AMD Polaris architecture – GCN 4.0

Rroff · 21 May 2016 at 14:50

Yeah will be interesting to see what happens again at the bigger end.

CAT-THE-FIFTH · 21 May 2016 at 14:56

queamin said:
Depends how cut down pascal get at the mid range which will go against polaris, Vega will be high end and that different from all accounts so it might handle AS differently.

The GP106 is around 200mm2 with a 256 bit memory controller.

Also have people found out if Polaris supports conservative rasterisation,etc?

I mean if Nvidia can use that for certain effects it will blunt any AMD advantage might have for async.

Ferrari100 · 21 May 2016 at 15:44

Looks like Total War Warhammer will have similar Async benefits to Ashes of the Singularity. Don't forget, just because it is a DX12 game does not mean Async compute is being used.
http://www.pcworld.com/article/3073...erformance-preview-radeon-reigns-supreme.html

This never ends well · 21 May 2016 at 15:58

^^^^ Whats the point in posting that when their no 980ti or fury's.

Or they didnt want to show the fury x spanking the 1080.

Ferrari100 · 21 May 2016 at 16:04

jamiehard1989 said:
^^^^ Whats the point in posting that when their no 980ti or fury's.

Or they didnt want to show the fury x spanking the 1080.

I guess they never had a Fury X to bench, that's all.
I would imagine performance will fall around 10-12% behind the 1080 like in AOTS?.

TheRealDeal · 21 May 2016 at 16:05

There is only 29% between a 390x and 1080 at 1440p. The fury X should be pretty close with the 1080.

The 1080 pulls up to 44% at 4k stretching it's legs some.

queamin · 21 May 2016 at 16:13

Don't let Cat see the 960

, but in all honestly it is early yet.

Rroff · 21 May 2016 at 16:14

^^ They are different resolutions - its a somewhat incomplete set of benchmarks with mis-matched cards to say the least but I guess that was all they had to work with - in which case making any kind of decisive summary is kind of premature.

TheRealDeal · 21 May 2016 at 16:16

Rroff said:
^^ They are different resolutions - its a somewhat incomplete set of benchmarks with mis-matched cards to say the least but I guess that was all they had to work with - in which case making any kind of decisive summary is kind of premature.

Yea the color code is at the top though so you can compare the results to the cards using the same colors.

D.P. · 21 May 2016 at 16:26

Do we know if the nvidia cards, specifically 1080, have a code path with appropriate async implementation? The AOTS bechamrks are invalid for testing pascal async performance because the developer hasn't had a chance to develop a pascal async code path yet. Since you can't even buy the cards.

queamin · 21 May 2016 at 16:29

D.P. said:
Do we know if the nvidia cards, specifically 1080, have a code path with appropriate async implementation? The AOTS bechamrks are invalid for testing pascal async performance because the developer hasn't had a chance to develop a pascal async code path yet. Since you can't even buy the cards.

Not sure but it is doing well without it really, it depends on Vega and ti, titan how it stands up to them, as will it come close to them or left behind a lot.

Lokken86 · 21 May 2016 at 16:29

The fury might pull ahead at 1440p in that game. Good stuff. It will best the 980ti for sure.

Ferrari100 · 21 May 2016 at 16:31

D.P. said:
Do we know if the nvidia cards, specifically 1080, have a code path with appropriate async implementation? The AOTS bechamrks are invalid for testing pascal async performance because the developer hasn't had a chance to develop a pascal async code path yet. Since you can't even buy the cards.

It's not that they have not developed a path. It's due to Nvidia requesting it be switched off. There is nothing stopping Nvidia asking it to be turned on. Nvidia originally requested to the developer Oxide that they not use Asynchronous compute at all. This request was not adhered to.

flopper · 21 May 2016 at 16:34

Ferrari100 said:
It's not that they have not developed a path. It's due to Nvidia requesting it be switched off. There is nothing stopping Nvidia asking it to be turned on. Nvidia originally requested to the developer Oxide that they not use Asynchronous compute at all. This request was not adhered to.

crytek says it best,
Nxxxxa holding back async compute and PC gaming atm.

"the relative immaturity of the hardware that supports it, where many GPUs come with sometimes problematic restrictions on pre-emption granularity, etc."
http://www.dsogaming.com/interviews...-tracing-physically-based-rendering-e3-demos/

LoadsaMoney · 21 May 2016 at 19:57

Crytek on Async Compute: Interesting, But It Adds Non-Trivial Complexity; DX12 Support To Improve With Time

http://wccftech.com/crytek-async-compute-interesting-adds-nontrivial-complexity-dx12-support-improve-time/

Calin Banc · 22 May 2016 at 08:31

D.P. said:
Do we know if the nvidia cards, specifically 1080, have a code path with appropriate async implementation? The AOTS bechamrks are invalid for testing pascal async performance because the developer hasn't had a chance to develop a pascal async code path yet. Since you can't even buy the cards.

I think they said in a post a while back that there is no specific code for async, just that nvidia has it turned off by default. I think you can still turn it on through a .ini config or at least that's what I've heard. Even if they had some special code for async, it looks like Pascal doesn't truly support it (it just doesn't take a performance hit while doing it), so it's rather useless anyway.

At this point I'd say nothing is really "trivial" in dx12 and most devs won't say much about which GPU is limiting or supporting what (no matter if it is AMD or nVIDIA), due to obvious reasons.

AlamoX · 22 May 2016 at 08:39

GFX-Kid · 22 May 2016 at 10:26

With all nvidia's recent antics, I am left waiting for Polaris.

It has been said that Polaris shines bright and it is a star that will guide us to Vega!

fs123 · 22 May 2016 at 11:43

Calin Banc said:
I think they said in a post a while back that there is no specific code for async, just that nvidia has it turned off by default. I think you can still turn it on through a .ini config or at least that's what I've heard. Even if they had some special code for async, it looks like Pascal doesn't truly support it (it just doesn't take a performance hit while doing it), so it's rather useless anyway.

At this point I'd say nothing is really "trivial" in dx12 and most devs won't say much about which GPU is limiting or supporting what (no matter if it is AMD or nVIDIA), due to obvious reasons.

Originally Oxide said the Nvidia drivers were advertising that Async was supported but when the game code executed the async codepath, there were severe performance problems.
Rather dubious on Nvidia's part to fake Async support in this way since drivers can be made to show support for any feature even if it's not actually supported at all. Programs like GPU-Z, Hardwareinfo, etc would show that card as supporting a feature by interrogating the driver only.

We will find out soon enough if the 1080 really does support it or it's the same again.

old gamer · 22 May 2016 at 11:58

GFX-Kid said:
With all nvidia's recent antics, I am left waiting for Polaris.

It has been said that Polaris shines bright and it is a star that will guide us to Vega!

I think your ocuk profile has been hijacked by flopper