Polaris 12 Support Being Sent In To Linux 4.10 Kernel

[phoronix]

Phoronix: Polaris 12 Support Being Sent In To Linux 4.10 Kernel

AMD is looking to land initial support for upcoming "Polaris 12" graphics processors into the in-development Linux 4.10 kernel...

http://www.phoronix.com/scan.php?page=news_item&px=Radeon-Polaris-12-4.10-Fixes

[davidbepo]

I wonder how Polaris 12 achieves the (supposed) 2x instructions per second compared to Polaris10/11. Is it just "marketing speech" which actually means the addition of compute units and running the GPU at higher frequency, or is it an actual increase of IPC?

that was vega afaik

[ResponseWriter]

I wonder how Polaris 12 achieves the (supposed) 2x instructions per second compared to Polaris10/11. Is it just "marketing speech" which actually means the addition of compute units and running the GPU at higher frequency, or is it an actual increase of IPC?

That's Vega. There's a bit about it here: http://www.anandtech.com/show/11002/...cture-teaser/2

Knowing what marketing people can be like, they could be referring to the doubling of FP16 instructions per clock, which would be more useful to certain compute workloads than it would for gaming. Still, gamers are only part of the market for these chips and there are plenty of improvements for everyone.

[davidbepo]

Ah, yes. There is little information about Polaris 12.

yes but the little is known leads me to think that is an improved polaris11 with more clock speed and less power draw

[rabcor]

gieb vega support plz

[smitty3268]

That's Vega. There's a bit about it here: http://www.anandtech.com/show/11002/...cture-teaser/2

Knowing what marketing people can be like, they could be referring to the doubling of FP16 instructions per clock, which would be more useful to certain compute workloads than it would for gaming. Still, gamers are only part of the market for these chips and there are plenty of improvements for everyone.

If someone saw 2x instructions per clock, that's almost certainly regarding the fp16 speed.

However, that link has a slide that mentions 2x geometry perf which is probably what atom had seen. It's doubtful that will make a big difference in most games, but it might in a couple here and there. Most games are more fragment shader bound than geometry bound.

This in turn is a piece of the bigger picture when looking at the next improvement in Vega, which is AMD’s geometry pipeline. Overall AMD is promising a better than 2x improvement in peak geometry throughput per clock. Broadly speaking, AMD’s geometry performance in recent generations hasn’t been poor (it’s one of the areas where Polaris even further improved), but it has also hurt them at times. So this is potentially important for removing a bottleneck to squeezing more out of GCN.

And while AMD's presentation and comments itself don't go into detail on how they achieved this increase in throughput, buried in the footnote for AMD's slide deck is this nugget: "Vega is designed to handle up to 11 polygons per clock with 4 geometry engines." So this clearly reinforces the idea that the overall geometry performance improvement in Vega comes from improving the throughput of the individual geometry engines, as opposed to simply adding more as the scalability improvements presumably allow. This is one area where Vega’s teaser paints a tantalizing view of future performance, but in the process raises further questions on just how AMD is doing it.

[pal666]

Yes. After browsing the Internet for more Vega details my conclusion is that FP32 IPC is equivalent to previous GPUs.

pretty dumb statement in context of videocards. vega will have much higher ipc because it will have much more compute units