Jump to content

Recommended Posts

On 6/17/2020 at 4:20 AM, ACG_crane said:

VRSS is showing "Not supported for this aplication". Any clue why?

VRSS requires developer support - sort of - which is why it is defaulted off for all titles and then explicitly enabled title-by-title by Nvidia.

The reason is that it makes use of a title's built in support for variable rate shading - and forward rendering. It needs this support to increase the shading rate within a central region of the display to increase the degree of AA super-sampling within that region and to dynamically vary the size of the "fovea" region in response to GPU headroom changes. For those that have no idea what VRSS is - it is Variable Rate Super Sampling. You must enable MSAA for it to work at all. If you then set MSAA to say 4x on a VRSS supported title the driver will selectively  supersample the central region of the view at 4x. If MSAA is 8x then it will supersample at 8x. The radius of the supersampled region grows and shrinks in response to GPU headroom changes.

It is up to the developer to submit their title to Nvidia for whitelisting in the driver, but 1CGS have not - probably due to their use of deferred shading.

 

Quoting the Nvidia dev blog announcing VRSS back in January 2020:

Quote

VRSS can be applied to all VR games and applications that are DX11, Forward Rendered and have MSAA.  The VRSS functionality resides in the NVIDIA Driver.  The best part is that developers do not have to write any code to integrate VRSS, they just need to submit their VR game or application to NVIDIA for VRSS testing. NVIDIA will test the VR game or application, and if it benefits from VRSS, then it will be considered for whitelisting in the NVIDIA Driver.

 

It's interesting that the introduction of FXAA coincided with the change to deferred shading and that most people's complaints about the shimmering and aliasing began with 4.006. Because FXAA is a common workaround for the negative impact deferred shading has on antialiasing - it kinda breaks it. But deferred shading itself is a workaround for the exponential growth in GPU load caused by dynamic lighting - which is what delivers the post 4.006 lighting effects such as realistic aircraft skin reflections, canopy reflections and instrument glass. I suspect that FXAA was added rather than just as a "faster anti-aliasing method" (relative to existing MSAA performance) to some extent as a workaround for the impact deferred shading would have had on MSAA.

 

I like the new lighting made possible on current hardware by deferred shading but I'm personally frustrated by the fact that even on an overclocked RTX3090 I get a HMD full of jaggies that even 8x MSAA doesn't remove.

 

Quote

The main advantage of FXAA over conventional spatial anti-aliasing is that it does not require large amounts of computing power. It achieves this by smoothing undesirable jagged edges ("jaggies")[4] as pixels, according to how they appear on-screen, rather than analyzing the 3D model itself, as in conventional spatial anti-aliasing.[1] Since it is not based on the actual geometry, it will smooth not only edges between triangles, but also edges inside alpha-blended textures, or those resulting from pixel shader effects, which are immune to the effects of multisample anti-aliasing (MSAA).[5]

The downsides are that high contrast texture maps are blurred; FXAA must be applied before rendering the HUD elements of a game lest it affect them too; and that polygonal details that are smaller than one pixel that would have been captured and rendered by MSAA and SSAA will not be captured and rendered by FXAA alone.

 

So FXAA can help with the texture aliasing that I think is the cause of the shimmering landscape but it does it by blurring everything kinda like a vaseline filter and definitely hurts spotting.

Edited by Dave
Link to post
Share on other sites
1 hour ago, Dave said:

 

SSAO is quite costly - it essentially uses a simplified ray-tracing algorithm to approximate global illumination (indirect bounce lighting from nearby surfaces) to improve the realism of shadowed regions which might otherwise look unnaturally dark and flat. In IL2 I personally think this adds almost nothing to the visuals in VR while costing precious FPS. Try turning this off and if you cant tell the difference you can enjoy the FPS boost without feeling like you are missing anything.

 

Not as been tested by @SCG_Fenris_Wolf  

 

Link to post
Share on other sites
49 minutes ago, Dutch2 said:

 

Not as been tested by @SCG_Fenris_Wolf  

 

Respectfully, I think Fenris' suggested settings are wrong. I am getting <9ms/frame with my Reverb on a 3090 with much better visuals than I did after trying Fenris' suggestions. Following his advice the image suffered and so did FPS.

Rather than just contradict though, I made a point of explaining why certain options are better or worse so people can consider the rationale and make informed choices. Good on Fenris for doing his own testing and sharing - I just don't agree with some of his results.

 

It is also worth noting that what works well for one hardware combination will not necessarily work well for another - so take any advice that isn't supported by explanation of the underlying features with a grain of salt.

Edited by Dave
The content rather than the writer
  • Like 1
Link to post
Share on other sites
9 hours ago, Dutch2 said:
11 hours ago, Dave said:

 

SSAO is quite costly - it essentially uses a simplified ray-tracing algorithm to approximate global illumination (indirect bounce lighting from nearby surfaces) to improve the realism of shadowed regions which might otherwise look unnaturally dark and flat. In IL2 I personally think this adds almost nothing to the visuals in VR while costing precious FPS. Try turning this off and if you cant tell the difference you can enjoy the FPS boost without feeling like you are missing anything.

 

Not as been tested by @SCG_Fenris_Wolf  

I went back and edited that just now because the way I had written it didn't make clear the comparison was with no AO. My personal observation of this setting and the performance penalty stand. As our other settings differ quite a bit there is the chance enabling SSAO for him is interacting favourably with another option he has chosen which I have not.

For example, if the Ultra base profile enables ambient occlusion (which I can't know because the exact content of each of those profiles is completely undocumented) then adding SSAO on top of Ultra will improve performance. If the High base profile, which I use, does not provide AO then turning on SSAO combined with High will decrease performance. It would be helpful to know exactly what each of the base graphics profiles set - as it stands we are just using trial and error combined with assumptions. It is very common for GPU driver configuration to expose coarse grained settings (eg AA mode) which you then refine with finer grained settings (eg 2x, 4x, multi-frame interpolated). It is also common for pairs of options to work against one another (eg negative LOD bias with anisotropic filtering) so it would be extremely helpful to know what the starting state is before tweaking fine-grained options.

Edited by Dave
  • Upvote 1
Link to post
Share on other sites
On 1/3/2021 at 7:20 PM, Dave said:

I am getting <9ms/frame with my Reverb on a 3090 with much better visuals than I did after trying Fenris' suggestions

 

Mind sharing your settings? I am also struggling with G2/3090 and surprised how low the performance is...

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...