VFX studio's Unreal Engine supervisor reports 50% failure rate for Intel Raptor Lake CPUs, prompting switch to AMD

[midian182]

What just happened? Intel is receiving more verbal hammerings over its crashing 13th- and 14th-generation CPUs. This time, an Unreal Engine supervisor at ModelFarm has revealed that it is experiencing a 50% failure rate for its systems powered by Core i9-13900K and 14900K processors. If that wasn't bad enough publicity for Intel, he also revealed that the company plans to replace Team Blue's chips with the upcoming Zen 5 Ryzen 9 9950X processors.

Dylan Browne, an Unreal Engine Supervisor and Feature Film VFX at the ModelFarm visual effects studio, posted the damning Intel-chip stats on X. He confirmed that the affected machines were exhibiting the widely reported crashes and instability.

For those curious at work our failure rate for our 13900k and 14900k machines is about 50% so far, any new machine builds going to be 9950x's, production environments need reliability#intel

– Dylan Browne (@DylserX) July 23, 2024

Browne wrote in a follow-up post that two of the chips exhibited instability problems straightaway, while the others took a while longer. He added that the 50% failure rate was within a year and that they are likely using Asus ROG motherboards – the instability is present even when power limits are lowered.

I'm away on a shoot at the moment but can check Thurs or Friday, fairly sure most are ROG ones, they haven't "fully" failed but exhibit the instability even when lowering power limits, haven't tested new bios' yet been on loan to another company for 3 months aha

– Dylan Browne (@DylserX) July 23, 2024

RAD, the company behind the Bink video codec and Oodle data compression technology, said in February that the problem with the chips was a combination of overly optimistic BIOS settings and the high clock speeds and power usage of Intel's processors. It emphasized that there were no bugs in Oodle or Unreal causing the issue.

The crashes have been causing problems for owners of the 13th/14th-gen CPUs for a while now. Trying to play demanding games, often those powered by Unreal Engine 5, results in errors when compiling shaders. Other CPU-intensive tasks such as running Cinebench and Handbrake causes similar problems.

Intel initially recommended reducing overclock and voltage settings to help mitigate some contributing factors. It only identified the problem this week: a faulty microcode algorithm that is mistakenly overvolting the processors.

The company will release a microcode mitigation patch in mid-August via BIOS updates from motherboard OEMs and Windows updates. This is more to prevent the problems occurring on chips that are currently unaffected as those that are experiencing issues are likely to have been damaged and need replacing by Intel.

AMD will undoubtedly be experiencing schadenfreude at Intel's expense, especially as Browne said his firm is switching to Team Red. He said that the Ryzen 9 9950X processors' single-threaded performance was "pretty damn good," adding that his own "5950X is good for single-threaded [performance], and they get better every [generation], not to mention having all core being equal unlike most of Intel's chips."

h/t: Tom's Hardware

Permalink to story:

VFX studio's Unreal Engine supervisor reports 50% failure rate for Intel Raptor Lake CPUs, prompting switch to AMD

 

[winjer]

"Silicon lottery" just took a whole new meaning.

 

[VaRmeNsI]

"AMD will be feeling smug"

I imagine the first thing they did when this came out was frantically check they weren't also affected.

 

[rmcrys]

"AMD will be feeling smug"

I imagine the first thing they did when this came out was frantically check they weren't also affected.

It can happen but Intel is being playing with fire fire quite some time: they’ve been working with high temperatures and frequencies and energy use/charges to overcome the architectural challenges (and lack of innovation); they’ve been working with “shortcuts” in the architecture so they get faster chips at cost of much more vulnerabilities. My 8th gen Intel with the vulnerabilities updates got extremely slow as an example.

AMD has been much better on this matter: cooler reliable chips; usually better pricing; socket lasts long.

My Legion Go with the Z1 Extreme is now my portable console and my desktop computer (attached to a RTX 3060Ti on a eGPU case and a monitor/keyboard/mouse). It’s used very little energy and is very snappy. I hope the new Ryzen AI meets the expectations as well as the RDNA4

 

[hahahanoobs]

50% failure rate in one year better spur more than this one complaint. I would expect at least 2 or 3 more companies to come forward in the coming days. Until then I can't get too excited about this. Kinda odd he didn't confidently know what mobos were being used offhand. More odd was how the conversation almost turns into a "AMD is mostly faster anyway." Then why didn't you buy AMD instead? I need to hear more.

 

[HardReset]

More odd was how the conversation almost turns into a "AMD is mostly faster anyway." Then why didn't you buy AMD instead? I need to hear more.

They didn't buy AMD because they "thought" Intel is more reliable, stable, has better features etc, just because it's Intel.

If based about facts only, there are quite small amount of user scenarios this decade where Intel is really better option than AMD.

 

[Puiu]

They didn't buy AMD because they "thought" Intel is more reliable, stable, has better features etc, just because it's Intel.

If based about facts only, there are quite small amount of user scenarios this decade where Intel is really better option than AMD.

People can always use them to warm up during the harsh winters of the northern regions/countries.

 

[trparky]

More odd was how the conversation almost turns into a "AMD is mostly faster anyway." Then why didn't you buy AMD instead? I need to hear more.

Because like IBM, nobody got fired for buying Intel.

Now? Not so much.

 

[MarcusNumb]

Big chance for AMD this time, I really hope that they will gain more market share this year and next year.

 

[hahahanoobs]

Because like IBM, nobody got fired for buying Intel.

Now? Not so much.

Doubt.

 

[pencea]

"AMD will be feeling smug"

I imagine the first thing they did when this came out was frantically check they weren't also affected.

Ironically, they actually did check.

 

[daffy duck]

Couldn't help the dig at AMD could you Techspot. I'm sure Steve will still spin this as AMD's fault though.

 

[Skjorn]

Still kinda feels like a Unreal Engine problem. It's UE games and synthetic benchmarks. Oodle says they have absolutely impeccable coding and there is absolutely zero errors in their code...yeah ok..I don't doubt the voltages are messing with the chips but to say you have errorless code is unbelievably arrogant.

 

[kiwigraeme]

A lot of you are a cup half empty , accentuate the positive , that's 50% that are still working ( well at least for now )

A lot of the server companies were running these undervolted, and they still had over 50% success rate

Spinning rust has competition, I still think Intel is short on the Intel

 

[HardReset]

Still kinda feels like a Unreal Engine problem. It's UE games and synthetic benchmarks. Oodle says they have absolutely impeccable coding and there is absolutely zero errors in their code...yeah ok..I don't doubt the voltages are messing with the chips but to say you have errorless code is unbelievably arrogant.

Code is probably not errorless. Still, it works with "fresh" CPU but that same code does not with "rot" CPU. That tells problem must be on hardware.

 

[Puiu]

50% failure rate in one year better spur more than this one complaint. I would expect at least 2 or 3 more companies to come forward in the coming days. Until then I can't get too excited about this. Kinda odd he didn't confidently know what mobos were being used offhand. More odd was how the conversation almost turns into a "AMD is mostly faster anyway." Then why didn't you buy AMD instead? I need to hear more.

Buying AMD for "servers" is usually not something smaller companies do since it usually requires retesting of all of their server software to ensure 100% compatibility and stability.

Time is money.

Unfortunately for Intel, they f-ed up hard. Losing the "we are the most stable solution" reputation will hurt them a lot. They were already losing in many performance metrics.