Intel processors crashing Unreal engine games (and others)

[DAPUNISHER]

Start around minute 12 to see the cpu voltage during cinebench runs after running the new baseline settings on the mobo. From 1.3v to 1.6v.

He was fairly succinct about his testing on twitter. If any of you can stomach going there.

 

[uzzi38]

That s still too much, power should be around the 7950X3D value, or at worst like a 142W PPT 5950X, guess that it s what was planned before they changed their plans
to keep up with Intel s power guzzling orgy.

AMD always planned for AM5 to run up to 230W, they didn't specifically just push the chips that far at the last minute. Motherboards and sockets were designed with that 230W power limit for the upper SKUs in mind.

 

[Abwx]

AMD always planned for AM5 to run up to 230W, they didn't specifically just push the chips that far at the last minute. Motherboards and sockets were designed with that 230W power limit for the upper SKUs in mind.

230W was more related to an eventual future core count increasement or a new uarch, if you look at the early AMD announcements they didnt target as much perf increasement for the 7950X, it s just that after the 5950X Intel used 241W as a mean to be on par and it made no doubt that they would keep this inflated TDP for their next SKUs, hence the 7950X being pushed at its limits as a preemptive strike, remember that they were not sure of the final power and were undecided till the last moments.

FTR they targeted 30% better perf in their early slides, and the 7950X ended 38% faster than the 5950X, at 142W PPT it s still 31% faster, so you can see the pattern, 6% more perf were required to outmatch a soon to come 253W 8 + 16 RPL, after wich Intel increased even more their TDP to be on par, at 335W they match the 230W PPT 7950X.

 

[Mopetar]

So many fine words to describe this mess, so few of them allowed on the forum.

A hot mess is certainly apt and allowable.

Plus, this way none will have jumped on the story and hence Intel's PR people can hardly blacklist them!

So either way they're getting burned by Intel?

I kind of want to see the company double down though. 700W Arrow Lake! A baseline so hot you can use it to freebase the drugs you'd need to be on to consider purchasing one. A most premium product, by which we mean your fire insurance premiums to have one in your house will be the most. The i9 won't come with a cooler in the box, but will ship with its own substation transformer.

 

[S'renne]

A hot mess is certainly apt and allowable.



So either way they're getting burned by Intel?

I kind of want to see the company double down though. 700W Arrow Lake! A baseline so hot you can use it to freebase the drugs you'd need to be on to consider purchasing one. A most premium product, by which we mean your fire insurance premiums to have one in your house will be the most. The i9 won't come with a cooler in the box, but will ship with its own substation transformer.

Is this what this forum is like now, just joining the hate train without any meaningful replies

 

[H433x0n]

Is this what this forum is like now, just joining the hate train without any meaningful replies

It’s been that way.

 

[igor_kavinski]

Intel is using the Dark Side of the Force so we are giving them the hate they need to get even stronger.

 

[gdansk]

Is this what this forum is like now, just joining the hate train without any meaningful replies

Better than deflecting with FUD.

Until Intel announces they'll replace any CPU exhibiting this error they deserve all the criticism they can get.

 

[Markfw]

Better than deflecting with FUD.

Until announces they'll replace any CPU exhibiting this error they deserve all the criticism they can get.

Actually, until they introduce a CPU that does suck power like crazy, and is somewhat competitive they deserve all the criticism they can get.

 

[H433x0n]

Actually, until they introduce a CPU that does suck power like crazy, and is somewhat competitive they deserve all the criticism they can get.

Hasn’t that been beaten to death at this point?

 

[gdansk]

Actually, until they introduce a CPU that does suck power like crazy, and is somewhat competitive they deserve all the criticism they can get.

No, there's a difference here. It's fast and hot but Intel said it'd work like that. It seems now, with some time, that this doesn't apply to everyone.

 

[S'renne]

I'll probably come back after a year once this rolls over, I just came here for the leaks not the flames(and you guys are going to take this out of context again reinforcing my reasons)

 

[Markfw]

I'll probably come back after a year once this rolls over, I just came here for the leaks not the flames(and you guys are going to take this out of context again reinforcing my reasons)

The last post I saw from you aside from this train of thought was that AMD CPU's are unstable that you got from someone. If you bad mouth the only good current CPU (as far as stability and power usage) what do you expect ? Intel is this thread, and AMD has proven stable for 7 years, so what do you expect ?

Truth is what I see here, not anything else.

 

[Bigos]

I'll probably come back after a year once this rolls over, I just came here for the leaks not the flames(and you guys are going to take this out of context again reinforcing my reasons)

Maybe just look at the leak threads then?

 

[Mahboi]

Hasn’t that been beaten to death at this point?

The fact that it has been beaten to death is the reason I scarcely bother to read Intel threads.
There is nothing to say. Intel is a burning wreck. That's it. The "good" side of intel is as credible to me as the "fix" for RDNA 3. Never found.

 

[igor_kavinski]

Intel can partially redeem themselves to me by giving us back the AVX-512 with E-cores also enabled and letting us choose to run AVX-512 software by pinning it to P-cores.

 

[coercitiv]

Is this what this forum is like now, just joining the hate train without any meaningful replies

It's always like this, we alternate between the few bits of information we get from external sources, the little we can investigate ourselves, and then give the company the proper treatment. It's the same way with AMD and Nvidia, nobody gets a free lunch. The more they delay and obfuscate, the more people jump on the train.

Also, you need to understand this issue is the result of something we saw coming ever since Coffee Lake Refresh. That's almost five years since they started messing with the stock settings. I remember arguing on the forums that pushing power limits beyond spec is unhealthy, and for all intents it should be considered overclocking. At the time I did not find many sympathetic ears because our collective definition of overclocking was rooted into the idea that OC meant operating beyond stock turbo clocks. Maybe now, after folks see how power limits affect voltage and stability in the upper frequency region, opinions will change.

Also, look at the difference in behavior towards Intel with respect to this issue and with respect to their GPUs and driver features/stability. It's the same forum, the same people, the same company, very different outcome in behavior. We are actually cheering for Intel and patiently waiting to get their act together. Intel is also behaving differently, they are actively communicating through a competent employee and correcting issues. Compare their GPU PR efforts with the awkward silence on this CPU stability topic. They are investigating for over 2 months now, and will issue a public statement in May. Meanwhile mobo makers release underbaked firmware fixes that are supposed to fix something they're not willing to tell us about.

 

[DrMrLordX]

That s still too much, power should be around the 7950X3D value, or at worst like a 142W PPT 5950X, guess that it s what was planned before they changed their plans
to keep up with Intel s power guzzling orgy.

230W PPT never damaged any Zen4 that I'm aware of. And 7950X3D never gets there anyway, it's the original 7950X that did so.

It's the same forum, the same people, the same company, very different outcome in behavior.

Intel seems to be pointing the finger at mobo OEMs and sitting this one out. Something they can't exactly do with their video cards. It's gonna be up to the enthusiast community to fix this problem with a combination of manual UEFI tweaks and/or XTU tweaks to get systems stable and performing as close to launch numbers as possible.

 

[igor_kavinski]

manual UEFI tweaks and/or XTU tweaks to get systems stable

At the very least, Intel should mandate the mobo makers to replicate the XTU settings in their UEFI screens and then update the XTU to warn if any parameter value is running out of spec.

Come on, Intel! Not a gargantuan task, is it?

UNLESS...

Whichever executive or division president is responsible for this mess, has no idea about these things...

EDIT: That would be, Michelle Johnston Holthaus, If Google isn't wrong.

In her previous role, Holthaus was chief revenue officer and general manager of the Sales, Marketing and Communications Group.

Yep. It definitely makes sense now.

The Client frickin' Group is being spearheaded by a marketing person!

And she looks just like my company's extremely annoying Chief Marketing Officer. I mean, same cheeks!

I would bet my non-existent billions that she has no clue about platform power delivery...

 

[DrMrLordX]

At the very least, Intel should mandate the mobo makers to replicate the XTU settings in their UEFI screens and then update the XTU to warn if any parameter value is running out of spec.

Different CPU samples may need to be tweaked . . . differently. It's not going to be absolutely uniform.

 

[DAPUNISHER]

From a user over at PCMR this morning -

I have a i9-14900k and for the life of me I can not play a game that uses unreal engine 4 because they all crash. I changed the bios setting to intel fail safe and still nothing changed. Usually its the shader decompress error or it won’t say anything and will just crash. Anyone know a solution because it really is irritating.

I recommended they RMA.

Where there's smoke there is usually fire. These failures are prevalent enough that the SNAFU has persisted instead of being the scandal du jour that burns out quickly. To the contrary, the fire keeps growing as more users experience the failures with the same gaming workloads.

Gaslighting, astroturfing, and deflection is not going to make it go away. Nor should this issue be resolved by owners of the hardware, except as a stop gap. The onus is on the manufacturer. Their delay in addressing this isn't only due to the "investigation". But also because the legal department is advising at every step to shield the company from as much liability as possible.

These threads become war wagons with purpose: To raise awareness for those searching for solutions and answers as to why their games are crashing or erroring out. And to do our part, however small, to force action from the companies.

 

[igor_kavinski]

Different CPU samples may need to be tweaked . . . differently. It's not going to be absolutely uniform.

Intel should provide guidelines for that since it seems they can't figure out an automated way to do that.

 

[KompuKare]

Intel should provide guidelines for that since it seems they can't figure out an automated way to do that.

Intel's response:
"If we were able to find golden tickets in the silicon lottery via software, do you not think we'd all be buying lottery ticket?"

 

[igor_kavinski]

Intel's response:
"If we were able to find golden tickets in the silicon lottery via software, do you not think we'd all be buying lottery ticket?"

My conspiracy theory: All their real engineers jumped ship to AMD as soon as Zen 3 was released

And no, they are not all working on CPUs. Some of them might be doing junior level stuff and enjoying it, after the politics they suffered at Intel.

 

[CouncilorIrissa]

I mean, what kind of reaction does one expect from such a thread otherwise?

The fact of the matter is that Intel priced their high-end parts based on their performance in benchmarks that are run with unlimited power. What we’ve learnt is that the CPUs either can not sustain this level of performance or they are degrading over time. So now we have two options: either they knew that was the case and kept silent to score good benchmarks on launch (i.e. intended to mislead their customers), or they didn’t know, which makes them incompetent.

Now I don’t know which one is worse, but staying silent isn’t really an option. An enthusiast might be able to figure out how to fix the problem, but an average user can’t, nor should be expected to.

This is something that should've never made it market.