Re:Intel AVX10: The Successor to AVX-512
by shmerl on Monday July 24, 2023 @06:27PM
(#36855128)
Why did it take around 10 years for AMD to implement AVX-512 and will they need to wait as long for this too? Doesn't seem to be patent related (patents are 20 years and AVX-512 was introduced in 2013?).
[ Parent
| on HN ]
Don’t Intel and AMD have cross licensing agreements for instruction set extensions?
That's what I thought too. But somehow AVX-512 took a long time for AMD.
Re:Intel AVX10: The Successor to AVX-512
by colejohnson66 on Monday July 24, 2023 @06:50PM
(#36855341)
Implementing an instruction set isn’t as easy as “we have access to IP”
Because Intel implementations of AVX-512 were pretty much unusable or not worth using until 2019-2020, so nobody was writing much AVX-512, so they didn't need to.
AMD practically resurrected AVX-512 by supporting it across both desktop and server products, so it's weird to criticize them for taking a while when Intel nearly killed the thing via segmentation and poor implementation.
I'm not criticizing it, just wondering why it took that long and if it was due to patents or not.
AMD probably decided that their transistor budget was better spent elsewhere.
If I were to guess:
AVX512 vector and mask register take space on the die, which is a finite resource and could be allocated to other things.
Moreover, power consumption may have also been a challenge, as seen in the early Intel implementations of AVX512 (especially Skylake-X).
And while AVX512 can improve single-threaded performance, that's only for applications that make use of it. And up until recently, that number was extremely limited.
I'm guessing the reason AVX512 took so long is that doubling the size increases complexity by a lot. This doesn't introduce wider vectors, so I don't see why it would take so long.
But who knows if AMD sees value in these new instructions/if there's market demand for them.