Intel processors crashing Unreal engine games (and others)

[igor_kavinski]

https://www.reddit.com/r/intel/comments/17aob62/13900k_wheas_19_or_bsod_during_shader_compilation

That one WHEA 19:
Corrected Machine Check
Translation Lookaside Buffer Error

The fix was so easy. Intel just needed to have a BIOS setting to disable the TLB.

/s

Sane settings being discussed here:

https://www.reddit.com/r/intel/comments/1d003xo/my_intel_14700k_bios_setting_after_the_14th_gen

 

[Markfw]

First and foremost, I blame Intel for being irresponsible in creation of their chips and bios recommendations to MFG's to win benchmarks. It makes no sense for a motherboard manufacturer to want to win benchmarks and have customers blame them for problems with their CPUs, they want sales and happy customers. I have been against it since day one of alderlake. I got rid of it due to insane power use, and was told by someone that I lied, but now I have proof that I was right.

 

[igor_kavinski]

(and would that even catch the CPU making a mistake?)

As shown in the link I posted above, ECC won't help when the CPU's own buffer gets borked!

 

[igor_kavinski]

I have been against it since day one of alderlake. I got rid of it due to insane power use, and was told by someone that I lied, but now I have proof that I was right.

Yeah, for normal use it's fine but any time all cores are in use, the fans go crazy and i've had CBR23 10 minute benchmark shut down the PC after about 5 or 6 minutes of MT when my 12700K is mildly OCed (4.9 GHz all core) at 260W. For 24/7 workloads, it will need to be tamed with lower all core clocks or heavy duty cooling which is not required with Ryzen.

 

[Thunder 57]

First and foremost, I blame Intel for being irresponsible in creation of their chips and bios recommendations to MFG's to win benchmarks. It makes no sense for a motherboard manufacturer to want to win benchmarks and have customers blame them for problems with their CPUs, they want sales and happy customers. I have been against it since day one of alderlake. I got rid of it due to insane power use, and was told by someone that I lied, but now I have proof that I was right.

I wouldn't go that far. Even 20 years ago we had motherboard makers trying to win benchmarks by fudging the FSB. I vaguely remember MSI having a 103MHz FSB for example.

 

[Markfw]

Yeah, for normal use it's fine but any time all cores are in use, the fans go crazy and i've had CBR23 10 minute benchmark shut down the PC after about 5 or 6 minutes of MT when my 12700K is mildly OCed (4.9 GHz all core) at 260W. For 24/7 workloads, it will need to be tamed with lower all core clocks or heavy duty cooling which is not required with Ryzen.

Mine was running avx-512 ! That was early bios, before it was disabled.

 

[coercitiv]

i've had CBR23 10 minute benchmark shut down the PC after about 5 or 6 minutes of MT when my 12700K is mildly OCed (4.9 GHz all core) at 260W.

Play "mild oc" games, win thermal shutdown prizes.

 

[igor_kavinski]

Play "mild oc" games, win thermal shutdown prizes.

Curious why it shuts down. Shouldn't it throttle and keep going?

 

[lakedude]

Curious why it shuts down. Shouldn't it throttle and keep going?

Good question!

Back in the day (before AMD had built in thermal protection) Tom's Hardware published video of them taking the heat sinks off several processors. I watched in horror as the AMD chips let out the magic smoke.

Early Intel chips locked up but didn't burn up. Then I watched in amazement as the newer (at the time) Intel chips just slowed down but kept working with no heatsink.

Are we saying the current chips have lost that ability?

 

[Markfw]

Good question!

Back in the day (before AMD had built in thermal protection) Tom's Hardware published video of them taking the heat sinks off several processors. I watched in horror as the AMD chips let out the magic smoke.

Early Intel chips locked up but didn't burn up. Then I watched in amazement as the newer (at the time) Intel chips just slowed down but kept working with no heatsink.

Are we saying the current chips have lost that ability?

There is a picture here somewhere, where I was having a problem with a new build, and it was running at 500 mhz. I figured out the problem, that I had not taken the plastic off the HSF before mounting. I think it was a 7950x. It definitely slowed down by itself to that speed, and has run @100% load 24/7 ever since I fixed it. So I can tell you for sure, AMD has this protection.

 

[lakedude]

I can tell you for sure, AMD has this protection.

I figured.

Back in the day our company avoided selling AMD specifically because of this issue.

More recently we are rocking an AMD 7600x desktop and an AMD 7735hs laptop.

 

[igor_kavinski]

Are we saying the current chips have lost that ability?

Let's wait for coercitiv to test his chip

 

[coercitiv]

Let's wait for coercitiv to test his chip

First they throttle, if the throttling isn't enough, they shut down.

When a core exceeds the set throttle temperature, it will reduce power to maintain a safe temperature level. The throttle temperature can vary by processor and BIOS settings. If the processor is unable to maintain a safe operating temperature through throttling actions, it will automatically shut down to prevent permanent damage.

 

[igor_kavinski]

So the heat buildup is so great that even throttling is unable to prevent temps from exceeding 100C!

 

[zir_blazer]

Back in the day (before AMD had built in thermal protection) Tom's Hardware published video of them taking the heat sinks off several processors. I watched in horror as the AMD chips let out the magic smoke.

There is a major drama history that belongs to the realm of digital archaeology about that video. At that time, Tom's Hardware was heavily sponsored and thus biased towards Intel, which happened somewhere after the Pentium 3 Coppermine 1.13 GHz recall that Intel began to put some big money on ads.
I think that Tom''s Hardware did a first video where they killed an Athlon Thunderbird. AMD incorporated thermal shut down in the next Athlon XP, but it required compatible Motherboards. Tom's Hardware tested it (And killed the Processor) on a random board instead of some Fujitsu-Siemens parts that were the earliest ones to implement the new feature. Additionally, I recall that having watched a video where they remove the heatsink from a Pentium 4 Willamate while playing Quake 3 and it throttles but doesn't crashes nor shut downs, and when they pass a termometer over it is marks some low 30°C or so temperature which was basically impossible (Since it throttles at around 100°C plus a further limit for critical shut down), then put the heatsink back again and it resumes playing full speed near instantly. These things caused some massive community uproar. If it wasn't due to linkrot you may be able to get forum threads about that matter and "live in the moment", I recall having googled a few hours about that in early 2010 or so.
Heck, I do even recall having scolded users even during 2007-2008 that refered to how worse AMD is because "they can burn". Tom's Hardware did cause some long term brand damage with that video...

 

[Thunder 57]

There is a major drama history that belongs to the realm of digital archaeology about that video. At that time, Tom's Hardware was heavily sponsored and thus biased towards Intel, which happened somewhere after the Pentium 3 Coppermine 1.13 GHz recall that Intel began to put some big money on ads.
I think that Tom''s Hardware did a first video where they killed an Athlon Thunderbird. AMD incorporated thermal shut down in the next Athlon XP, but it required compatible Motherboards. Tom's Hardware tested it (And killed the Processor) on a random board instead of some Fujitsu-Siemens parts that were the earliest ones to implement the new feature. Additionally, I recall that having watched a video where they remove the heatsink from a Pentium 4 Willamate while playing Quake 3 and it throttles but doesn't crashes nor shut downs, and when they pass a termometer over it is marks some low 30°C or so temperature which was basically impossible (Since it throttles at around 100°C plus a further limit for critical shut down), then put the heatsink back again and it resumes playing full speed near instantly. These things caused some massive community uproar. If it wasn't due to linkrot you may be able to get forum threads about that matter and "live in the moment", I recall having googled a few hours about that in early 2010 or so.
Heck, I do even recall having scolded users even during 2007-2008 that refered to how worse AMD is because "they can burn". Tom's Hardware did cause some long term brand damage with that video...

That was awhile ago and I was still pretty young, but I do recall Tom's being biased during the P4 era. It never made sense to me as Tom's was one of (if not) the first ones to sound the alarm about the P3 1.13GHz. It was them or HardOCP I think, then Anand got looped in.

 

[lakedude]

Tom's Hardware did cause some long term brand damage with that video...

IMHO it was AMD who damaged their own brand by leaving off a critical safety component, Tom's simply did the reporting.

I sent a friend an AMD system back then, the heatsink fell off in shipping and his brand new computer went up in smoke.

As I recall AMD left it up to the motherboard manufacturers to implement thermal protection but by the time the MB got hot enough to shutdown it was too late.

That was not all that different from the games Intel is playing now. Both damaged their own reputation.

 

[zir_blazer]

That was awhile ago and I was still pretty young, but I do recall Tom's being biased during the P4 era. It never made sense to me as Tom's was one of (if not) the first ones to sound the alarm about the P3 1.13GHz. It was them or HardOCP I think, then Anand got looped in.

Tom's Hardware had a guy named Van Smith as Senior Editor (Who I don't recall if it is the same one from the Pentium 3 1.13 GHz recall, but he claims to have been related to articles against Rambus) that quit somewhere around that time and based on the aftermatch seems to not have been in good terms.
Oh, here is more about the videos I mentioned about.

 

[igor_kavinski]

based on the aftermatch seems to not have been in good terms.
Oh, here is more about the videos I mentioned about.

Wow. Thanks so much. I somehow missed this bit of drama and always wondered why people doubted the journalistic integrity of THG. No wonder!

I wish Van Smith had succeeded in his opensourcemark/minibench venture. Looks like he either gave up or maybe no longer among the living.

EDIT: He seems to be a current AMD employee. Great.

 

[DAPUNISHER]

When the doctor himself owned and ran Tom's, it was always under the microscope. They have always had good contributors IMO. Just not all of them. Similar to digital foundry where Alex is concerned. It becomes digital founder's edition more oft than not when he is the contributor. He switched to the 7800X3D so there's that. When my tinfoil hat slips off the voices tell me that's because Nvidia does not directly compete with them on that front...yet. When the hat is firmly seated, my brain suggest the reason for the switch was that he was probably getting inundated with viewer comments telling him his 12900K? was bottlenecking the results.

Circling back to Tom's; never forget this contributor shamelessly shilling, and going to war with other tech news outlets over it-

On topic - The crashing issues being resolved with settings that lower performance, is as some of us stated early on, going to result in lawsuits. The popular view is that it falls under false advertising. Since you do not get the performance indicated by every bigger bar better review without risking CPU degradation. If enough proof/evidence emerges there are indeed shenanigans over RMAs of these damaged CPUs, that'll be a bigger scandal yet. We will see various countries governments get involved including the FTC here in the U.S.

It is a bummer that dirtbag corpos have so many juicy scandals going on, to the point Gamers Nexus has not been able to focus on this one. If Steve had not proven himself tirelessly for years, I'd be suspicious that his relationship with TAP is influencing his journalism. I will leave it to others to decide if his journalistic integrity is beyond reproach or not. I personally think it is, for what it's worth. And that he simply had too many other offenders under the microscope already.

 

[Hitman928]

TLDW:

HUB got confirmation it is CPU degradation.
Board partners are not happy with Intel, the treatment they are getting, and lack of clear guidance.
Intel is trying to sweep this under the rug. It seems to be working. Attention spans are short...squirrel!
You need to watch the part about Intel pulling shenanigans over RMAs yourself.

The fact that Intel communicated first with the tech media while leaving their board partners in the dark shows that they are more interested in trying to spin the narrative to avoid bad PR than actually making things right for their customers. Board partners having to go to media outlets to try and figure out what is going on is just a terrible situation all around. It's also sad that Intel doesn't actually have any kind of real certification program for their board partners to go through so even if they do finally put an actual spec out, they have no system to keep the board makers in check and make sure that they are actually following spec. At the very least, hopefully Intel learns from this and fixes their approach before the next gen boards arrive.

 

[DAPUNISHER]

We have seen Intel shrug off reputation damage again and again. Every court judgement they either string out indefinitely and/or it amounts to a great ROI compared to what the judgement was over.

I suppose my point is, they will learn nothing from this. If they ever need to crank it up to 11 again to make bigger bar better, they will do it without hesitation. Unless an unprecedented financial judgement and other penalties are levied against them. Which based on precedence seems highly improbable.

I think Igor's Lab was the first to nail it as permanent damage? If so, props to him for calling it early.

 

[Joe NYC]

It is a bummer that dirtbag corpos have so many juicy scandals going on, to the point Gamers Nexus has not been able to focus on this one. If Steve had not proven himself tirelessly for years, I'd be suspicious that his relationship with TAP is influencing his journalism. I will leave it to others to decide if his journalistic integrity is beyond reproach or not. I personally think it is, for what it's worth. And that he simply had too many other offenders under the microscope already.

What is TAP?

 

[DAPUNISHER]

What is TAP?

Tom 'TAP' Petersen, the face of Intel Graphics. Ryan Shrout was dead weight.

 

[Saylick]

Tom 'TAP' Petersen, the face of Intel Graphics. Ryan Shrout was dead weight.

I think Tom’s middle initial is A, hence TAP.