AMD's ACE in the hole - only game in town for VR

[antihelten]

With regards to the implementation of LiquidVR vs. GameWorks VR, I think the below interview is somewhat enlightening:

https://www.youtube.com/watch?v=asduqdRizqs&t=5m29s

Basically when Nate Mitchell (VP of product and co-founder of Oculus) is asked about Nvidias VR middleware (i.e. GameWorks VR), he is basically one big question mark, and then starts talking about both Nvidia and AMD in very general terms.

 

[Udgnim]

With regards to the implementation of LiquidVR vs. GameWorks VR, I think the below interview is somewhat enlightening:

https://www.youtube.com/watch?v=asduqdRizqs&t=5m29s

Basically when Nate Mitchell (VP of product and co-founder of Oculus) is asked about Nvidias VR middleware (i.e. GameWorks VR), he is basically one big question mark, and then starts talking about both Nvidia and AMD in very general terms.

my understanding of Liquid VR is that it's Mantle 2.0 and geared towards VR development

not very familiar with GameWorks VR but would not be surprised if it's similar to GameWorks in that it provides additional effects but geared towards VR

so AMD has an API geared towards VR while Nvidia does not and AMD might actually be ahead on the software front for once against Nvidia

then again, if Nvidia cards perform better in VR apps than AMD, it does not really matter unless AMD starts pulling proprietary maneuvers which I doubt they'll do

 

[geoxile]

my understanding of Liquid VR is that it's Mantle 2.0 and geared towards VR development

not very familiar with GameWorks VR but would not be surprised if it's similar to GameWorks in that it provides additional effects but geared towards VR

so AMD has an API geared towards VR while Nvidia does not and AMD might actually be ahead on the software front for once against Nvidia

then again, if Nvidia cards perform better in VR apps than AMD, it does not really matter unless AMD starts pulling proprietary maneuvers which I doubt they'll do

Most likely LiquidVR is not a new API, but rather a set of dev tools + library (like gameworks) implementing features for existing or upcoming APIs. I don't think I've seen anyone claim it's a Mantle 2.0.

 

[dacostafilipe]

I don't think I've seen anyone claim it's a Mantle 2.0.

We know it's Mantle, but if it's 2.0, 1.1, 1.* that's unknown ...

 

[VR Enthusiast]

I talk with them. They promoted LiquidVR on E3, I know it. They want you to play these games on AMD, because NVidia can't provide the same experience. But this doesn't mean that they won't support GameWorks VR. Even if it's worse than LiquidVR, it will help to create better experience compared to the base Oculus SDK.

You seem knowledgeable about VR - do you work in the industry?

I have heard that Crossfire is twice as fast as VR SLI currently and AMD also has half the latency compared to Nvidia. Can you comment on this?

Also this -

Both SDK can work with any API. Most VR games will be based on DX12 and Vulkan. Gameworks VR works with API specific extensions, while LiquidVR works with a VR optimized Mantle version.

From what I can tell GameWorks VR is (currently) Dx11 only?

 

[zlatan]

You seem knowledgeable about VR - do you work in the industry?

Yes. On a PS4 project.

I have heard that Crossfire is twice as fast as VR SLI currently and AMD also has half the latency compared to Nvidia. Can you comment on this?

LiquidVR using a different multi-GPU solution. AMD can do a more efficient approach with Mantle. VR SLI using NVAPI which is more like a GPU services library, so the ability to control the workload is much more limited. But VR SLI works in every aspect, just not as good as affinity multi-GPU.

From what I can tell GameWorks VR is (currently) Dx11 only?

Currently is, but NVAPI is a universal services library, so they can build support for any API.

 

[VR Enthusiast]

Zlatan, thanks for the info.

I had wondered how VR on the consoles was coming along. I know a lot of people will automatically assume that the PS4 and XBone aren't powerful enough for VR but I guess we'll find out. Obviously they both have access to ACEs through GCN.

Do you believe Nvidia when they say that Pascal (I assume) will be up to scratch here? Doesn't that seems unlikely given GPU development timescales? Even if so it's still probably 18 months+ distant...

It's difficult to believe how far behind Nvidia is on VR. Surely even AMD can't mess up this advantage?

 

[flopper]

B.
The timewarpis also different in LiquidVR. GameWorks VR works with draw-level preemption, which is very inefficient with long draws, because it will delay the context switch. AMD use a compute based pipeline, which is more efficient, and there is an extension for Radeon R9 285/380/Fury, because these GPUs can support fine-grained preemption. This is the ultimate solution for VR, and NVidia told me they will switch to fine-grained preemption when their new architecture is ready.

For the best possible VR experience you should buy a GPU with fine-grained preemption support.

Thanks.
if VR happens then I be ready for it.

 

[TheProgrammer]

Thanks zlatan, really useful info. Basically if you're spending $200 or over (cards fast enough that they'd support one eye in VR ~1440P so R9 380 and up), you want GCN.

Of course the newer GCN the better but per the thread title.. the only game in town if you want the best possible VR experience is GCN ACE.