Ashes of the Singularity Revisited: A Beta Look at DirectX 12 & Asynchronous Shading
by Daniel Williams & Ryan Smith on February 24, 2016 1:00 PM ESTClosing Thoughts
Wrapping up our second look at Ashes of the Singularity and third overall look at Oxide’s Nitrous engines, it’s interesting to see where things have changed and where they have stayed the same.
Thanks to the general performance optimizations made since our initial look at Ashes, the situation for multi-GPU via DirectX 12 explicit multi-adapter is both very different and very similar. On an absolute basis it’s now a lot harder to max out a multi-GPU configuration; with reasonable quality settings we’re CPU limited even up to 4K, requiring we further increase the rendering quality. This more than anything else handily illustrates just how much performance has improved since the last beta. On the other hand it’s still the most unusual pairing – a Radeon R9 Fury X with a GeForce GTX 980 Ti – that delivers the best multi-GPU performance, which just goes to show what RTG and NVIDIA can accomplish working together.
As for the single GPU configurations, I’m not sure things as they currently stand could be any more different. NVIDIA cards have very good baseline DX11 performance in Ashes of the Singularity, but they mostly gain nothing from Ashes’ DX12 rendering path. RTG cards on the other hand have poorer DX11 performance, but they gain a significant amount of performance from the DX12 rendering path. In fact they gain so much performance that against traditional competitive lineups (e.g. Fury X vs. 980 Ti), the RTG cards are well in the lead, which isn’t usually the case elsewhere.
Going hand-in-hand with DX12, RTG’s cards are the only products to consistently benefit from Ashes’ improved asynchronous shading implementation. Whereas our NVIDIA cards see a very slight regression (with NVIDIA telling us that async shading is not currently enabled in their drivers), the Radeons improve in performance, especially the top-tier Fury X. This by itself isn’t wholly surprising given some of our theories about Fury X’s strengths and weaknesses, but for Ashes of the Singularity performance it further compounds on the other DX12 performance gains for RTG.
Ultimately Ashes gives us a very interesting look at the state of DirectX 12 performance for both RTG and NVIDIA cards, though no more and no less. As we stated at the start of this article this is beta software and performance is subject to change – not to mention the overall sample size of one game – but it is a start. For RTG this certainly lends support to their promotion of and expectations for DirectX 12, and it should be interesting to see how things shape up in March and beyond once the gold version of Ashes is released, and past that even more DirectX 12 games.
Facebook
Twitter
RSS
153 Comments
View All Comments
Beany2013 - Wednesday, February 24, 2016 - link
You are aware that Mantle and DX12 are actually different APIs, yeah?zheega - Wednesday, February 24, 2016 - link
AMD just released new drivers that say are made for this benchmark. Can we get a quick follow-up if their performance improves even more??http://support.amd.com/en-us/kb-articles/Pages/AMD...
AMD has partnered with Stardock in association with Oxide to bring gamers Ashes of the Singularity – Benchmark 2.0 the first benchmark to release with DirectX® 12 benchmarking capabilities such as Asynchronous Compute, multi-GPU and multi-threaded command buffer Re-ordering. Radeon Software Crimson Edition 16.2 is optimized to support this exciting new release.
revanchrist - Wednesday, February 24, 2016 - link
See? Every time when there's a pro AMD game tested, there'll be much butt hurt fanboy comments. And i guess everyone knows why. Because when you bought something, you'll always want to justified your purchase and you know who's got the lion share of the dGPU market now. Guess nowadays people are just too sensitive or has a heart of glasses, which makes them judging things ever so subjectively and personally.Socius - Wednesday, February 24, 2016 - link
For anyone who missed it:"Update 02/24: NVIDIA sent a note over this afternoon letting us know that asynchornous shading is not enabled in their current drivers, hence the performance we are seeing here. Unfortunately they are not providing an ETA for when this feature will be enabled."
ToTTenTranz - Wednesday, February 24, 2016 - link
"Unfortunately they are not providing an ETA for when this feature will be enabled."If ever...
andrewaggb - Wednesday, February 24, 2016 - link
Makes sense why it would be slightly slower. Also makes through benchmarks less meaningfulExt3h - Wednesday, February 24, 2016 - link
"not enabled" is a strange and misleading wording, since it obviously is both available and working correctly according to the specification.Should be read as "not being made full use of", as it is only lacking any clever way of profiting from asynchronous compute in hardware.
barn25 - Thursday, February 25, 2016 - link
If you google around you will find out nvidia does not have asynchornous shading on its DX"12" cards. this was actually first found out in WDDM 1.3 back in windows 8.1 when they would not support the optional features which AMD does.Ext3h - Thursday, February 25, 2016 - link
I know that the wrong terminology kept being used for years now, especially driven by major tech review websites like this one. But that's still not making it any less wrong.The API is fully functional. So the driver does support it. Whether it does so efficiently is an entirely different matter, you don't NEED hardware "support" to provide that feature. Hardware support is only required to provide parallel execution, as opposed to the default sequential fallback. The latter one is perfectly within the bounds in the specification, and counts as fully functional. It's just not providing any additional benefits, but it's neither broken nor deactivated.
barn25 - Thursday, February 25, 2016 - link
Don't try to change it. I am referring to HW Asyc compute, which AMD supports and NVidia does not. Using a shim will impact performance even greater.