VR Benchmarking Results, using FCAT from nVidia

@MarcG just like to say thank you, because if you hadn't done this thread and posted your finding regarding SVR SS I would have never thought to investigate. The difference for me is night and day, both with graphical detail and fidelity.

I wish R3E would have played nicely but I've had to do a fresh install of the game, let it rebuild my save & updated to the latest GPU drivers to get the performance boost i had working last night. I'm expecting to boot R3E tomorrow and find that I have regressed again to a stuttering mess, but for now your finding have changed both these games for me, so thank you.

 

I Liked this, I think I like it? What is "SVR SS"?

Thanks!

 

Steam VR Super Sampling

EDIT:
@vegaguy5555 the reason why it has changed so drastically for me is that my Steam VR had defaulted to 2.0 SS after I had changed a setting in the steamvr config file many months ago for a game I play (iL-2), prior to SS being a slider on the SteamVR settings menu. This was really dropping my performance in rf2 combined with using a 1.6PD in OTT.

I now run:
PD @ 1.6
AA @ 2
PP - low
AFx16
All Details = FULL
Shadows = Low
Soft Particles & all Reflections = Low
AI = 23 (all visible)

With these settings and testing a variety of mods (GP3, Clio, Skip, GT500, USF, T5) around a few tracks (Silverstone, Portland, Toban, Croft) I am able to get to a steady 90fps within 2 laps of starting a race at the back of the grid once the pack spreads. Tbh there might be drops here and there as I have given up using the performance HUD overlay in the last day as everything was smooth and I was happy. But i don't see any graphical warping/inconsistencies once the pack spreads. There are a couple of tracks that I have issues with passing the pits/start/finish (Oulton park) but i had the same issues running Multiview in triples.

 

I am actually fairly impressed with rF2's VR. I had some blurring when going around corners with everything maxed out and then noticed I forgot to reset Oculus DBT which was SS at 1.7 for Dirt Rally.

It is almost livable at 1.7. Tried turning everything down but it didn't seem to help much.

AC is looking kind of burry lately. Has it gotten worse or has everything else just getting better?

 

I find this hard to believe. I have a 1080Ti and I run it at SS@1.4, AA Level2, PP off and medium details, including medium shadows to get stable 90fps (I disabled all forms reprojection so no ASW-type cheating). Occasionally selecting an unoptimized or too detailed car will drop me below 90fps.

 

I'm no making things up. The things I have done:
Taken HT option off in BIOS
The Controller.json file to read:
"Use thread":true,
"Use thread#":"Use a separate thread to issue FFB commands which may block with some drivers"
I'm Steam VR Supersamplimg level is at its lowest and the tick box unchecked.
I've not changed anything in Nvidia control panel apart from prefer max performance.

Things really changed for me when I took down Steam vr settings to 0.65.

 

I don't think I'll be using my triples much anymore!!

I would go as far as saying rF2 has the clearest VR now.

There is still some jumping. I tried my triples again and there is jumping there too, always was now that I think of it. Not as noticeable without VR.

Does anyone else experience this jumping?

PS
Thank you, thank you for not having auto exposure Studio 397.

 

Ah, here we go. So you've conveniently left out the most important bit from your original claim. Running SteamVR supersampling at 0.65 means your net supersampling is PD @ 1.6 x SS @ 0.65 = 1.04.
Meaning, you're not running any supersampling at all.

So essentially, you could have just left both PD and SS at default.

 

@Milopapa
I didn't leave anything out if you go back and read my posts. We did not know how steamvr interacted with OTT's PD and this was stumbled on by Marc benchmarks. But thank you for explaining to me how these 2 interact!

 

@Milopapa
A question, does the Steam VR SS value still come into your equation if the 'Enable Advanced SS Filtering' is unchecked?

I ask because I have this unchecked/checked in tests I ran a few posts back.
Logically unchecked would mean a way to bypass Steam VR's SS (??)

 

This is just a post-processing filter and since many people didn't like it in the beta, they now offer it as an option. Should have a small sharpening/blurring effect - should not impact the supersampling itself. You enable that by setting a multiplier different from 1.0.

 

@Milopapa so "Enable" is some sort of light FXAA?

 

@Milopapa

I'm not convinced by your equation of how Steam SS interacts with Oculus PD & the reason I say that is;

I've just run a variety of different settings of SS : D and most significantly;

1:1.3 (which according to you equation = 1.3) runs better than 0.6:1.6 (which as you pointed out = 0.96 - sorry it was 0.6 not 0.65), I've now gone to AA=3 & 1:1.3 and have a clearer picture with a little less performance than I had at 0.6:1.6.

I am testing GP3 mod @ Silverstone with 23AI. Starting at the back of the grid it's 45fps but within no more than 2 laps it steadies at 90fps. With AA=3 & 1:1.3, I get a slight dip to 45fps once at 90fps going around Village, it happens regardless if a car is near by or not. Otherwise performance is on par with 0.6:1.6

I'm not doubting the interaction between SS : D settings, but I'm certain it is not as you state it. can I ask where you have derived this formula and where can I read about how these 2 setting interact with one another? Thanks.

 

Are you running SteamVR beta or stable? There was a recent update which changed the multiplier interpretation. Previously it was the same as PD for Oculus (quadratic), now it's linear. Meaning, what was previously 1.5 is now 2.25 in SteamVR beta.

I'm not sure which is which in you X:Y description so I can't validate.

As for the source of knowledge: both Steam and Oculus do supersampling, they just call it differently (or at least Oculus does). The formula is pure common sense, applying two supersampling methods one after the other. Think about supersampling as upscaling video. If you downscale to half the resolution and then upscale to twice the resolution, you're just where you began, except you lost a bunch of details due to downsampling.
See this example:

 

From what I read, disable is. You'll need to test for yourself, I just use the default enable. I believe off is sharper.

 

I'm using Steam VR Beta. The ratio was SS : PD, so originally I had 0.6:1.6 but 1:1.3 performed better but visible change was negligible but I'm wondering if I am downscaling, then upscaling, as you state in your eg.

So I'm clear, with SteamVR linear settings, are the following PD-SS conversions correct;

0.78 = 0.6, 1=1, 1.3 = 1.69, 1.5 = 2.25 etc ?

And in my original scenario of using 1.6 on OTT & 0.6 on Steam VR I essentially upscaled to 1.6 and then downcscaled it by 0.78 = 1.248 but losing clarity in the process?

Now using 1.3 on OTT & 1 on Steam it is 1.3 without any loss of clarity?

I think tonight I am going to disable OTT and just solely try various SteamVR SS settings. Using Steam VR is there any way to get a FPS counter in the headset?

 

I tried steam vr ss but it didn't seem to do any. Do I have to enable it?

 

Interesting. Are there benchmarks/details, to see how it performs with different CPU's? I'm in the fence on whether its worth $ to upgrade from i5-4670k to i7-4790k, as I get mixed feedback.

 

Post #1 Updated with the latest Build Benchmark.

Results suggest an improvement in performance especially at Race Starts (I've added the Benchmark below), you can see the improvement between 5sec-10sec, this is the first tight right turn at Malaysia South Loop where the cars all bunch up. After that there is a slight improvement through to the end of the Benchmark, the sharp spikes are New Frames being added as you can see in the Plot Graph (below) from the first Benchmark to todays.

Remember the lower the Frametime (ms) the better the Performance/FPS. I think it's safe to ignore Build 1879993 Result due to the issues at the time with SteamVR itself (lots of small updates & rollbacks.)

 

@MarcG Just so I'm clear, are you no longer running Pixel Density but setting it solely via Steam VR Super sampling @ 1.9 to get approximately that 1.4 value you had before?