Intel processors crashing Unreal engine games (and others)

[BFG10K]

Intel with 12,13,14 generation did EVERYTHING to win benchmarks against AMD, too much voltage, too much heat, no protections (or way too few), too high GHZ (they got that only through too much voltage and heat). They deserve any bad press or sales hits due to this.

This pretty much agrees with what I put in the OP. Both Intel and mobo vendors know exactly what the cause is but they're hoping the issue issue blows over rather than confronting it. They won't dare do anything that risks lowering benchmarks or a class action lawsuit.

Nothing new here, Intel does this every time AMD starts beating them. Industrial water cooler @ Computex, Prescott throttling @ stock, faulty 1.13 GHz P3 recall because it couldn't compile the Linux kernel.

Heck, E-cores are nothing more than Cinebench accelerators specifically designed to win benchmarks because Intel can't compete with real cores. To hell with the people trying to actually use the chips for real work and finding all sorts of scheduling problems with the async design.

The solution? An "Intel accelerator utility" that disables them. You know, for all those other times when you aren't constantly looping Cinebench.

First reported in February, here we are in April, Intel still "investigating". How come a $161 billion corporation has so much trouble figuring this out when according to some people in this very thread, they've solved it? Just point Papa Gelsinger over here and things should be sorted quickly, amirite?

 

[Ranulf]

Nvidia recommends consulting Intel if having stability problems with raptorlake in latest gpu driver notes.

NVIDIA recommends Intel Raptor Lake CPU users to consult Intel in case of stability issues - VideoCardz.com

Intel Raptor Lake 13/14th Gen Core stability issues now mentioned in GPU driver notes NVIDIA is aware of Intel problems, recommends consulting CPU maker in case of stability issues. NVIDIA has acknowledged reported issues affecting Intel CPUs, particularly the 13th Gen Core and 14th Gen Core...

videocardz.com

Edit to add earlier report that Intel is investigating the issue:

Intel 13th/14th Core "Raptor Lake" gaming instability is now being investigated - VideoCardz.com

Tekken 8 instability issues on Intel 13/14th Gen Core series sparks investigation South Korean media report that Intel is now investigating the problem of Raptor Lake CPU crashes. According to ZDNET Korea, Intel is now investigating the issue related to gaming instability as observed in games...

videocardz.com

 

[igor_kavinski]

Use your new 13/14th gen CPU as much as possible at max speed before it degrades, folks! Time is running out!

 

[AdamK47]

Not good for Intel. That's what happens when a CPU runs at the ragged edge out of the box.

How much are motherboard manufacturers to blame?

 

[DAPUNISHER]

How much are motherboard manufacturers to blame?

Dr. Ian Cutress places most of the blame on them. He explains why in the vid I posted. It amounts to their engineers always pushing the limits for bigger bar better. Of course Intel could squash all of that if they really wanted. I don't think you'll find anyone without a horse in the race that doesn't think they bare some of the blame for at the very least, tacitly allowing their board partners to engage in these practices.

 

[lakedude]

Reminds you what... of the days of the Pentium /// 1.0 Ghz CPU? That was later recalled?

Oh man I remember those fire breathers.

My buddy had a 1GHz laptop that may have had a desktop CPU in it, regardless it put out heat like a 1500 watt hair dryer.

I was such a jerk, LOL. I'd take his Coke and put it in the path of the hot exhaust. After maybe 15 minutes the Coke would be undrinkable.

Sometimes he would notice right away but other times that Coke got as warm as coffee or tea.

 

[Mopetar]

Not good for Intel. That's what happens when a CPU runs at the ragged edge out of the box.

How much are motherboard manufacturers to blame?

They're selling the board. It's their product. Does Intel say that their boards can't explode and spray acid in a customer's face. Probably not, but would anyone be okay with a board that did that? I think all of this used to go without saying, but it's apparent that now it needs to be said.

No one company will stop because actually staying within specifications means they'll be dead last on a bar graph and that's going to hurt sales. Intel should at least force them to require users flip a switch or type in the power overwhelming cheat code if they want to go lay with fire.

Customers should have to be paranoid about their technology being configured to destroy itself out of the box. As others have pointed out, Intel isn't losing much and if they still really feel they need that last few percentage points, just put it behind a BIOS setting that warns the user and then it's squarely on them.

 

[BFG10K]

HWUB weighing in on the matter.

 

[tamz_msc]

ASUS has outed new BIOS with updated settings profile called Intel Baseline Profile.

Finally this crap is going to end. Now Intel just needs to enforce this profile on initial setup across all vendors.

 

[bononos]

.........
Heck, E-cores are nothing more than Cinebench accelerators specifically designed to win benchmarks because Intel can't compete with real cores. To hell with the people trying to actually use the chips for real work and finding all sorts of scheduling problems with the async design.
..........

Which benchmarks expose this issue directly. From a quick look at the AT benches, Intel seems to be doing ok except for the scientific-simulation computational suite like the 3D particle movement, ONNX runtime, Tensorflow benches. Are they multithreaded benches?

 

[coercitiv]

Finally this crap is going to end.

It's going to end if they make it the default profile and don't mess with it when enabling XMP. We'll have to wait and see.

 

[igor_kavinski]

From a quick look at the AT benches, Intel seems to be doing ok except for the scientific-simulation computational suite like the 3D particle movement, ONNX runtime, Tensorflow benches. Are they multithreaded benches?

What do you mean except? Did AT point out that the CPUs got too hot or unstable in these benches?

 

[bononos]

What do you mean except? Did AT point out that the CPUs got too hot or unstable in these benches?

I said in simulation and computation benchmarks AMD beats Intel by alot 30-50% even up to 8x in one 3d particle movement bench, wondering if that is because of the e-core architecture.

 

[tamz_msc]

I said in simulation and computation benchmarks AMD beats Intel by alot 30-50% even up to 8x in one 3d particle movement bench, wondering if that is because of the e-core architecture.

3DPM is Ian's code optimised by an ex-Intel guy for AVX-512.

 

[igor_kavinski]

I said in simulation and computation benchmarks AMD beats Intel by alot 30-50% even up to 8x in one 3d particle movement bench, wondering if that is because of the e-core architecture.

Which review are you referring to by the way? Must be a bit outdated. I think a direct die cooled 14900KS would reduce that lead by quite a bit. Could even turn the tables in non-AVX-512 benchmarks. You do pay too much for that in terms of heat, noise and power consumption though.

 

[DAPUNISHER]

Bryan says Intel blacklisted him. He goes after LGA 1700 fairly hard for this latest problem and tells his viewers do not buy for any of the higher end CPUs past the 12700K. And he advises turning off e-cores on it. After hearing his story about how he was treated, I'll state yet again - they let the wrong people go during those mass layoffs.

 

[SolidQ]

ASUS has outed new BIOS with updated settings profile called Intel Baseline Profile.

That's big drop

 

[tamz_msc]

That's big drop

Meh. People who care about such things should probably not update.

It's quite funny seeing folk on the overclock.net forums upset about this cause according to them if you're buying a $700 CPU with a $700 motherboard, you're doing it wrong if you don't overclock.

 

[yottabit]

I can confirm Intel has an issue with Unreal Engine games. I played PUBG on a 7600K and the game would run perfect and then introduce lag, latency and what seemed like video driver crashes but would recover. I think it's a CPU issue. I never thought the day would come where an i5 processor would be cut down by such an old game as PUBG. It's been around for 7 years. This reinforces the problem Intel CPU's have with unreal engine games. I hope they fix the problems.

You can’t really go by age with live service games. As an extreme example World of Warcraft in 2004 minimum requirements were a Pentium III 800 MHz. Nowadays it requires a Haswell/Zen quad core @ 3 GHz.

I can still run Half Life v1.0 retail fine on my Pentium 233 MMx but I can’t exactly go install the latest version off Steam and expect that to work.

PUBG specifically has gone through numerous updates including now using a much newer version of Unreal Engine. Each newer version of game engines tends to be a bit more “heavy” and I believe PUBG is currently in the middle of migrating to UE5 so, performance probably isn’t going to get better. Not to mention all the Spectre/Meltdown mitigations and Windows OS updates that have come out since launch. It’s also possible you’re getting some actual hardware related failure due to age of the system

 

[Hitman928]

Gamers Nexus has commented on it, not much in their actual video but there is some interesting discussion in the comments section.

We obviously have no way to verify this claim or how much volume this SI does to see if it is a significant amount, but GN seems to at least anecdotally confirm based upon viewer reports in a reply.



At this point, it's hard to give any kind of number on percent of CPUs effected, but if you are having this issue, it seems you should make the BIOS adjustments to follow Intel spec and if you are still having issues, RMA your CPU.

 

[S'renne]

I said in simulation and computation benchmarks AMD beats Intel by alot 30-50% even up to 8x in one 3d particle movement bench, wondering if that is because of the e-core architecture.

Does the benchmarks account for data integrity? I know an engineer that regularly compiles full Linux kernels that states AMD is indeed faster than Intel in those workloads, but does not have full precision unlike Intel systems

 

[igor_kavinski]

Confirms what I've suspected and expressed here before: Alder Lake is just a cobbled together mess of an architecture originally designed to be used in a chiplet based product. (I've surmised on the hurriedly cobbled together part many times on this forum)

Thank you so much, @DAPUNISHER

 

[igor_kavinski]

but does not have full precision unlike Intel systems

What does that even mean?

 

[S'renne]

What does that even mean?

Data corruption/errors, they're really costly, along with software optimisations that AMD didn't assist with for such workloads

 

[Hitman928]

Data corruption

Uhh, no. There is no issue with AMD CPUs and data corruption unless you've pushed a memory or CPU overclock too hard, but that's true of any CPU.