126% Custom Resolution vs. FXAA vs. No AA (HP Reverb)?

[eliteqs]

Assuming no aliasing in ED, HMD 1.0 in ED, and all your other settings not changed, which one do you like best and why?

I've done this comparison by turning on FXAA in NVIDIA Control Panel for that test only and using only 100% custom resolution in the FXAA and no AA test. I used 126% for the custom resolution test because SteamVR only increments in even percentages and 150% is too much for Titan V/2080 TI/Titan RTX tier with HP Reverb. Using latest SteamVR and Windows Mixed Reality for SteamVR Beta at 90Hz. I ask both generally for different VR sets and specifically about the HP Reverb as that is the headset I have and am curious what other people prefer among these three settings.

 

[jnTracks]

1.25 super sampling looks WAY better than AA and obviously better than nothing. But you’ll see the performance hit.
ED’s implementation of AA is very weak so I think that’s the main reason it’s not great. seems to only apply to certain aspects of the game render while leaving other jagged. A good temporal AA would probably be good since the main problem in VR is the shimmering, but alas, we don’t have that.

I don’t have a reverb, but this wouldn’t vary by headset, would it? Whatever the base res is, applying SS or AA would have the same effect.

 

[Kyokushin]

I am using 130% on steam (2519x2463 per eye) and SMAA ingame, full details (high/ultral on the reverb and 1080ti.
in game SS and HMD SS at 1.0.
WMR portal set to 90Hz and all to 'very high'. WMR Plugin and Steam VR on beta channels.

 

[eliteqs]

Super sampling is a form of AA.

Turning FXAA on in the NVIDIA Control Panel (or AMD equivalent) removes the question of ED's implementation of FXAA.

Mathematically, the result of each approach will the same relative to the base resolution of each headset, but preferences could be different for each headset resolution.

The differences between these three settings are subtle on the Reverb. I'm interested to know what people who try these three settings prefer and why.

 

[jnTracks]

Super sampling is a form of AA.

Turning FXAA on in the NVIDIA Control Panel (or AMD equivalent) removes the question of ED's implementation of FXAA.

Mathematically, the result of each approach will the same relative to the base resolution of each headset, but preferences could be different for each headset resolution.

The differences between these three settings are subtle on the Reverb. I'm interested to know what people who try these three settings prefer and why.

Are you sure the NCP forced AA is working in VR? I've found that a lot of those settings that do what they say on the tin in the case of flat monitors... Don't actually work in VR, so I think there is a question there.
And generally, driver forced FXAA is lower "quality" that game engine implementations since it simply blurs the whole screen, rather than detecting edges at all. So it would be lesser than the games higher AA options, but possibly better than the games FXAA since it might actually work on the parts of the game that don't get it normally. That is, if it's even working in VR.
Overall, SS is a form of AA, but since it's truly rendering higher and then filtering down, it's far superior quality than the other kinds, and comes with the performance cost to match.

From what I've read and personal testing (Rift S), the way VR looks is so different from flat monitors, that AA methods (other than true SS) are just not as effective, and maybe aren't worth the performance cost, however slight. I've found games with a temporal AA are seems to be most effective, so that might be an exception. But other than that, after all my trial and error, I decided to turn off all AA and get supersampling as high as I can afford. That gives me a result I'm most happy with. Except in Skyrim VR which has the temporal AA (and doesn't involved a lot of detailed reading of things, so I can deal with a slightly stylized, smooth look)

Another note is that the SS in steam VR isn't measured the same way as it is in ED (or Oculus' driver control software).
Steam VR is linear, so 200% is double the pixels, 1.5 SS in game would match that.
So if you're setting 126% is steam, that's about the same as 1.13 -ish in game.
(and it's multiplicative if you have it set in more than one place.)

 

[eliteqs]

I am using 130% on steam (2519x2463 per eye) and SMAA ingame, full details (high/ultral on the reverb and 1080ti.
in game SS and HMD SS at 1.0.
WMR portal set to 90Hz and all to 'very high'. WMR Plugin and Steam VR on beta channels.

You are using a Reverb with a 1080ti at a 130% custom resolution plus SMAA in game? What are your frame timings (can see in SteamVR settings or the in-game graph enabled through SteamVR settings Developer tab). In space versus in stations? Are you trying to maintain 90FPS or always using reprojection?

 

[Kyokushin]

In empty space i am hitting 90, not always of course, in stations or asteroids i am on reprojection.

repro works good in my oponion on WMR. Mostly i am on reprojection.

 

[aXeL]

Kyokushin, I'm using pretty much the same approach as you except I may push SteamVR SS to around 300% although above around 180% it seems to be diminishing returns.