VR Benchmarking Results, using FCAT from nVidia

Interesting findings Marc.
I also noticed no fps changes using postprocessing 0 or 1.
For AA it is also strange that highest 3 levels show identical fps. I didn't get to test above level1 considering I already saw that I cannot afford going higher.

 

BTW, Marc shared collected .csv's with me. I did some math and it seems that what we suspected but had no evidence is true - VR needs a lot of CPU speed.

I saw CPU time between 7.5 - 12(ms) per frame (Ivy Bridge), and this is just CPU time in game, that excludes GPU rendering and Oculus Runtime costs (which to me were surprisingly high). In order to have 90FPS we need everything done below 11ms.

 

Rather annoyingly (but probably in a good way for the future, who knows!) the recent SteamVR update changed the way they work with & display SuperSampling, so instead of 1.4 they now call it 1.9 but it's not exactly the same as what 1.4 was (that's the annoying bit! see link below for a better explanation).
https://www.reddit.com/r/Vive/comments/6c3o6w/psa_supersampling_is_now_a_linear_scaling_of/

I have now effectively "turned off" the SteamVR SS (putting it to 0.6) so I now rely solely on the Oculus setting, something I probably should've done from the start but oh well never mind

 

Ah good to know. Turning that off too now!

Can you do some night tests too Marc? I had some troubles racing online during the night.

 

I'm just trying to understand it at the moment, putting in 0.6 then 1.9 there was no visual difference, but the Benchmark showed a massive performance boost in 0.6 as though the SteamVR SS runs in the background regardless of the setting

I've not run much at night at all so will take a look soon

 

Ok ignore all that, Oculus Tray Tool wasn't up to date after the recent Oculus Drivers update (my bad), just ran some visual tests and I can't tell whether SteamVR over rides the Oculus PPD or not...or the other way round. I'll come back at it with fresh eyes another day!

I just hope we get native Oculus support one day as that would make life much simpler

 

I did spend a bit of time testing with just the framerate counter, and I found no difference in setting PPD via SVR or Oculus. I would dare to assume SVR simply passes multiplier to Oculus Runtime Now however SVR added some checkbox below the slider, not sure what that does yet.

I really wish SVR was more clear about how each feature works on Oculus vs Vive.

 

I just tried again and it appears something drastic has changed and whatever it was I'm seemingly getting better performance, but it skews all my benchmarks out of sync and I don't fully understand why, one of the pitfalls of benchmarking an ever evolving platform - by that I mean not just RF2 itself but also updates to Oculus & SteamVR, plus today throw in an nVidia Driver update and who the hell knows what's what any more

Also after the last SteamVR update the FCAT Benchmark only provides 2 CSV Results logs which are Merged (I show the Merged file in the Benchmark Results as FCAT describe in their tutorial), whereas before it would give you 3, the now missing one being an OpenVR CSV. So again no idea why that's suddenly not being recorded, maybe a change in the way SteamVR do things I dunno, all it means is my Benchmarks are a lot more sporadic and not necessarily 'precise'.

But anyway, I shall endeavour to continue until it becomes completely pointless!
Post #1 Updated with Build 1879993

 

@MarcG I think you have stumbled across something quite big with regards to performance boost on the OR. On STeamVR (Beta) I have deselected 'Enable Advanced Supersampling Filtering' and placed the SS at 0.6 and just using Oculus Tray Tool to have a PD of 1.6 and ASW on Auto and the game plays smoothly. ASW/45fps when at the back of the grid and for the 1st few laps and as the grid spreads I get 90fps.

I need to add that I am running AA=2, AF=x16, ALL graphical details on FULL, Shadows= Low, SP = Low.

I think you need to add this little tip to the VR Tips & Tricks Thread.
I should also mention that I have now edited the line in 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"

 

Maybe on to something, I'll have another play another day if I get time. I was getting frustrated with the whole thing earlier so I wasn't thinking straight when trying to find a cause but if you say you've found something too then yeah maybe we're onto something. Regarding the OTT it is very finicky, you have to reboot a lot for the settings to take effect as I found the "Restart Oculus Service" did not do anything some times. SO be careful that it's not throwing a Placebo at you like I think what was happening to me earlier

I may go back to using the Oculus Debug Tool just to be safe, feel free to put what you've found in that thread, maybe others will find the same results

 

I've just tried R3E using the 0.6 settings & 1.5PD on OTT, but if I deselect the Enable Advanced SS Filtering option, it renders the images in the eyes incorrectly. So enabling that and choosing 0.6 I am getting great performance gains in R3E too. Significant in that I have upped the graphics details in that too, GT3 19AI around Spa and hardly noticing ASW kicking in, where as just a few hours earlier it seemed like the whole front window kept warping. My next test is iL-2-BoX.

Just to add, with regards to OTT, I'm not changing my setting there at all. I set them up a while back and only thing I changed was taking ASW on rF2 from 45fps to Auto, that definitely kicked in as the I put the performance HUD up.

 

Magus, have you checked if you there's a difference between 0.6 and 1.0 in SVR and 1.5OTT? I agree with Marc, OTT is quite unreliable.
In any case, if both SVR and Oculus PPD value apply Supersampling, that's a scary mess

 

@The Iron Wolf For the least hour I've been trying to do as you requested and I chose R3E as the warping in that is very evident when ASW kicks in but I've been having issues with it recognizing my fanatec wheel. I've now deleted my profile and started again and seems to be recognised in R3E but now Steam VR is giving me an error 306. Just verified it & now working. SO I will try again. Before notching down m SteamVR settings to 0.6 it was previously on 2.0. SO I will try 1.0 & 2.0 & 0.6 again and report back shortly,

 

Fanatec driver has for sure some problems where it changes the order of how joysticks are enumerated, very frustrating thing, especially with little time we have to race
It would be helpful to keep OTT/Oculus Debug Tool out of tests, to first see how (if) new SVR SS works with Oculus. I also hope to drive a bit tonight, and I'll try SVR only, also in R3E.

 

I went back to rF2 as R3E keeps losing my profile and doesn't recognise my wheel unless I keep deleting my profile. Will get on Sector 3 forums to see what is going on. But anyway, here are my findings:

OTT was not touched and everything left as it was, just enabled the performance HUD. Tested USF @ Toban Long with 23AI starting at the back, all details on FULL, Shadow Low, SP = Low, AAx2, AFx16, PP=Low

0.6 SSF unchecked = a steady 45fps from start, headroom was in minus figures so ASW was working withing 1st few corners but smooth sense around those corners. Once the grid spread, straigh to 90fps with headroom between 5-40% but never dropped from 90fps whilst I remained on track and around 2-3 cars.

1.0 SSF checked = 45fps at start with slight stutters, managed to reach 90fps but with around 5-20% headroom and drops at times.

2.0 SSF Checked = Didn't go beyond 45fps and would drop below, stuttering constantly, even when pack spread.

I'm not sure why but my Steam VR setting was by default on 2.0 all this time, and for me I am experiencing a HUGE performance boost. Much more subtle if you have been using 1.0. Now I want to reverse the test and only use STeam VR settings and completely shutdown OTT but not sure how to monitor the performance without the performance HUD

 

Thanks. Does the image quality change between 0.6/1.0 and 2.0? I am not 100% sure I understand your situation, but SVR update changed supersampling scale to linear (before it was mutiplier for width/height). If it defaulted to 2.0, that means you had it 1.2-1.4 previously, and they converted multiplier to new value.

 

Okay I see, yes I had it set at 1.3 for Il2 via SteamVR.config file. No I don't see any noticeable difference in image clarity running OTT @ 1.6 : SVR @ 2.0 or 1.0, just the frames are worse.

 

Have you tried not having OTT override? Also, why do you do that? I am not picking up, just trying to understand Thanks for sharing your findings.

 

Yes I want to completely close OTT and just use the SteamVR settings and see what happens now.

A few months back SteamVR didn't have a PD setting so we had to manually configure it via the SteamVR.config file and also in iL2 you would get the blue chaperone ring around you in the cockpit when in VR so we also needed to edit that out in the SteamVR.config file too. I should probably delete that file and let it build a new one.

EDIT: the file is actually called steamvr.vrsettings now found in the Steam/Config folder.

 

It's hard for me to tell as I don't have the performance HUD to look at but at 1.0 SS in SteamVR, OTT OFF and all other conditions the same it was all very smooth, a few stutters at the start especially when dirt kicked up but I think the performance under OTT PD@1.6 was better than SteamVR @1.0 (ASW kicking in where needed probably makes it smoother), that is a completely subjective statement as I had no way of measuring but I just felt a few more stutters under SteamVR.

Anyway, that's my testing for tonight, I think i am going to switch the SS options OFF in Steam VR & leave OTT at 1.6/Auto and then see what you guys find out.