Intel's “Cripple AMD” Function | Hacker News Reader

Intel's “Cripple AMD” Function

agner.org

95 pointsjigglypuffs10 comments

It's interesting how this is still a major issue today, Last month I learned about the Matlab AMD crippling issue. With the workaround AMD performance on a 3960x increased by 30%.

"[...] the Intel MKL uses a discriminative CPU Dispatcher that does not use efficient codepath according to SIMD support by the CPU, but based on the result of a vendor string query. If the CPU is from AMD, the MKL does not use SSE3-SSE4 or AVX1/2 extensions but falls back to SSE1 no matter whether the AMD CPU supports more efficient SIMD extensions like AVX2 or not."

https://www.extremetech.com/computing/302650-how-to-bypass-m...

https://www.reddit.com/r/Amd/comments/e24jqp/extremetech_how...

Wasn’t there a lawsuit about the ICC doing this a few years ago?

For these kind of stuff it was particularly shocking (or telling of the raw power Zen 2 achieves) that AMD Epycs surpassed Intel in one of their own benchmarks this year.

"Windows ain't done 'til Lotus won't run."

Not sure if it's still the case... would be interesting to see a post-compiler tool that bypasses the GenuineIntel checks for certain cases.

Intel is still doing it [1], at least in some of their libraries.

[1] https://www.extremetech.com/computing/302650-how-to-bypass-m...

This is a message board thread from 2009 FYI.

There is a recent post discussing the idea that AMD Phenom processors are unfairly disadvantaged because they do not support SSE3 and modern games are targeting SSE3/4.2. I think that's what this post it trying to point out. I really don't see any modern relevance to this post though since modern AMD processors support these instructions.

I'd still be curious to know if Intel is doing this though.

I guess there are some other Ryzen related conversations too that are relevant.

> modern AMD processors support these instructions

... and that's the whole point of the post. AMD CPUs support the instructions, but Intel's compiler ignores that support on non-Intel processors.

There is a useful recent note, toward the end, about an environment variable that, apparently, may be set when running a program built with Intel's compiler, and that will enable the program to run at full speed on an AMD cpu.

Reply on news.ycombinator.com