Performance testing for Odyssey+ VR versus clarity

  • I used the idea of watching the frametime from SteamVR in order to measure impact to performance along with a clarity study. I decided to repeat that kind of experiment and I used our flight simulator (Aerofly FS2 Q400 which has excellent performance and has some detailed scenery from ORBX (Ultra/Insane settings) for Innsbruck airport ) instead of car racing as a test case since there are many cockpit instruments with small fonts to get the best impression of image clarity (can I read the number without leaning). The ASW column refers to using motionvector or not, FPS was provided by the in-game printout, and I did try adjusting RenderTargetScale. A summary of the results for my CPU (9700Kat 5Ghz) / GPU (RTX2070) show that the optimal spot for me is as follows:

    ASW=motionvector, SteamVR Video Tab at 212% and Application at 100%, In-game SS = 1.6, RenderTargetScale = 2.0, all of which provides a B+ image clarity. I wish we could try a "sharpening" filter (Unreal Engine has it) that might get it to an A- but maybe I have to wait for Gen 2. For this application, RenderTargetScale has a small clarity effect but significant AA effect. I did have to always check to see that I was in the small sweet spot of the Odyssey.


    Here is a table of the test: (Sorry about the incorrect spacing)

    ==================================================================================

    WMR Benchmarks (ASW = on => motionvector ) For 90fps = 11.1msec and 45fps =22.2msec max
    Test location is parking at ORBX Innsbruck
    Time FPS ASW SteamVR_SS IngameSS RenderTargetScale Clarity Comments
    12.8 45 On 200 1.6 2.0 B+
    15.8 58 OFF 200 1.6 2.0 B+ Why did the frametime go up a little??

    10.7 70 OFF 100 1.6 2.0 C+
    17.9 52 OFF 250 1.6 2.0 B++
    17.9 45 OFF 250 1.7 2.0 A- slightly clearer but strong aliasing begins
    15.0 42 On 250 1.6 2.0 B++ Clear but flicker on instruments
    12.3 45 On 212 1.6 2.0 B++ Optimal?
    12.7 45 On 212 1.6 1.0 B- less clear and with noticeable aliasing
    12.7 45 On 212 1.6 3.0 B++ noticeable AA improvement over 1.0
    10.5 45 On 150 1.6 2.0 B- Less clear but provides FPS headroom

  • I read that the default WMR rendertarget has just been changed to 2.0


    Also i can't help feeling it would be more processor efficient to only do one SS activity rather than supersample in steam and then again in game. Are they the same sort of thing?


    You mention sweet spot. Would you agree that with Ody you need eyes forward more than Rift to avoid some blur? With head fixed but glancing down with your eyes it isn't so sweet or is that just because i'm wearing glasses?

  • Phil,

    I wear glasses too (waiting on my widmovr lenses to experiment ) and yes the "sweet spot" is difficult to keep right there for flight sims (have to keep checking and bumping the HMD ) - not as noticeable in many games or demos (really like theBlu and Apollo 11). I have progressive lenses and I also have to get them lined-up to get the most clear image. Even so, I like the image better than the Rift - things just look beautiful with the O+

    I my experiment above, I used the engine readouts in the center MFD of the Q400 for the clarity test without leaning in but with the eye at the reference eyepoint where the triangle device on the windshield is correct. The best was that I could barely read them.


    Dave W.

  • Cumulative 1.40 SS in Steam seems to be the clarity sweet spot.


    No need to change settings in Aerofly, and Vulcan and High quality AA seal the Deal for me.

    Devons rig

    Intel Core i7 8700K @ 5.0GHz / 32.0GB G.SKILL TridentZ Series Dual-Channel Ram / ZOTAC GAMING GeForce® RTX 2080 Ti Triple Fan / Sound Blaster Z / Oculus Rift VR Headset / Klipsch® Promedia 2.1 Computer Speakers / ASUS ROG SWIFT PG279Q ‑ 27" IPS LED Monitor ‑ QHD / 2x Samsung SSD 850 EVO 500GB / Windows 10 Pro 64-bit /Gigabyte Z370 AORUS Gaming 5 Motherboard

  • Cumulative 1.40 SS in Steam seems to be the clarity sweet spot.


    No need to change settings in Aerofly, and Vulcan and High quality AA seal the Deal for me.

    You mean 140% in Steam on top of some FS2 supersampling or set FS2 to 1.0 ?

    It will be nice to get a definitive that we all agree on.

  • You mean 140% in Steam on top of some FS2 supersampling or set FS2 to 1.0 ?

    It will be nice to get a definitive that we all agree on.

    No extra supersampling from Aerofly. The Odyssey higher resolution requires Steam ss at 1.40 to match the headsets natural resolution. Vulcan and High Aa takes care of the rest and gives a beautiful picture, if your computer can handle it.


    Steam will automatically try to set you at 200. Using Individual steam settings for aerofly you can remove 70 to achieve a total setting of 140 to match the hmds native resolution.


    [Blocked Image: https://imageshack.com/a/img924/3304/oEwsUJ.jpg]

    Devons rig

    Intel Core i7 8700K @ 5.0GHz / 32.0GB G.SKILL TridentZ Series Dual-Channel Ram / ZOTAC GAMING GeForce® RTX 2080 Ti Triple Fan / Sound Blaster Z / Oculus Rift VR Headset / Klipsch® Promedia 2.1 Computer Speakers / ASUS ROG SWIFT PG279Q ‑ 27" IPS LED Monitor ‑ QHD / 2x Samsung SSD 850 EVO 500GB / Windows 10 Pro 64-bit /Gigabyte Z370 AORUS Gaming 5 Motherboard

  • I am the OP and my conclusion was as follows:

    "A summary of the results for my CPU (9700Kat 5Ghz) / GPU (RTX2070) show that the optimal spot for me is as follows:

    ASW=motionvector, SteamVR Video Tab at 212% and Application at 100%, In-game SS = 1.6, RenderTargetScale = 2.0, all of which provides a B+ image clarity. I wish we could try a "sharpening" filter (Unreal Engine has it) that might get it to an A- but maybe I have to wait for Gen 2. For this application, RenderTargetScale has a small clarity effect but significant AA effect."


    I set the Reprojection in the default.vrsettings file to "motionvector" so the FPS was a constant 45 - this allowed me to make all the other changes and still keep the frametime under 22.2 milliseconds. I wanted a fairly high density scenery viewpoint so I picked ORBX Innsbruck - at a parking spot so as to keep the viewpoint repeatable.


    What I wanted to do globally was to set the SteamVR Video tab (global SS) to 200% and the Application specific at 100% for all applications and then if the application has an internal SS setting, adjust that. First, I verified by actual test (measuring clarity and frametime (not FPS) in milliseconds) whether those SteamVR SS settings had any effect - I changed the 200% to 150%, 250%, and 100% and the change in frametime (and clarity) verified that there really was a change. Then I went back to the 200% Global - 100% Application in SteamVR and now only adjusted the Aerofly FS2 RenderScaleFactor from 1.2 to 1.7 and again I saw BOTH frametime (not FPS) and clarity change.


    The summary of all this was to show a concept of how to evaluate your settings - to see how much you can increase your settings towards maximum image clarity but still get the fluid 45 fps with Reprojection. I really don't want to get the stutters! The values that I chose were only for my CPU/GPU combination which is maybe at the 80% point of max performance in AFS2 with todays hardware. YMMV!!


    Flying around in VR within the sweetspot of the Odyssey+ and smooth reprojection gives a great flight - just beautiful.


    Dave W.