Re:Intel's “cripple AMD” function (2019)
by colejohnson66 on Saturday August 29, 2020 @09:25AM
(#24315899)
Why is it just the compiler maker’s job to report it may (read: will) underperform on AMD and not also the program developer too? If I paid for software that performed worse on AMD because it deliberately hobbled itself (and was not informed), I’d want a refund.
It’s straight up anti-competitive, but consumers aren’t smart enough to understand that it’s a problem; A consumer just sees biased benchmarks that show Intel outperforming AMD, and then choose Intel.
It's a bit worse than that. Intel has no obligation to support optimizations that aren't unique to AMD; they're allowed to disable SIMD extensions that AMD processors declare support for, while at the same time using all of those SIMD extensions on Intel CPUs. They just have to include the disclaimer that their compiler and libraries may be doing this.