Crashes in DecompressShadergroupWithOodleAndExtraLogging

Hi,

we’re seeing a surge of PC-only crashes in DecompressShadergroupWithOodleAndExtraLogging. We end up with a message like:

LowLevelFatalError [File:.\\Runtime/RenderCore/Private/ShaderCodeArchive.cpp] [Line: 413] DecompressShaderWithOodleAndExtraLogging(): Could not decompress shader group with Oodle. Group Index: 477 Group IoStoreHash:ffba225bcd2c82a6697cb909 Group NumShaders: 48 Shader Index: 4553 Shader In-group Index: 477 Shader Hash: B027FC7291957FA03DAD133B22EB62034CBEC4CA. The CPU (13th Gen Intel(R) Core™ i7-13700KF) may be unstable; for details see Intel Processor Instability Causing Oodle Decompression Failures

We see it on a plethora of GPUs, crashing on different shader indices, no common denominator but just a few CPUs, also reported by the OODLE message itself.

So far:

  • Intel(R) Core™ i7-13700K
  • Intel(R) Core™ i7-13700KF
  • Intel(R) Core™ i9-13900K
  • Intel(R) Core™ i9-13900KF
  • Intel(R) Core™ i7-14700K
  • Intel(R) Core™ i9-14900KF
  • (non exhaustive)

Anyone ran into it too? Perhaps detect the exact revision of the CPU and warn the users to upgrade?

Cheers all!

Hi Pavel,

This one is tricky to handle. The errors come from faulty hardware and can be intermittent. Intel has shared some details on the cause of the problem in an article that is linked by our post on the Oodle website. https://community.intel.com/t5/Blogs/Tech\-Innovation/Client/Intel\-Core\-13th\-and\-14th\-Gen\-Desktop\-Instability\-Root\-Cause/post/1633239

As stated by Intel, the problem comes from “reliability aging under elevated voltage and temperature”. From what we understand, this damage is irreversible and the affected CPUs must be replaced. Intel has pushed a microcode update to the motherboard manufacturers and users must update their BIOS to protect their CPU from being affected by the problem.

We are not aware of a way to detect if a CPU is damaged other than hitting a hardware fault. Oodle is just a red herring here as it has some code that detects the decompression did not complete successfully.

Regards,

Martin

Thanks Martin, yeah, I feared that. We’ll communicate it to the users via support then.