VR Discussions, Questions & Feedback

I don't understand how you can be satisfied with the performance achieved with 29 AI and the rain.
With my i8700 5.1 Megahertz, RTX 2010 ti 1850 megahertz,, 32 gb ram 3.6 megahertz, only AMS installed and highly optimized PC, with my HDM HP Reverb in the same conditions it is absolutely unplayable for me.
Are you a lucky man ... or have you never tried anything better

 

I wouldn't be satisfied either if my PC was running x1000 slower than it should!

 

Just out of curiosity i tried a quick race with 29AI@Spielberg in heavy rain, monitoring performance via the performance hud from the oculus debug tool.

I had 45 fps for the first seconds and before reaching the first corner it switched to 90 fps and after driving two laps i had exactly zero dropped frames.

For reference: System is a 9700K@5GHz, a GTX1080 and 3600CL15 but only using a cv1. Obviously my gpu is a bottleneck but thankfully, it's much easier to circumvent a gpu limit than a cpu limit.

I remember the oculus vr support in AC for example being much better than openvr from a performance standpoint.

Maybe the reverb just does not perform as good as the native oculus driver? The resolution of the HP Reverb is higher though, maybe that's the reason?

 

Ok so I just did other tests with my configuration and weirdly I have some noticeable differences with you. The way I test those settings are easy, I don't put the headset, it's on the side of my desktop with some stuff to trigger the proximity sensors, so I can see impact on settings on the Logitech gaming application on my android phone with GPU and CPU usage. I tested on test day, fixed weather, so no other cars.

Resolution of the monitor and Vsync have no impact. it can be low (800x600) or at max (1920x1080x60 on my TV monitor), and with VSync ON or OFF, it have 0 impact at all. My GPU stay fix at 72% Usage.

Testure Filtering : No impact at all, can go from Trilinear to x16 without changes in GPU usage.

Car detail have also 0 impact, can be Low or ULTRA, always 72% GPU usage and I can see my car don't have any change in quality, so I'm sure it only impact other cars in the distance (I'm alone as a test day).

AA settings : MSAA is a killer, MSAA at Medium and GPU usage raise to 84%. In comparison, SMAA at medium is 72% and SMAA at ULTRA raise at 78%. So the order of AA impact for you (SMAA Low - MSAA Low - SMAA Medium - MSAA Medium - SMAA High etc.) is really different than mine (NVidia vs AMD I guess).

I didn't test Grass as I don't do grass and forget to test Shadow, but I know I'm happy with the Low setting so far so I'll stop here

 

Yes, i think resolution of Reverb is the problem, but in Iracing and AC i can run at 90 FPS whit 8 xAA and 210 SS whit really clear image.
I don't see this big improvement in graphics compared to Iracing and AC that can justify such a big drop in performance, obviously in good weather conditions, without rain.
I agree with those who claimed that IR and AC have old graphics engines, but basically the visual differences are minimal but the performance is extremely lower.
For VR actually nothing is best that IR and AC in the world of racing simulation whith HP Reverb.
Really same as 1080 p monitor.

 

From what I understand about how SMAA is applied, it will not do anything to the image displayed in a HMD. From tests I did myself, I could see no difference between off and ultra. This was on a Rift CV1.

 

SMAA make a huge difference on my machine in my Rift CV1 for sure, from off to low to Ultra, easily see at the border of the car, on white lines on the road, dashboard, etc. This differences appear also clearly on the monitor. Took me time to find the SMAA medium setting to be the good balance in my headset...

 

Has anyone noticed that high frame-times at night and dawn/dusk seem to be associated with having the driver and dash in your line of sight? It seems like as the driver and dash come out of your field of view the frame times drop down significantly, despite the road in front of you and all of the complex track surroundings still being visible in your field of view. I was noticing this when using the Formula V12 on Imola. I wonder if there is some potential for graphical optimization here.

 

Ive been really enjoying this sim but have been running into an annoying issue while using my wmr headset. When I use my reverb, i find that I often have to take the headset off and alt tab out and back into the game. In the headset it looks like the game hangs. However, when i alt-tab forward and then alt-shift-tab back to the game or select it with the mouse on my running programs start menu bar than it goes back to normal. This never happens in a race but frequently happens within the menus. I also have a quest (with link) and I never see this happen while iam using it. I was able to try it out with an Odyssey+ and it also seems to have this issue. Anyone else running into this with their wmr headset/figured out a way around it?

 

It happen often to me with my Rift CV1 but only visual hang and the game still running in background (I can drive blindly, sounds ok, etc.). Sometimes it get back after few seconds, sometimes I have to exit the game and relaunch.

 

Tried it again and yes you're right it makes a difference on the border of the cockpit to the outside. However the underneath of the roof of the grandstand and anything with complex detail in the distance is still a shimmering mess using SMAA ultra compared to MSAA medium. It seems to have an effect on some assets but not others.

 

Did you restart the game during each change? Even though it doesn't ask you too I noticed a lot of times that made changes, but didn't restart, the next time I did everything was way off.

Will give this a try. I've since dropped to 'Medium' and I can't really notice a difference.

Could be. I'm not really a huge fan of SMAA, I see a lot of shimmering. I'm hoping that the improvements to Anti Aliasing that are 'coming' will open up more options here.

All that said there are so many variables, and there appears to be so little wiggle room, it's not so strange that we all have different experiences.

 

AMS 2 freezes when it is not the currently active program, so it sounds like something on your system is stealing focus somehow. Try something like Windows Focus Logger to determine if this is happening and what's causing it.

 

To be clear, I really prefer using MSAA over SMAA. SMAA don't fix all the shimmering in the distance, just work fine near the camera. MSAA even at low is simply too much fps loss for my machine.

 

Ok, you nailed it : for texture filtering you're right I didn't relaunch the game as it didn't ask for it, and shame Reiza!, you need effectively to relaunch the game to have this setting take in account.

So I just remade those test and here results on my machine:

Trilinear GPU Usage 70% Stable
Anisotropic x2 GPU Usage 69/70%
Anisotropic x4 GPU Usage 70/71
Anisotropic x8 GPU Usage 71/72%
Anisotropic x16 GPU Usage 72% Stable

So I'll make other visual test (above was made headset on my desk like always) to see if it really matter in VR and if not the x2 will be my choice as it's the less costy one.

Edit : Oh and I tested yesterday race with Car Detail LOW and can't see any difference in VR for my car as I said and for other cars, so I'm gonna test FPS impact on race start, but it will be more difficult as fps is more fluctuant in this situation.


Edit #2 : With Anisatropic x2 or x4 you can see the limit not that far from your car and all other things in the distance are blurred. So I'll stick to Trilinear.

 

Yeah, I find having MSAA on Low, with trilinear filtering is the best visuals-to-performance compromise for me.

 

Oops, sorry, this was BS, had a proper look about on drone cam and they are all 2d sprites. Apart from the marshals, they are 3d models.

Something I had never noticed before is that the trees are 2d sprites as well. If you look out the side of the car when passing through the trees they all turn to look at you as you go past. Good place to try this is Brands Hatch GP in a Caterham. Quite spooky!

 

A few settings I've found that IMHO give a very nice image with not much performance penalty if you have an nvidia card.

In the nvidia control panel in 3d settings set Antialiasing - FXAA to on and Multi-Frame Sampled AA (MFAA) to on. Then in game use MSAA - Low. If you also use sharpening in the oculus settings xml file this gives a pretty good output and solid 90fps for me on a GTX1080.

 

I checked on PCars2 and some of the crowd in that are 3d but most are 2d. Checked again in AMS2 at Monaco / Azure, and some of the crowd are 3d. It's reserved for the people that are animated, so a guy waving, someone taking a photo, another waving a flag etc.

 

Ok, tested. GPU raise to 85% vs 72% with only SMAA Medium and I agree image is nicer with your settings.

My tests trigger a weird bug. After removing all AA settings in NVidia control panel to retrieve my usual situation with only SMAA medium in game I made test with only MSAA, and MSAA High have a nice bug:
MSAA OFF & Post AA OFF = GPU 64%
MSAA Low = GPU 74%
MSAA Medium = GPU 82%
MSAA High = GPU 59%

Yes MSAA High is the lowest fps cost (better than no AA, what?) and image is pretty nice, AA is applied for sure as I can see that on the car, fences, etc.
what happen here? I reproduced this around 10 times just to be sure, switching between different AA and MSAA High.
I guess I need to post that on Bug report side.
Can somebody which have a way to see GPU usage test this also?

EDIT : Ok, found the "bug", my FPS went from 90 to 45 fps ASW with MSAA High How stupid am I...