Intel processors crashing Unreal engine games (and others)

[BFG10K]

As if we needed any more evidence that these Intel furnace monstrosities are being pushed beyond their limit just to win performance benchmarks. No wonder even 480mm rads or 1.5kg coolers aren't enough for these solar flares.

These parts are being redlined because they're horrifically inferior to AMD X3D processors for gaming, and yet there are still some people on the internet defending Intel's version of Bulldozer.

This is the result of an absolutely incompetent monopoly that sat on 14nm+++ for seven years and was happy to keep selling you +5% performance + new motherboard every 18 months.

Recent high-end Intel CPUs are crashing Unreal Engine games

There have been several reports of Core i9-13900K and Core i7-14900K processor users experiencing crashes in games that show an 'out of video memory' error. The issue...

www.techspot.com

Is your Intel Core i9-13900K crashing in games? Your motherboard BIOS settings may be to blame — other high-end Intel CPUs also affected (Updated)

Apparently a 4096W power limit might be a bad idea.

www.tomshardware.com

 

[zir_blazer]

This is not explicitly Intel fault. Is simply than Motherboards vendors leave unlimited Power Limits or MCE (MultiCore Enhancement) on by default, so you are overclocking out-of-the-box without knowing it.

 

[Dave3000]

What are the chances that setting to XMP in the BIOS can play a big part on this instability? Afterall, XMP is considered a form of overclocking as in it runs the memory at speeds faster than what the CPU officially supports, and I wouldn't be surpised if there are people who build their own systems and don't pay attention to what the memory controller on the CPU actually supports and just buy the fastest memory they can afford and that motherboard they buy supports and then set to XMP in the BIOS to get their memory to run at it's advertised speed, even if they don't have the intention of overclocking. Also I used to own a motherboard that if I enabled XMP, it would overclock the all-core turbo speed to it's single-core turbo speed for all core loads.

 

[Markfw]

This is not explicitly Intel fault. Is simply than Motherboards vendors leave unlimited Power Limits or MCE (MultiCore Enhancement) on by default, so you are overclocking out-of-the-box without knowing it.

I don't think most vendors (if any) leave this on by default. Intel experts chime in here ?

Apparantly tomshardware says SOME (but does not say who, or how many) do set this to 4096 watts and 4096 amps, but I find it hard to believe that many would do something so insane.

 

[jpiniero]

This is not explicitly Intel fault. Is simply than Motherboards vendors leave unlimited Power Limits or MCE (MultiCore Enhancement) on by default, so you are overclocking out-of-the-box without knowing it.

Unlimited power limits shouldn't cause it to crash. MCE might... but that would mean you got a dud that can't do ACT stably enough on all of the P cores.

 

[CouncilorIrissa]

Intel absolutely have the means to force mobo manufacturers to run their CPUs within spec, they simply chose not to because it makes benchmarks look better.

 

[H433x0n]

I don't think most vendors (if any) leave this on by default. Intel experts chime in here ?

Apparantly tomshardware says SOME (but does not say who, or how many) do set this to 4096 watts and 4096 amps, but I find it hard to believe that many would do something so insane.

If you enable XMP you’ve inadvertently enabled MCE for the popular motherboards. This setting means unlimited current and uncapped power limits.

 

[adroc_thurston]

This is not explicitly Intel fault. Is simply than Motherboards vendors leave unlimited Power Limits or MCE (MultiCore Enhancement) on by default, so you are overclocking out-of-the-box without knowing it.

Intel can use the boot™ and enforce limits in DIY PC like they do pretty much everywhere else.

 

[tamz_msc]

These parts are being redlined because they're horrifically inferior to AMD X3D processors for gaming, and yet there are still some people on the internet defending Intel's version of Bulldozer.

Holy hyperbole!

 

[tamz_msc]

Most, if not all Intel motherboards have some performance enhancing settings turned on by default.

This has been a problem since the days of Coffee Lake.

Intel just needs to stop allowing concessions to OEMs that are used by them to exploit loopholes in the specifications.

 

[tamz_msc]

As expected, the culprit is ICC_max set to above spec - it should be 307 A for the 14900K.

https://www.reddit.com/r/overclocking/comments/1axepvu

ICC_max is more important than the PL1/PL2 values.

Much ado about nothing.

 

[gdansk]

Much ado about numerous vendors shipping their motherboards with bad defaults is not nothing.
Most the people reporting the issue had no idea they were overclocked. Intel can and should put the kibosh on vendors shipping firmware like this.

 

[tamz_msc]

Much ado about numerous vendors shipping their motherboards with bad defaults is not nothing.
Most the people reporting the issue had no idea they were overclocked. Intel can and should put the kibosh on vendors shipping firmware like this.

This has been going on for 7 years at least. It was first observed with 8th gen Coffee Lake. Only it wasn't a problem back then probably because 14nm++ is more tolerant of transients than Intel 7.

 

[IEC]

I set PL1 and PL2 to 105W on my Z690 rig just for efficiency sake, but I had no idea ICC_max was even a configurable setting.

Once you have to change out of the box settings you're already past 90% of users, whether Intel or AMD...

 

[Saylick]

I set PL1 and PL2 to 105W on my Z690 rig just for efficiency sake, but I had no idea ICC_max was even a configurable setting.

Once you have to change out of the box settings you're already past 90% of users, whether Intel or AMD...

Yeah, if enthusiasts dont even know about how to fix this issue, average Joe would not stand a chance.

 

[BFG10K]

As expected, the culprit is ICC_max set to above spec - it should be 307 A for the 14900K.

https://www.reddit.com/r/overclocking/comments/1axepvu

ICC_max is more important than the PL1/PL2 values.

I'd never heard of this setting until now and had to look it up. I know how to write a double-linked list in C using free() and malloc(). I bet most people in this forum can't do that, so I'm not a computer illiterate moron by any stretch of the imagination.

A lot of tech enthusiasts don't even know about XMP or how to set a high refresh rate,, much less something as obscure as ICC_max.

Much ado about nothing.

Then it should be absolutely no problem for Intel to proactively contact all major CPU reviewers and ask them to retest @ spec. Amirite?

 

[tamz_msc]

I'd never heard of this setting until now and had to look it up. I know how to write a double-linked list in C using free() and malloc(). I bet most people in this forum can't do that, so I'm not a computer illiterate moron by any stretch of the imagination.

And I absolutely detest writing computer programs (I find it terribly boring - I'd rather do a math problem), and yet I know about ICC_max.

Then it should be absolutely no problem for Intel to proactively contact all major CPU reviewers and ask them to retest @ spec. Amirite?

Yes, obviously.

Then the conclusion of Intel CPUs guzzling power would magically disappear, which will then go against the prevalent narrative.

 

[BFG10K]

And I absolutely detest writing computer programs (I find it terribly boring - I'd rather do a math problem), and yet I know about ICC_max.

Nobody cares what you find boring because it's irrelevant.

Then the conclusion of Intel CPUs guzzling power would magically disappear, which will then go against the prevalent narrative.

The performance will magically disappear too, that's why Intel happily look the other way when benchmarks are posted.

And it's not really a narrative, more like objective proof. 14900K overall slower than X3D for gaming while absolutely chewing through power. Also throttles @ stock even with a 360mm AIO liquid cooler.

 

[coercitiv]

I can't feel properly vindicated about this because I know it must have been very painful for people to figure out why these systems were unstable (both users and game devs). Still, it does feel nice to finally have public proof that all this irresponsible board maker behavior is toxic for the consumer. Also, the biggest irony is the folks affected are the ones buying the premium hardware.

Like @zir_blazer said, this is not explicitly Intel's fault. Still, they were complicit. Now they'll pay the PR price.

 

[coercitiv]

I don't think most vendors (if any) leave this on by default.

Here's how it happens on my MSI Z690 board. Any time I make a change that affects my CPU multiplier or enable XMP, I get this pop-up message in UEFI. You'll notice there is no Cancel button. MSI decided that consumers don't need Cancel buttons, that would only confuse them. And just in case someone tries to say "but", one gets this prompt even if they try to underclock, lower CPU frequency below stock.



So what is Water Cooler setting?! Well, I'm SO GLAD you asked! Let me introduce you to the Power Menu à la carte:



Current limits:

• Boxed Cooler 240A
• Water Cooler 512A

but I find it hard to believe that many would do something so insane.

I would expect that most boards with Z chipsets and most B/H chipset "gaming" boards with powerful VRM stages will exhibit this behavior. The more expensive the board, the more likely it will behave like this. For the example above, mine was one of the cheaper Z690 models and still does it. In other words, all the boards that one would buy together with a 13900K or 14900K will fit the profile.

 

[tamz_msc]

Nobody cares what you find boring because it's irrelevant.

Nobody cares if you are a coding wizard either, it's irrelevant as well. But this is a forums for PC hardware enthusiasts, and it reflects poorly on those who flaunt the tag 'enthusiast' if they don't know about the basic things they ought to know about their hobby.

The performance will magically disappear too, that's why Intel happily look the other way when benchmarks are posted.

They perform practically the same if you remove the outlier reviews like Techspot/HWUB (which favour AMD) and some other reviews by various European outlets which favour Intel.

The only difference is the power consumption while gaming - and yes, this thread is about why that is the case.

 

[BFG10K]

Nobody cares if you are a coding wizard either, it's irrelevant as well. But this is a forums for PC hardware enthusiasts, and it reflects poorly on those who flaunt the tag 'enthusiast' if they don't know about the basic things they ought to know about their hobby.

Nah, it's very much relevant. Someone that builds PCs and codes in C has vastly more knowledge than the average PC user. If you're seriously suggesting all 14900K owners should know ICC_max is intentionally out of spec and causes issues, your argument is pure delusion at its finest.

Your first two posts didn't even mention ICC_max. You obviously furiously ran some google searches after and then posted a follow-up how it's "obviously expected" to be the cause. Oh, you tech searcher, you.

They perform practically the same if you remove the outlier reviews like Techspot/HWUB (which favour AMD) and some other reviews by various European outlets which favour Intel.

"Outliers", because they don't fit the Intel apologist(tm) handbook. Intel could enforce these limits on reviewers and board makers (including mandating PL1/PL2), but they don't. Hmm, what could Intel possibly gain by allowing their hardware to run out of spec for internet benchmarks?

And...I wonder... is it the same idea that was behind their SPEC cheating?

SPEC Invalidates Over 2600 Intel CPU Benchmark Results Due to Custom Compiler Optimizations

The Standard Performance Evaluation Corporation (SPEC) has found that Intel's oneAPI DPC++ compiler was optimized in a way that breached SPEC's guidelines, leading to the disqualification of over 2,600 SPEC CPU 2017 benchmark scores for Intel processors.

www.guru3d.com

Or how about this one?

Intel: We 'Forgot' to Mention 28-Core, 5-GHz CPU Demo Was Overclocked

We met with Intel and asked a few pointed questions about its now-infamous 28-core CPU demo.

www.tomshardware.com

 

[Thunder 57]

And I absolutely detest writing computer programs (I find it terribly boring - I'd rather do a math problem), and yet I know about ICC_max.

Yes, obviously.

Then the conclusion of Intel CPUs guzzling power would magically disappear, which will then go against the prevalent narrative.

And so would their performance. Intel gets creamed at lower wattages.

A Lighter Touch: Exploring CPU Power Scaling On Core i9-13900K and Ryzen 9 7950X

www.anandtech.com

 

[tamz_msc]

And so would their performance. Intel gets creamed at lower wattages.

A Lighter Touch: Exploring CPU Power Scaling On Core i9-13900K and Ryzen 9 7950X

www.anandtech.com

Please make strawman arguments elsewhere. Nowhere did I say anything about testing at lower wattages.

 

[BFG10K]

Nowhere did I say anything about testing at lower wattages.

But that's exactly what testing "at spec" means. It includes forcing Intel's default PL1 / PL2, whose wattages are lower than the unchecked furnaces Intel allows in benchmarks.