• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Question Zen 6 Speculation Thread

Page 405 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Well compiler can use it for you like here https://godbolt.org/z/fT1WYfKvz (it's trivial example, but here we are summing 72 floats, there are 16 floats in 512b reg, so it compiles down to 4 512b ops and one 256b op (ymm are 256b regs, zmm are 512b ones).

So knowing compiler can use them for you and/or you can explicitly use them (for example for tail handling, small transpose or gemm kernels etc) it's really hard to give you more detailed answer than: "For sure".

Okay, thanks for presenting at least one example. My own experience with anything beyond AVX2 is pretty limited (since I was on a 3900X for so long), but I do remember ycruncher splitting their executable into a number of different executables targeting different platforms based at least in part on what ISA extensions were supported by the target hardware.

Regardless I don't recall there being a rash of software from (for example) the Rocket Lake or Raphael generation onward utilizing AVX-512 exclusively for all vector lengths. Probably because there's still so much hardware out there that does not support AVX-512 at all, including multiple Intel CPUs released after Rocket Lake.

Eventually it will make sense to target AVX-512 or AVX10 once enough recent generations of consumer hardware support one of those standards.

Your understanding here is still incorrect. Hybrid architectures aren't any easier with AVX10 than AVX512.

Could you elaborate on that?
 
How many units over say 5 year period?
That's hard to guess in advance, but it'll certainly be many millions.

Because what they're imo doing here is actually competing with mainstream OEM and DIY gaming PCs, not PS6.
And most of those gaming PCs they're competing with will be either slower, more VRAM-limited, or much more expensive.

Think about it: Even at a price of $999, $1099 or even $1199, they'd still seriously undercut most new gaming PCs of similar performance in price.

If you can get an entire Xbox PC for the price of a single 5080/6080, a lot of people will be going XbPC.
I mean, good luck getting a decent mobo, PSU, NVMe SSD, 12C Zen6, 32GB RAM and 5080-class GPU (and a legal Win11 license) for about $1K total in 2027.
Nevermind that this Xbox PC will have the advantage of a shared memory pool, so the GPU might have more than 16-18 GB at its disposal in games that need it.
With Xbox PC, MS can also ask a much higher price for the same hardware than they could for a "pure" console, and still be hella price-competitive vs. "normal" PCs.
In console-only space, they'd have a hard time against PS6, due to lack of exclusives and lack of graphics edge. But in PC space, they might be competing against PCs with 8C + 8-12GB gfx cards on price, yet against 12C + 5080+ on performance.

In fact, that might be one of the reasons AMD killed AT1:
They knew of MS' plans, and knew almost nobody would buy a 48WGP AMD dGPU at $1k+ anymore if - in addition to potentially strong competition from 6070Ti/6080 - people could get an entire XboxPC with a 34WGP GPU and large shared memory pool for the same money.
 
Back
Top