PCTuning fórum

Ache píše:

webwalker píše:http://www.dsogaming.com/news/oxide-dev ... benchmark/

Každý přece předem musel vědět jak to dopadne...

Personally, I think one could just as easily make the claim that we were biased toward NVIDIA as the only "vendor" specific-code is for NVIDIA where we had to shutdown async compute. By vendor specific, I mean a case where we look at the Vendor ID and make changes to our rendering path. Curiously, their driver reported this feature was functional but attempting to use it was an unmitigated disaster in terms of performance and conformance so we shut it down on their hardware. As far as I know, Maxwell doesn't really have Async Compute so I don't know why their driver was trying to expose that.

Given its growing market-share, NVIDIA could use similar tactics to keep game developers away from industry-standard API features that it doesn't support, and which rival AMD does. NVIDIA drivers tell Windows that its GPUs support DirectX 12 feature-level 12_1. We wonder how much of that support is faked at the driver-level, like async compute. The company is already drawling flack for using borderline anti-competitive practices with GameWorks, which effectively creates a walled garden of visual effects that only users of NVIDIA hardware can experience for the same $59 everyone spends on a particular game.

Our use of Async Compute, however, pales with comparisons to some of the things which the console guys are starting to do. Most of those haven't made their way to the PC yet, but I've heard of developers getting 30% GPU performance by using Async Compute. Too early to tell, of course, but it could end being pretty disruptive in a year or so as these GCN built and optimized engines start coming to the PC. I don't think Unreal titles will show this very much though, so likely we'll have to wait to see. Has anyone profiled Ark yet?

Krteq píše:Ale on mluvil převážně o DX12.1 a podpoře různých tierů. To že Maxwell 2.0 nepodporuje jednu ze zásadních funkcí (A/C) se tam, možná naštěstí, nezmiňuje

CageJ píše:A/S bude velka vyhoda sucasnych Radeonov.. uvidime ci ju dokazu aj vyuzit... v texte pise, ze narast vykonu je aj 30%! ...

On the Maxwell architecture, up to 31 simultaneous command lists (the limit of Maxwell in graphics/compute workload) run at nearly the exact same speed - indicating Async Shader capability. Every 32 lists added would cause increasing render times, indicating the scheduler was being overloaded.
On the GCN architecture, 128 simultaneous command lists ran roughly the same, with very minor increased speeds past 64 command lists (GCN's limit) - indicating Async Shader capability. This shows the strength of AMD's ACE architecture and their scheduler.

Interestingly enough, the GTX 960 ended up having higher compute capability in this homebrew benchmark than both the R9 390x and the Fury X - but only when it was under 31 simultaneous command lists. The 980 TI had double the compute performance of either, yet only below 31 command lists. It performed roughly equal to the Fury X at up to 128 command lists.

Furthermore, the new beta of GameworksVR has real results showing nearly halved render times in SLI, even on the old GTX 680. 980's are reportedly lag-free now.

Maxwell is capable of Async compute (and Async Shaders), and is actually faster when it can stay within its work order limit (1+31 queues). Though, it evens out with GCN parts toward 96-128 simultaneous command lists (3-4 work order loads). Additionally, it exposes how differently Async Shaders can perform on either architecture due to how they're compiled.