VR - experimental support

[KingOfTheRoad77]

I just did a couple of trips right now:
Sandpoint > Coeur d'Alene
Spokane > Coeur d'Alene
Coeur d'Alene > Spokane (Coastline Mining)

=> FPS were noticeable better than the days before, even in and around Spokane and Coeur d'Alene, for whatever reason. And they climbed up to 65-75 during the last trip as soon as I left Spokane and headed towards Coastline Mining.

fpsVR Report:
App: amtrucks HMD: Pimax Vision 8K X (75.000 Hz, IPD 63.0)
GPU: NVIDIA GeForce RTX 3080 (30.0.15.1179, Tavg 64.1, Tmax 70) CPU: AMD Ryzen 5 5600X 6-Core Processor (Tavg 50.2, Tmax 56)
Delivered fps: 56.27 Duration: 85.8min. Headset was active: 100%
GPU Frametimes:
Median: 16.5 ms
99th percentile: 26.2 ms
99.9th percentile: 29 ms
frametime <13.3ms(vsync): 21.5%
CPU frametime:
Median: 13.2 ms
99th percentile: 24.8 ms
99.9th percentile: 28 ms
frametime <13.3ms(vsync): 51.7%
Reprojection Ratio: 0.0% (for Index/Vive/VivePro headsets only)
Dropped frames: 0 or 0.0% (for Index/Vive/VivePro headsets only)
Max. SteamVR SS: 100%
Render resolution per eye: 4512x3948(by SteamVR settings, Max.) (HMD driver recommended: 4514x3948)

CPU usage was around 12-16% with a maximum of 19%, whereas the GPU always was between 75 and 99% (mostly >90%).

Still not sure what to think of it and what to do:
I could take a trip back to Sandpoint, then re-install the 5800X 3D and then do almost the same routes than today. But I don't think there'll be a big difference... might already be lack of a longer rain period I had now which might affect FPS significantly - or not...
On the other side: At least with the 3080, I certainly won't get beyond an avg. of 60 FPS or so anyway on these routes.
So, main reason for any further evaluation of the 5800X 3D would be the future upgrade to the 4080 in order to avoid any CPU limitation best as possible... But on the other side, there still would be plenty of time to evaluate this as soon as I have a 4080 and hopefully to save some money until then if the CPU's get a bit cheaper.


Well, I think if SCS (@Komat) would work on the VR programming and optimize it, integrate VRS and FFR, at least that the NVidia driver supports it officially, we already would have a huge huge gain and save even more money...

Just let's pray

[Krists]

I can see that your Pimax headset has 75hz refresh rate which means that you wouldnt see any benefits from 75+ fps in game, thus your CPU and GPU is measured by max frametime of 13.3ms. Can you do more tests with 5800x3d? I can see that 5600X can manage only around 50-60% of frames within the 75fps (13.3ms) range whereas for me with 5800x3d I get above 90% both in ETS2 and ATS for 90fps. I would say that you are somewhat bottlenecked with CPU, but there is much higher bottleneck with GPU, therefore you dont really feel this difference. If you would upgrade you GPU to 3080ti and above, you would probably start to see you CPU lacking.

I’ll see if I have energy later this weekend to swap the old components to see at what extent CPU/GPU and ram can affect the performance.

@Komat - how about we organize a crowdfunding to push you for more VR optimization?

[Krists]

Tried with 210% SS in SteamVR in ATS and performance was the same as before, but with better picture.
I also noticed that GPU load was around 80-95% and 11GB of memory usage (9.5GB at 150% SS), CPU was 10-20%.
Delivery from Portland to Seattle during daytime.

fpsVR Report:
App: amtrucks HMD: HP Reverb Virtual Reality Headset G20 (90.001 Hz, IPD 68.0)
GPU: NVIDIA GeForce RTX 3080 Ti (30.0.15.1277, Tavg 71.0, Tmax 76) CPU: AMD Ryzen 7 5800X3D 8-Core Processor (Tavg 60.2, Tmax 67)
Delivered fps: 74.91 Duration: 29.3min. Headset was active: 11%
GPU Frametimes:
Median: 9.7 ms
99th percentile: 12.7 ms
99.9th percentile: 15.8 ms
frametime <11.1ms(vsync): 89.2%
CPU frametime:
Median: 6.9 ms
99th percentile: 12 ms
99.9th percentile: 14.5 ms
frametime <11.1ms(vsync): 97.6%
Reprojection Ratio: 0.3% (for Index/Vive/VivePro headsets only)
Dropped frames: 28 or 0.0% (for Index/Vive/VivePro headsets only)
Max. SteamVR SS: 210%
Render resolution per eye: 4584x4484(by SteamVR settings, Max.) (HMD driver recommended: 3164x3095)

EDIT: did another try - cranked up stereo buffer to 1.2, took freightliner cascadia and was driving around riverton during rain and it was terrible - CPU and GPU charts were orange with some red frames for GPU and 11.5GB memory usage

fpsVR Report:
App: amtrucks HMD: HP Reverb Virtual Reality Headset G20 (90.001 Hz, IPD 68.0)
GPU: NVIDIA GeForce RTX 3080 Ti (30.0.15.1277, Tavg 73.1, Tmax 77) CPU: AMD Ryzen 7 5800X3D 8-Core Processor (Tavg 59.5, Tmax 67)
Delivered fps: 48.28 Duration: 8.1min. Headset was active: 100%
GPU Frametimes:
Median: 15.8 ms
99th percentile: 22.4 ms
99.9th percentile: 26.1 ms
frametime <11.1ms(vsync): 29.3%
CPU frametime:
Median: 8.6 ms
99th percentile: 18.3 ms
99.9th percentile: >30 ms
frametime <11.1ms(vsync): 67.7%
Reprojection Ratio: 20.1% (for Index/Vive/VivePro headsets only)
Dropped frames: 4 or 0.0% (for Index/Vive/VivePro headsets only)
Max. SteamVR SS: 197%
Render resolution per eye: 4436x4340(by SteamVR settings, Max.) (HMD driver recommended: 3164x3095)

[dhrto]

Thanks @Krists for posting your config and findings of your testing! Very informative.

It seems you have almost everything at max. compared to me and still manage to get nearly stable 90 fps. Maybe I should check it out if I can up my settings some more, starting with the 'r_manual_stereo_buffer_scale' / SS. I thought the 3080ti is not a lot faster than the 3080 (10-15%, no raytracing), but I might be mistaken.

[Krists]

Welcome.
The biggest take I can get right now for me is that ETS2 performance is way worse, but graphics is way better than ATS with the same settings (you can see around 50 FPS in ETS2 vs around 80 FPS in ATS and most importantly the frame time % of GPU (a stable frametime should be above 80 or 90%) and reprojections (to me above 10% reprojection ratio is very high and stuttery, and can lead to motion sickness).
If I have enough time this weekend, I can switch my every part from my old system to test how each item (RAM, GPU, CPU) affect the performance, however I think the aliasing and performance differences between ETS2 and ATS means that I have some software issue that wont give me the proper "clean" result for the test.

[dhrto]

ETS2 and ATS should perform and look similarly. Are you running any mods? I used to have pro mods for ETS2 and from what i remember that pro mods can be quite hard on the performance. I've been running vanilla for some time now, since the base map + all the map dlc's are covering more and more of Europe.

[Krists]

I did have truck, traffic and real company mods, but right now all mods are disabled in both games. It seems like I have higher stereo buffer settings “stuck” in ETS2, however config shows 1.0

After couple more testing with various SS modes and stereo buffers, for my system the best results (stable 55fps, with no stuttering or reprojection below 2%), I’m finally settled for 200% SS in ATS and 120%SS in ETS2 and both games having stereo buffer of 1.0 and all other graphics settings to max.

ATS still looks much more aliased than ETS2, but I couldnt figure out how to fix it - even tried 100% SS and 1.5 stereo buffer, but game was still aliased with unpleasant stuttering.

[KingOfTheRoad77]

I would say that you are somewhat bottlenecked with CPU, but there is much higher bottleneck with GPU, therefore you dont really feel this difference. If you would upgrade you GPU to 3080ti and above, you would probably start to see you CPU lacking.

Yes, that is my feeling, too. Perhaps something like 80% GPU and 20% CPU or the like.
But as mentioned, I don't know whether it is wise to spend 500 € now for a CPU that might help with a faster GPU in the future... when the 5800X 3D or a comparable 76/7800X perhaps might cost 100 or 200€ less - which I wouldn't really save as I'd need a new AM5 board and DDR5 It's really difficult...

Furthermore, I also had / have sessions like these:

ETS2
Delivered fps: 57.9 Duration: 35.1min. Headset was active: 100%
GPU Frametimes:
frametime <13.3ms(vsync): 13.7%
CPU frametime:
frametime <13.3ms(vsync): 78.8%

ETS2
Delivered fps: 62.25 Duration: 33.9min. Headset was active: 100%
GPU Frametimes:
frametime <13.3ms(vsync): 39%
CPU frametime:
frametime <13.3ms(vsync): 88.5%

ATS
Delivered fps: 64.33 Duration: 25.6min. Headset was active: 100%
GPU Frametimes:
frametime <13.3ms(vsync): 52%
CPU frametime:
frametime <13.3ms(vsync): 91.2%

So, it is not always that bad.
But when looking through the different logs since two or three years, I notice that ETS2 provides better CPU frametimes in average than ATS.

Last but not least, the truck sim's aren't the only games I play in VR. Have a look at these:

Subnautica
Delivered fps: 39.13 Duration: 19.1min. Headset was active: 100%
GPU Frametimes:
frametime <13.3ms(vsync): 0.3%
CPU frametime:
frametime <13.3ms(vsync): 91.8%

=> Absolutely GPU limited

Elder Scrolls Online via vorpX
Delivered fps: 66.35 Duration: 128.9min. Headset was active: 100%
GPU Frametimes:
frametime <13.3ms(vsync): 100%
CPU frametime:
frametime <13.3ms(vsync): 91.4%

=> This time, a little CPU limited, but still >90%

Elite Dangerous
Delivered fps: 71.27 Duration: 76.7min. Headset was active: Unknown
GPU Frametimes:
frametime <13.3ms(vsync): 83%
CPU frametime:
frametime <13.3ms(vsync): 99.8%

=> Here mainly GPU.

So, 1st, in general, it seems to be the GPU in most of the cases.
2nd, the truck sim's really aren't optimized at all, it seems.

I’ll see if I have energy later this weekend to swap the old components to see at what extent CPU/GPU and ram can affect the performance.

I hadn't, unfortunately.
But I will see to drive some trips with lower FOV (=decrease of resolution) and adjust SS in order to get close to Reverb G2-resolution. This way, we can do some comparison in the other way. Finally, it also might be that the HMD drivers themselves could have some impact, too.

@Komat - how about we organize a crowdfunding to push you for more VR optimization?

I'd agree
Or make it a DLC to pay for in order to compensate your work for it a bit.

[KingOfTheRoad77]

I lowered the FOV and increased SS to 112% (=> 3488x4180), which should almost fit G2 resolution in terms of pixels.

Trips:
1. Spokane (Coastline Mining) > Sandpoint (nightime)
2. Sandpoint > Coeur d'Alene (daytime)
3. Coeur d'Alene > Spokane (daytime)

Here's the result:

Delivered fps: 65.76 (before: 56.27) Duration: 57.6min. Headset was active: 100%

GPU Frametimes:
frametime <13.3ms(vsync): 52.8% (before: 21.5%)

CPU frametime:
frametime <13.3ms(vsync): 62.6% (before: 51.7%)

Max. SteamVR SS: 112%
Render resolution per eye: 3488x4180(by SteamVR settings, Max.) (HMD driver recommended: 3294x3948) (default: 4512x3948)

So, main gain is on GPU (not surprising ), but CPU frametimes did improve as well a little bit.
Would now be interesting so see how my 5800X 3D would perform.

In general, driving felt much smoother with the lower FOV / resolution, of course. But I don't know whether I can adapt to this - there's quite a lot of my usual image missing to the left and right But well, for performance research purpose right now, it's okay.
Dammit - I want a 4090 with max 350W NOW! Or working FFR...

[Krists]

I dont think 4090 will be below 400W

I am by no means a programmer nor game developper, but looking at our test results, the biggest and cost effective gains (from user perspective), could come from optimizing game itself.

Kinda hard to justify 1000 EUR upgrade for mostly one game and still not being able to run at stable 90fps with antialiased game (ATS), and this is exluding the cost of the remaining parts

Nevertheless, I still love playing these games, if possible, I will probably buy next gen graphics cards when they come out, and if I would have space in my home for motion rig, that would be the next purchase