the reviewer and a consumer will have to know that. Are the mobo mfgrs going to tell us which mobos are using which profiles? They don't tell us what power limits they use now in specs/manuals. Not to mention now it will give them another excuse to upsell you on a higher-priced mobo. Hey, our Apex board has better CPU performance than our Tuf board, you should buy that instead! Secondly, now a normal person is expected to buy a CPU cooler not based on Intel's ARK advertised power limits, but by the mobo they buy. That's insanity. That would be like buying a 6000 MT/s memory kit and the mobo saying, i'm a great mobo for memory, not only am i not going to use JEDEC, nor the XMP advertised number, i'm going to set 6400 MT/s. I have no problem with mobos having a popup when first entering into the BIOS that says "Your CPU is set to Intel's default power limits. Do you want to increase these to X to improve performance? WARNING: This may affect your thermals, power usage and stability" YES/NO. Simple. Why is this so hard?

[@deepSilent0]

I went with Intel since everyone was saying that’s the “no tweaking, works out of the box” option… Now I have a worse product that doesn’t even work out of the box. Reviewers and “experts” also need to be held accountable.

[@janbenes3165]

Which ones? Most reviewers were in agreement for a while that in most cases AMD is a better option for the time being.

[@punishthemeatpocket]

To be fair a lot of them lack basic knowledge of pc hardware, they just spout whatever intel tells them and throw up cinebench charts. Power users tend to figure these things out first.

[@betag24cn]

to be fair, the problem was not happening since day one, so no one saw it untill the out of ram happened not long ago

[@lukilapaj]

Can we set AC/DC Load Line in Gigabyte BIOS manualy?

[@Synth3ticMind]

Yes

[@tamamoko9725]

Yes but the problem is that we don't know the correct value to set to

[@dolareta8257]

so ASUS 2024/04/19 bios with "Intel Baseline Profile option" is garbage and doesnt reflect this guidelines? my 14900k shat itself like 2 months ago, sitting now on 13700k waiting for safe working bios

[@devans83]

Just shipped off my 14900K for RMA today. Popped in a 13700K I have to do a build for my daughter. I have ASUS' latest bios update, and I am beyond confused as to what settings I need (I don't overclock). The new ASUS "Intel Baseline Profile" easily puts performance at a 10% loss. I don't understand why Intel isn't putting more of a priority on this debacle.

[@dolareta8257]

@@devans83 my 13700k btw just died as well, same thing, played new UE5 game for like 10 hours, then crashes started, then they became more frequent and now i cant even start a game, mobo asus apex encore fuck intel and fuck asus

[@guineapigofdoom]

I found that setting the following on my Gigabyte Z790 Aorus Elite Ax helps CPU speed with CEP turned on : Tweaker>Advanced voltage settings>CPU/VRM Settings>CPU Internal AC/DC Load line: Setting: Turbo = AC 70 / DC 90 (This sets the AC LL to 70. It leaves the DC LL at 90) My Gigabyte LLC Settings are set to: Standard P1 setting is 125 P2 setting is 253 CPU is an i7 13700KF Cinebench score is no longer cut in half. It is back to where it was with CEP turned off.

[@kry1429]

i added the intel base profile 'performace' asus bios update from 1 month ago 14900ks with sp score of 99 cooler points 178 STILL THERMAL THROTTLES 90 DEGREES ON CINEBENCH AND GAMING BRUH

[@thomaspvh]

Does anyone know if 12th gen has these issues? Why did it get wonky in only 1 gen when its pretty much the same CPU no?

[@NINEWALKING]

I have had 0 issues with 12900K. At low loads especially it was drawing less power. Watching KZbin was like 9 to 12 Watt. Now it's 20-ish. Full load within the thermal limit was also way lower. Unlimited, with my cooling setup I have seen 404 Watt maximum draw in Prime95 small FFT's 24 threads. That's a lot of power. I have limited long term paper to 330 Watt but that apparently let's it still draw 360-ish Watt. No benchmark gains from letting it draw 400 Watt. 14900K is a strange CPU.

[@deadlymecury]

Same, had no issues with 12900k. And it's been more than a year I am facing application crashes and os issues after switching to 13th and then to 14th gen "to fix issues with 13th gen". Kinda yeah, it's "the same cpu" and somehow using "the same cpu" they pulled frequencies up to 6.0 ghz for 13th gen and up to 6.2ghz for 14th gen. Try to overclock your 12900k to that and I guess you will get that "13th/14th gen experience" for free with ability to revert back lol.

[@NINEWALKING]

@@deadlymecury actually it is not the same CPU :( Those additional E Cores and additional power draw and cache and speed and what ever else makes it completely different beast. If you thought 12900K is a fire breathing monster wait until you use 14900K.

[@deadlymecury]

@@NINEWALKING I know. I used 12900k in a prebuilt (and before it was 9900k), then I built my system with 13900ks, when I replaced it with 14900ks to solve stability issues - just to have more issues. It's a "completely different beast" based on the same architecture, same process etc and overstressed to the limit that it failed completely. Also in case of gaming I don't quite see huge difference between 12900k and 14900ks. E-cores especially do nothing for that.

[@andrey7268]

I didn't test my 12700K on Asus Z690 Hero at full stock, but with E-cores disabled, the default auto AC LL is too low to be stable in Prime95. As far as I'm concerned, this stability issue applies to 12th gen as well.

[@jayhsyn]

My asus 790 dark hero sets the amperage to 511A as default out of the box, here it says 400A on extreme. Haha

[@Chevalier_knight]

Can someone tell me like im a 5 year old what am i meant to set a dc and ac for a z790 p with a i13900k and 4090? I have no clue wh at this guys talking about 90% of the time do i just keep cep off and auto for dc and ac and just change my lp1/2 and current to the same ammount.

[@onlinegaming4826]

i us Gigabyte 790 elite ax Intel Recomended Spec Enhanced PCore 54 ECore 43 XMP1 Vcore auto W253 A400 Loadline Medium C State Enable (ALL) EIST Enable TVB Enable CEP enable Not the best scores 36k or 37 but safe

[@andrey7268]

20:02 I think you misunderstood this note. "VR and BIOS Load Line values must match" means literally the way the voltage regulator is configured (according to the LLC mode set in the BIOS) must match the DC LL value that is programmed in the CPU on boot. This, and the previous note, basically give the motherboard vendors the permission to provide multiple LLC modes, according to their MB designs, but require that the AC/DC LL values programmed in the CPU match those modes upon boot. When you see the 0.4/0.9 mOhms LLs in HWINFO, you're seeing the values that the BIOS set for the LLC mode that is configured there. Change the LLC mode and leave the DC LL auto and you will see that DC LL (and likely AC LL) will change. Regarding AC LL, I'm not sure how MB vendors choose that value, but I suspect that's where all those "AI overclocking" and "SP rating" stuff come into play. Basically, every vendor is free to guess "the right" value for any specific CPU based on whatever logic they can come up with. For example, you might guess AC LL must be higher for an i7 than for an i9 (because of worse silicon quality), or for a CPU with a higher V/F curve (compared to whatever reference the MB vendor has conducted via testing), or whatever. I'm sure, there is a fair amount of CPU testing being done by MB engineers to come up with reasonable reference values. But this fundamentally doesn't guarantee CPU stability for any given CPU. And Intel isn't helping in this regard, other than by saying "it'll work if AC LL is equal to DC LL". Which is fine as the criteria for differentiating between a functional and faulty CPU for the purpose of the warranty, but doesn't provide guidance for choosing better (i.e. lower) AC LL values. Because running with AC LL == DC LL will result in high operating voltages, way higher than necessary on motherboards with decent VRs. What Intel needs to do is to define a set of specs where they basically define AC LL, DC LL and max allowed voltage undershoot and duration for a number of LLC modes, at which CPUs are supposed to function. Basically, they need to do themselves what they are asking to do the MB vendors now, and put the results in a spec.

[@deidian635]

Intel can't tell a board maker what AC_LL is the right one: what they do is setting a max value which essentially sets a minimum bar for the power plane in the board. AC_LL is just the board informing the CPU how much impedance there is in the socket + VRM(physically the board) so the CPU can calculate the final SVID request for the VRM: more impedance the CPU has to ask higher SVID so the Vcore actually meets the fused VID point the CPU knows it needs.. It can certainly be misused to under volt or over volt the CPU if the board lies to the CPU about that impedance: but that's on the liar, not the CPU.

[@Shonkuk1]

DC Load line must match AC load line is a translation error or document error DC Load Line must match Load Line (LLC)

[@ActuallyHardcoreOverclocking]

DC load line must match VRM load line so that the power calculations are correct. AC_LL must match DC_LL so that the CPU requests the correct voltage.

[@Shonkuk1]

@@ActuallyHardcoreOverclocking DC Load Line must match VRM load line (LLC) someone has presumed this is AC_LL its a mistake/wrong/typo/whatever DC load line has always matched LLC Asus sync's this so are correct at 1.1 gigabyte has always locked in a wrong value of 0.9 instead of 1.1 no matter how llc is set, if you want to confirm this go change llc on an asus z690/z790 board and watch the DC_LL change to match it

[@deadlymecury]

​@@Shonkuk11.1 is not the correct value but maximum allowed variable. Correct value depends on PCB layout, VRM components, socket etc. There is a chance that 0.9 is correct LLC for gigabyte and that's why they use it every where in their vcore LLC profiles changing only AC: when you change profile from "power efficient" - "performance" - "turbo" - "extreme" - "superpuperultradouble extreme" - it changes from 50/90 to 60/90, 70/90 etc. I would take it as it should be 90 and value below is undervolt. And they use 90 for DC just to get correct power reading. But only gigabyte can tell if it's correct assumption or not and "this little maneuver gonna cost us 51 years". Still waiting for 14900ks bios lol.

[@Shonkuk1]

@@deadlymecury 1.1 is the correct value for llc being standard/normal, You can lead a horse to water but you can't make him drink

[@deadlymecury]

@@Shonkuk1 just rewatch video / reread that part of the specs where they tell about tool to measure impedance of the board.

[@raq5558]

where can i find intels recommended default limits?

[@user-tc4tz8ww1z]

At 41:26 in the video. Intel removed the full spec sheet from their website

[@raq5558]

@@user-tc4tz8ww1z why did they remove it ? Smh

[@user-tc4tz8ww1z]

@@raq5558 I don't know, someone provided me a link but when you go there it says "oops cant find file"

[@raq5558]

@@user-tc4tz8ww1z wow. Thank u man

[@Tech-Dino]

Can someone translate this table to Asus Bios named settings?

[@ashryver3605]

These things should never have been released at 5600mhz (13900ks), etc Basically all of these things are 100mhz higher than they should be, similar to how it 12th gen mobo manufacturers actually lowered Intel's intended 5.1 down to 4.9 cuz they wanted to be safe and assume It would be hard to be stable at consumer default. You can observe it on any 14900k/s and 13900ks, if you downclock it just by 100mhz on the P-Cores, at stock, the system is basically stable at everything, the window of stability suddenly becomes enormous.

[@xBINARYGODx]

sorry to burst your bubble - but most of them do 5600 just fine - and that was already lowered from 5700 to begin with (which was the problem). Also - downclocking almost ANY cpu 100Mhx from default can drastically increase stability or rather, let you dump the voltage down more than you might at first assume, etc - blah blah blah.

[@deadlymecury]

@@xBINARYGODx it's quite funny to read your rant while using 14900ks that is still not stable under 6.0/5.8 ghz and still having app crashes. Guess it's time to test 5.9/5.7. Or crank LLC from 65 to 70.

[@Mangokr]

With the new intel baseline guidance (125W pl1 / 188w pl2), there is a chance to take them to the court? Since is not “as advertised” anymore.

[@ActuallyHardcoreOverclocking]

where in the intel chart do you see 188W?

[@Mangokr]

@@ActuallyHardcoreOverclocking in videocardz there is an article about it videocardz.com/newz/intel-wants-default-settings-with-pl1-pl2-at-125w-188w-to-be-implemented-by-motherboard-vendors-by-the-end-of-may

[@blgDemon]

@@Mangokr wait for the official intel statement. All that is just rumours at this point

[@cemsengul16]

@@ActuallyHardcoreOverclocking Well technically the PL1 and PL2 on Asus is set to 253 on the new Baseline profile however if you run Cinebench you will not hit 253 watts.

[@ActuallyHardcoreOverclocking]

@@cemsengul16 I think the ASUS baseline sets ICCMAX to 280A. Which is why it won't hit PL1 or PL2

[@Seth22087]

i am getting more and more of feeling like Intel spec sheet ends up as toilet paper and their tool as cup holder... so some employee can punch in random numbers based on what they think will win benchmark and if they notice those numbers aren't quite right they patch them with BIOS update. And there I thought that worst thing that happened to my previous CPU, Ryzen 2600X was me, but apparently it was Gigabyte. :-D P.S. Also if you aren't sure if things are set correctly, you are probably more often right if you assume things aren't set correctly. :-D Like, there is Murphy's Law and then there is Intel board partners level or two worse than that. Also Intel shouldn't be asking at this point. They should be more in lines of come in and say "Listen here you little s***, here is what you are going to do...". I think asking was more for time when Intel generations were still single numbers. Or in other words, go nVidia on them. No board partner ever goes against what nVidia wants. They rather quit business entirely than go against what nVidia says.

[@bastordd]

I have a MSI BOARD on my 13900K... MY IA Loadline AC/DC: 0.010/0.010 How the fk is this possible?

[@YGadgETech]

So much stress it’s seems like a total disaster what’s going on here , seriously I don’t know who to listen to =/ Should I return my parts ? Didn’t opened it just yet but seriously I want to give it a try as I have hope , 14900KS + apex encore and 8400mhz g skill I really be glad if someone could help me out to stable it and make it work as it should

[@NINEWALKING]

Well who ever gives you one line answer is wrong sadly. No one can even know what is your draw of luck going to be accept if he pretests the actual parts that you are going too get. KS and K does not guarantee you anything. Plus mainboard and Bios version will play the major role as well. Now to less obvious issue 8400 memory? Well you'll need tons of luck there as well. First two memory slots only highest end mainboard is needed to have any base to reach anything 7800 plus. Then you need to win the silicone lottery for the CPU and it's memory controller and even then it is not that simple. Plus if it is all working fine it still doesn't guarantee you much performance gain over something like 7200 cl34 that is going too work with most of the new mainboards and most of the CPU's. I went even with 6400 CL32. The main thing is SK Hynix and A die memory chips and you can play with speeds yourself. One more sad thing. I have both EK Velocity 2 full nickel 1700 and Apex Core1. Despite everyone and their aunts saying Alphacool water block is better I get 1000x better mount with my EK block and way better temperatures. So I use my Alphacool block on my AMD system now. So it seems that it is a lottery as well in the IHS of the CPU. If it is less curved you will profit from the flatter block. So my EK matches better my 14900K IHS. We live in the sad times where getting 14900K, RTX4090, Soundblaster AE-9, Intel network card, Studio monitors etc. etc. etc. does not even guarantee you will be able to play even 5 years old game :(

[@YGadgETech]

@@NINEWALKING I completely agree that this is Silicon Lottory, the board I have is the APEX ENCORE which is the best for OC and especially for high RAM what is left for me is to know how to correctly configure the processor and the RAM so that the RAM works at this speeds stably

[@NINEWALKING]

@YGadgETech sadly you need the luck with the CPU memory controller as well. Some will not handle anything above 7200. Most good mainboards and CPU'S cap at 7800. Your mainboard is great. Only a few mainboards can handle 8000 plus even with good CPU. Some do not want to work both loading the XMP profile, but you can enter it yourself, and it would work. Good luck mate.

[@RurouTube]

My main issue is that from what I see, you can't realistically get the advertised performance at default settings. Even using extreme profile, you're not going to get the performance that Intel advertised in their own slides. For example, the Cinebench 2024 multicore bench from Intel stated that 14900K is 1.06X faster than 7950X. You look at the extreme profile benchmark (like the one Hardware Unboxed use since they use Asus baseline which is basically extreme profile) and it can only match 7950X. Only the number with basically unlimited power they can get around 1.06X performance vs 7950X. So basically without mainboard vendors going beyond the extreme profile, they can't really match the number Intel put up in the slides. Also some 13900K and 14900K can be unstable at 253W. It might be stable in some workloads, but on a different workload it can crash. People might not notice that their CPU is unstable at extreme profile when they only do gaming thus happily think that their CPU is 100% stable when it might not actually be 100% stable. Also the CPU that is currently crashing after using basically unlimited profile probably will not be stable using extreme profile which is why I'm surprised Intel actually recommending the highest profile that the mainboard can use which is the extreme profile because I doubt it is enough to solve the current instability issue.

[@chiffry830]

Wake up babe! He dropped another one!!! “Kendrick???” BUILDZOID!!!!

[@notreya]

Yeah i'm done. RMA'ing my 13900k. Going to sell it, the motherboard and ram. Bought a 7800x3d, Gigabyte B650 Aero G, Corsair 32GB (2x16GB) 6000MHz CL30 DDR5 EXPO Ive spent like the equivalent of days trying to get this stable. It will be stable for like 2 months and then just fall apart again. I start getting "Render Device Lost" errors in Overwatch. I lose SR and I have almost been banned from competitive play. I just need my shi to function lmao.