zzyzx (zzyzx.delete@this.zzyzx.sh) on May 22, 2022 2:46 am wrote:
> Jukka Larja (roskakori2006.delete@this.gmail.com) on May 21, 2022 9:48 pm wrote:
> > Some time ago there was a "scandal" about a game not running on some (mostly old AMD) CPUs.
> > Turned out the game was using POPCNT, which according to Steam Hardware Survey was missing
> > from about 1-2 % of Steam users' CPUs at that time. It's rather surprising that Assassin's
> > Creed Odyssey even tried to require AVX. It would be interesting to know how they actually
> > fixed it. Did they just drop AVX altogether, ship two binaries or make a runtime choice?
>
> I don't know how they fixed it.
>
> The POPCNT one sounds familiar now that you mention it, and IIRC there was
> another game in the AC:Od timeframe that launched requiring AVX but made less
> of a fuss because the devs were clear from the start that they'd fix it.
>
> At this point (4 years later) I bet you could just about get away with requiring AVX for something with already
> high-ish sysreqs. Intel only very recently stopped fusing it off on Pentiums and Celerons, but I was surprised
> at the time of the AC:Od thing that hardly anyone even mentioned those (and by now the CPUs without AVX are
> weak enough that it's understandable). All of the concern was about Nehalem and Westmere.

You can get away with pretty much anything, if you just communicate the requirement clearly. AVX is currently in Steam Hardware Survey around 95 % supported (AVX2 is around 87 %, AVX-512 7.5 % by the way). Some of the remaining 5 % are dual-cores, some will have low-end GPU, so if the game requires strong CPU ("quad-core", in system requirements speak. Hard to be much more precise really) and GPU, it probably doesn't matter much if couple of percent of potential customers are lost.

It does raise the question, is AVX enough of a reason to lose any customers though? Some people will always play on a system with less than minimal requirements, if the game just works without crashing.

> > I admit that the newest generation of consoles are different, but the last time there was a big jump
> > in console CPU performance (PS3/Xbox360 to PS4/Xbox One), SIMD usage dropped. My understanding is that
> > the reason was that GPU performance and programmability increased even more (also: CPU SIMD performance
> > per core dropped, while core count and per core non-SIMD performance went up), which is again the case
> > with newest generation (this time the CPU SIMD performance per core has also gone up though).
>
> I haven't seen any confirmation on this, but if the PS5's FPUs are 128-bit as it looks on die
> shots, this gen should be another significant decrease in effective vector:scalar ratio (to whatever
> extent the PS5 is the lowest common denominator and determines how things are built).

Ah, I didn't even consider that. Good thing is I haven't actually read the docs about PS5 CPU. Wouldn't be able to speculate, if I had (due to NDAs and so) :D .

-JLarja