Intel processors crashing Unreal engine games (and others)

[coercitiv]

That's really weird. What kind of power usage causes degradation in Raptor Lake CPUs over the long term? Alder Lake was pushing 200W+ and has been for quite awhile without having these problems . . .

We don't know if it's degradation, at least not as the main cause. Buildzoid has a rant video on the subject, and one of the things he talks about is that folks who test multiple top tier K CPUs for binning will repeatedly stumble upon chips that fail to pass stability tests out of the box. That does not mean the chips are bad, it likely means that the board configuration is too aggressive for that specific CPU. It's rare, but it happens. In this case there's definitely no degradation, the CPU is brand new.

My hunch is that one can take such a "bad lemon" K CPU, shove it in a B760 board with locked CEP and conservative stock power/current limits, and it will work just fine long term. It just won't win any benchmarks.

PS: It seems I linked the wrong video above, he already has 2 videos on the subject. The first and longer one is here, but anyone looking for a "shorter" take should check out the newer one.

 

[DAPUNISHER]

Builzoid is ranting on twitter about the issues.

that board is literally on the latest BIOS. Intel is really quick to patch BCLK exploits with microcode updates and gives exactly 0 <insert profane word here > about shipping functional 700USD CPUs.

EDIT: We both posted about Buildzoid seconds apart.

Simul-jam

\

 

[CouncilorIrissa]

So consumers pay their hard-earned money for what is essentially an unstable OC. No wonder Silicon Lottery has closed down. lol

 

[coercitiv]

So the reddit comment that triggered BUildzoid's second video argues that Asus/Gigabyte and to a lesser extent MSI/AsRock undervolt the CPUs using AC Loadline settings to make them perform better in benchmarks. This is done by eating into the safety margins, they are essentially undervolting after they're done overclocking out of the box.

This would indeed also explain why "it feels" like the CPUs are degrading, because any aging effect on the CPU could make it pass the threshold to instability, since the safety margins were sacrificed to justify a price premium on those fancy new board and CPU. It could also be a matter of seasons, you buy the CPU in the winter and it's stable, summer comes with higher temps and... "not enough video memory"!

Here are are some quotes to take into consideration, emphasis mine:

All CPUs degrade over time but that's why their fused VF curves have a sizeable buffer to account for derating.

ASUS/Gigabyte and (from what I observed) to a lesser extent MSI/Asrock Z-boards set an AC load line (AC_LL) & LLC level that undershoots the VF curve significantly under load and eat into this buffer. Because AC_LL undervolts increase in size as CPU amps increase, unlimited power results in more undervolting relative to the VF curve.

Alder Lake and Raptor Lake CPUs also have a clock stretch feature (IA CEP) to account for Vcore undershoots, but due to the aforementioned AC_LL undervolt causing it to trigger constantly under load, it is disabled by default on the Z-series boards I have tried.

Regrading CEP, I should mention it was enabled by default on my MSI Z690 & 12700K combo, so this varies. Let's keep reading:

While the spec says AC_LL == DC_LL == LLC slope, in practice motherboard vendors undervolt by doing AC_LL << DC_LL* == LLC slope. The CPU does not request enough voltage to counter the LLC Vdroop so Vcore falls below the programmed VF curve that would be followed if AC_LL == LLC.

On a freshly reset ASUS Strix Z790-H with MCE off, here are the values I see:

Setting	Value
AC_LL	0.5
DC_LL	1.1
LLC	1.1 (LLC3)
VID 55x idle	1.30V
VID 54-55x CB24	1.26V
VF 54x	1.289V

Uh oh, I'm 30mV short out of the box at 253W for 54x and well over 50mV short of 55x.

ASUS LLC6 would be the closest to Intel spec for AC_LL 0.5, and that is a little better than AC_LL 1.1 + LLC3 in their Baseline profile IMO since it won't boost idle turbo VID to >>1.6V...

*DC_LL is only used by the CPU to compute post-droop VID for power measurement, AC_LL + motherboard LLC are the two Vcore levers that affect Vcore in an unthrottled state.

 

[poke01]

I don’t much about about this Intel issue but hopefully Intel will not have these sorts of issues on such high end CPUs later this year when Arrow Lake launches.

 

[DrMrLordX]

We don't know if it's degradation, at least not as the main cause. Buildzoid has a rant video on the subject, and one of the things he talks about is that folks who test multiple top tier K CPUs for binning will repeatedly stumble upon chips that fail to pass stability tests out of the box. That does not mean the chips are bad, it likely means that the board configuration is too aggressive for that specific CPU. It's rare, but it happens. In this case there's definitely no degradation, the CPU is brand new.

My hunch is that one can take such a "bad lemon" K CPU, shove it in a B760 board with locked CEP and conservative stock power/current limits, and it will work just fine long term. It just won't win any benchmarks.

Well okay, but again Alder Lake didn't have that problem either. Though if it is as you say and the mobo OEMs have started effectively undervolting Intel CPUs @ "stock" while not doing so on Alder Lake CPUs, then that might explain things a bit.

 

[Saylick]

Looks like Intel issued a statement on the situation:

Intel issues its first statement in response to 13/14th Gen Core i9 CPU stability issues - VideoCardz.com

Intel statement on Core i9 stability concerns Intel has issued the first statement on user reported Core i9 processors stability issues. The company has put the blame on motherboard vendors who do not follow Intel specs. There would hardly be any difference in motherboard performance if it...

videocardz.com

 

[igor_kavinski]

Does Ultimate performance power plan in Windows disable safeguards too???

Or is Intel saying that their CPUs are too sensitive to run at the Ultimate level?

 

[Kepler_L2]

Looks like Intel issued a statement on the situation:

Intel issues its first statement in response to 13/14th Gen Core i9 CPU stability issues - VideoCardz.com

Intel statement on Core i9 stability concerns Intel has issued the first statement on user reported Core i9 processors stability issues. The company has put the blame on motherboard vendors who do not follow Intel specs. There would hardly be any difference in motherboard performance if it...

videocardz.com

Intel blaming AIBs while they have known about this for many years, and even used the high-end boards with extreme settings in their 1st party benchmarks

 

[Abwx]

Looks like Intel issued a statement on the situation:

Intel issues its first statement in response to 13/14th Gen Core i9 CPU stability issues - VideoCardz.com

Intel statement on Core i9 stability concerns Intel has issued the first statement on user reported Core i9 processors stability issues. The company has put the blame on motherboard vendors who do not follow Intel specs. There would hardly be any difference in motherboard performance if it...

videocardz.com

In Computerbase review they didnt specify that the 14900KS was set at 253W contrary to other CPUs, and for a reason since it was stock set at unlimited power, as in Intels s recomendation.

https://twitter.com/x/status/1784207862463644010

 

[Mahboi]

Intel blaming AIBs while they have known about this for many years, and even used the high-end boards with extreme settings in their 1st party benchmarks

LeADeRSihP PaIRFormAnCe


Seriously the cracks just expand every single week these days. I can envision the dam breaking after Computex.

 

[csbin]

In Computerbase review they didnt specify that the 14900KS was set at 253W contrary to other CPUs, and for a reason since it was stock set at unlimited power, as in Intels s recomendation.

https://twitter.com/x/status/1784207862463644010

The CB is simply set up for 253W PL2, which is only about 36,000-37,000 according to the ASUS“ intel baseline”.

 

[coercitiv]

Lolz, Intel published a statement to announce they will publish a public statement in May 2024.

 

[Abwx]

The CB is simply set up for 253W PL2, which is only about 36,000-37,000 according to the ASUS“ intel baseline”.

View attachment 97961View attachment 97963

Why post the CB score without the power, and yet it s also measured at Computerbase for Cinebench.?.

Because you know that you are lying, full system power in Cinebench for the stock 14900KS is 522W, there s no other numbers for this CPU, and Computerbase always do tests at stock, they have a separate article for overclocking, btw CPU package power in Prime 95 is 355W at stock, about the same CPU power as in Cinebench.

Intel Core i9-14900KS im Test: Anwendungs-Benchmarks, Leistungsaufnahme & Temperatur

Intel Core i9-14900KS im Test: Anwendungs-Benchmarks, Leistungsaufnahme & Temperatur / Testsystem und Testmethodik für Anwendungen

www.computerbase.de

 

[coercitiv]

Well okay, but again Alder Lake didn't have that problem either. Though if it is as you say and the mobo OEMs have started effectively undervolting Intel CPUs @ "stock" while not doing so on Alder Lake CPUs, then that might explain things a bit.

ADL was in a better spot at launch relative to both prior gen and the competition. We could also argue It was also a new platform, with a learning curve attached for the board makers and a massive workload to get stable systems at launch, not to mention proper XMP support for both DDR4 & DDR5. Heck, Gigabyte was so busy theyr forgot to add NVME support on their launch UEFI version The landscape changed drastically for RPL and then the refresh, both in terms of pressure to deliver perfomance and in terms of time for agressive tunning. That's my impression anyway.

 

[DrMrLordX]

ADL was in a better spot at launch relative to both prior gen and the competition. We could also argue It was also a new platform, with a learning curve attached for the board makers and a massive workload to get stable systems at launch, not to mention proper XMP support for both DDR4 & DDR5. Heck, Gigabyte was so busy theyr forgot to add NVME support on their launch UEFI version The landscape changed drastically for RPL and then the refresh, both in terms of pressure to deliver perfomance and in terms of time for agressive tunning. That's my impression anyway.

No real arguments there, though there were certain lines they (they being Intel and the mobo OEMs) knew they shouldn't cross, but eventually they did it anyway.

full system power in Cinebench for the stock 14900KS is 522W

Is that a typo? 522W? Or did you mean 322W?

 

[igor_kavinski]

Is that a typo? 522W? Or did you mean 322W?

Intel Core i9-14900KS Review - The Last of its Kind

The Intel Core i9-14900KS reaches impressive clock frequencies up to 6.2 GHz, which yields remarkable performance in applications and gaming. However, this comes at the cost of serious power consumption, especially if you go beyond stock, where we saw over 500 W during testing.

www.techpowerup.com

What's a few more watts?

 

[KompuKare]

T

Intel Core i9-14900KS Review - The Last of its Kind

The Intel Core i9-14900KS reaches impressive clock frequencies up to 6.2 GHz, which yields remarkable performance in applications and gaming. However, this comes at the cost of serious power consumption, especially if you go beyond stock, where we saw over 500 W during testing.

www.techpowerup.com

View attachment 98031

What's a few more watts?

TPU do measure at the motherboard's 12V connector:

All power measurements on this page are based on a physical measurement of the voltage, current and power flowing through the 12-pin CPU power connector(s), which makes them "CPU only," not "full system." We're not using the software sensors inside the processor, as these can be quite inaccurate and will vary between manufacturers.

Whereas Computerbase do full system (and later package power using the internal sensors), although using Prime95 not Cinebench R20 so not directly comparable. For Prime95 using the CPU sensor they get:

Intel Core i9-14900KS im Test: Anwendungs-Benchmarks, Leistungsaufnahme & Temperatur

Intel Core i9-14900KS im Test: Anwendungs-Benchmarks, Leistungsaufnahme & Temperatur / Testsystem und Testmethodik für Anwendungen

www.computerbase.de

355W, which is actually more than those big Threadrippers although since Prime95 is not something they benchmark it is hard to see what that means.

We do know that for Cinebench R20, the 7980X is way over twice as efficient (scoring 37,070 using 505W vs the 14900KS scoring 15,629 while using 522W). And CPUs with a similar score to the 14900KS like 7950X use 329W, and a bit behind that a 7950X3D uses 233W (all whole system measurements).

 

[igor_kavinski]

We do know that for Cinebench R20, the 7980X is way over twice as efficient (scoring 37,070 using 505W vs the 14900KS scoring 15,629 while using 522W).

Wow. Insane data point. Someone somewhere (I think it was a github thread) was bemoaning buying the 14600K and saying that 180W isn't even enough to feed the 6 P-cores properly, let alone the additional E-cores. If that's true, what has Intel created, seriously? They really must badly want the poles to melt!

 

[CouncilorIrissa]

Aussie Steve chimes in.

ASUS and Gigabyte apparently have different ideas on what the Intel Baseline Profile is supposed to constitute.

 

[Mahboi]

Ballsy cameraman, getting this close to get a mugshot of a furious Aussie creature just about to pounce.