VR - experimental support

[dhrto]

@KingOfTheRoad77 Thanks for the elaborate test man! It really gives us some extra insight. So it seems the 4080 barely made a difference compared to a 3080. So that one is off the list for me as well. I generally tend to upgrade when the performance improvement is at least 80%, preferably 100%, which would mean 3080 > 4090 spec wise (and benchmarks in other games). But as your testing indicates, this does very little for ATS (and presumably ETS2). And also, the price of the 4090 is way north of what I can justify on a GPU. I'm still holding onto my 'hard' limit of €1000 (although the 3080 was a little bit over, yeah bought it at the wrong time).

I haven't done so much extensive testing as you, but what I see when driving around it that I'm mostly CPU limited too, especially in and near the cities and around new areas. I have yet to explore Texas (bought it, played a little, but unfortunately a little to busy with other things until Christmas break). I'm currently on a 5800X3D and am looking forward to what the 7000X3D series will bring to the table. On the other hand, I think the game engine unfortunately remains the limiting factor. If only it could benefit more from multicore processing.

@Phantom_Mark Thanks for sharing your insights on the 3080Ti > 4090 upgrade. It's really too bad it all seems pointless for this game at the moment. Regarding you 4090 'sleeping'. There's clearly some performance room left and it seems you're CPU limited (yes even with that monster of a CPU, I blame the game engine). Maybe you can use that performance room for a higher buffer scale, to get a clearer picture with less aliasing.

[Phantom_Mark]

I am still building the 13900k build here, so that should lift the CPU as high as it can get pretty much, be interesting to see if there is any gain.

I broadly have no complaints as brute force is seeing me get enough performance to enjoy the game, that is what matters most, and to be fair to SCS there is a ton of engines which have their roots from back in the early 2000s ,or before, which generally suck in VR performance. I do wonder that the long term goal for SCS is tho, do they intend to keep bolting on to this old engine or if they plan a new multi core enabled engine with proper VR support one day ?? IDK ? I have had better truck driving physics working in UE4 and UE5.0 when it comes to the driving model, which tells me UE5.0 could be a viable option for SCS maybe one day, it can handle the maps as well, would be a large shortcut to the next gen for them.

[Madkine]

There are many reason why UE5 is not a viable option. (including a need to rebuild every map). It would take far longer to build in UE5 than to keep upgrading their current engine. Anyway that's off topic for this thread.
If you want further info, see one of the several threads that have discussed this.

[KingOfTheRoad77]

... my experience at the moment in VR is at least stable and very playable to some degree, so anything better is better.

Yes, I generally can agree.
From my personal point of view, there are ~4 stages concerning FPS and playability (at Pimax 8K-X with 75 Hz):

1. 70-75 FPS: Perfect
2. 50-70 FPS: Well playable - but annoying if you have greater areas with 70-75 FPS since the stutter can be noticed at once. Less annoying, if it's always this low - except the fact I know I could be better
3. 35-50 FPS: Still playable - but more annoying - and even less enjoyable if you know there are areas with higher or even top FPS
4. <35 FPS: Playable - but well - let's not talk about this...

At the beginning, with SteamVR only, I had "3." and "4." on large parts of the map, "2." sometimes, and "1." only in (old map) desert areas in Nevada and Arizona. It was okay, since I didn't experienced better performance personally.
Then, I desicovered first fholger and afterwards OpenXR - and suddenly, even larger parts of California, Oregon, Washington, NM, Utah and Idaho became very well playable ("2." and "1.").

But then, something changed with newer map DLC's, slightly with Colorado, clearer with Wyoming, Montana and finally Texas now, as well as with reworked California areas. And all over sudden, not even OpenXR can handle stable 70-75 FPS there, especially in cities.

Of course, it is still playable - but somewhat annoying.


I've ordered a 5800X3D on Saturday evening, it hopefully will arrive tomorrow or on Wednesday.
So, more tests with the 3080 and the 4080 will follow, although this probably will be some work and monotone driving always the same route...
But okay - I totally agree with you, @dhrto , that such high costs for GPU and/or CPU really must be worth it: Of course, the 5800X3D is still one of the fastest CPU's especially for gamers. But if I don't play current AAA*-games and if older games such as ATS/ETS2 don't benefit from its architecture, the money is just wasted. This counts even more for a 4080 or 7900 XTX (not talking about a 4090), if the game engine is so much CPU-limited.


Therefore, @Madkine :
I don't think, the general discussion - or at least mentioning of how "aged" the engine is - is completely off-topic since it seems to affect VR performance of the games - unsurprisingly - even more significantly than flat monitor gaming.

Of course, we all know that VR support still is beta/experimental (even though it's working great since years), and that its implementation is the solo project of @Komat - which I (or may I say "we" ) appreciate very very much
I know also from the live streams that it's really a task to create all those large and excellent map as well as other DLC's - and this is appreciated also very very much - I'm always and gladly a first-day-buyer
And I know, that VR in general is still some kind of a niche, not talking about enthusiast-VR such as with my Pimax 8K-X.

Nevertheless, I think it's a pity that there are only so few people realizing the large improvement, gain of immersion and atmosphere VR is bringing to your games! And I think, I'm probably not the only one.

So, at least refering to the meanwhile 10 year old engine's core without modern CPU architecture optimization in VR context is somewhat understandable and logical, since VR is one of most demanding (gaming) technologies. Don't you think?
Certainly, it's not about discussing this here in this topic forever. But I think it's legitimate for it to be mentioned here, too.

[Phantom_Mark]

For me personally changing this made a massive difference to my stutters....as suggested to me here previously.

uset t_averaging_window_length "0"

[KingOfTheRoad77]

@Phantom_Mark
Thx

I tried and verified it with three different settings:
20 - default
0 - suggested by you
375 - =5x max FPS as suggested by someone else in another topic here in the forums

Result:
"0" might really be a little smoother - but the lack of FPS still is quite dominant and perceptible.
But except for further benchmarks, I will give this a few more trials in the upcoming days

@dhrto
"Maybe you can use that performance room for a higher buffer scale, to get a clearer picture with less aliasing."

I tried this just now, increasing SS from 100% to 125% = 4.272 x 3.556

=> Avg. FPS of 50 from Dallas to Lufkin

=> SS of +25% does cost about 10% FPS

BUT!! :
=> It's quite an impressive increase of image sharpness and quality, compared to 100%. And much less anti-aliasing and flickering, even though it is still noticeable.

So, just from this one test drive, I'd say, that a 4080 (if it gets cheaper) perhaps might be worth the invest. Or an even less expensive 7900 XTX, of course, if it performs the same or even better.

Looking forward to the 5800X3D...

[Phantom_Mark]

Thanks for your efforts here, I will try and get some more fact based feedback for you once I am finished setting up this new system etc, see how things compare this side.

[KingOfTheRoad77]

5800X3D arrived and installed; 3dMark TimeSpy and Port Royal everything as expected.
ATS test trials will follow later.
___

EDIT 18:30:

Unfortunately, I was not able to do more than one test drive since I'll be on the move till the rest of the evening.

So, just a little spoiler already:
Seems, as if there's not a cheap way to gain FPS - but it is possible

I hope I'll be able to complete testing until the end of this week, and then I'll share all results in a diagram.

[KingOfTheRoad77]

Okay.
I'm done with testing and benchmarking.
And I can honestly say: I know the routes between Dallas and Lufkin, and Sandpoint and Colville now by heart: Just set me adrift somewhere there, and I'll find my way back to any of those cities or settlements in between blindly And I won't accept any further deliveries between those towns in the next time

But all jokes aside - here are the results:

[ external image ]

This graph shows quite well that I was limitied both ways, CPU and GPU.
But surprisingly (for me), CPU-limit was even much stronger than expected, since on the 5600X I just gained 3 FPS when switching from the RTX 3080 to the 4080, whereas FPS increased by +8 when switching only the CPU while staying with the RTX 3080.

The fact that I was not able to notice that already during my first 5800X3D-trials half a year ago probably is, that I was still on SteamVR then and not on OpenXR, as suspected earlier already. And even fholger's SteamVR mod didn't help enough there. I will try to verifiy SteamVR influence somewhen later (although simply on the new 5800X3D as I don't want to install back the former 5600X once again and go back to the 5800X3D again afterwards - would cost me additional two hours just for that test).

Nevertheless, the RTX 4080 has quite a large impact, too - as soon as the CPU's handbrake is released => additional +8 FPS = +14,5% in average over both test routes.

Finally, if upgrading both CPU as well as GPU, I gain an average of +16 FPS = +31% in total.

Unfortunately, 68 FPS aren't equal to 75 FPS, and 59 FPS are even farther away... But compared to my starting point ten days ago - 52 respectively 45 FPS - much much better already.

Conclusion:
Now, the game engine being primarily CPU-limited and with OpenXR working with my Pimax, I won't get very far without upgrading the CPU first. I wished it would not be necessary - but meanwhile I am convinced, that not even a RTX 4090 would be able to achieve stable 75 FPS in ATS/ETS2 on my old 5600X. Even though 3DMark result statistics may say something else - well... doesn't help me... the one is an artifical benchmark test, the other one a real game


Due to the hint with the overhead of GPU power at least on the 5600X, I did a second set of test drives in order to evaluate the impact of SuperSampling:

[ external image ]

(Unfortunately, at the time being, I did not have in mind already to do a comparison with the 5600X and the 3080 @100% and 125% SS - so those values are missing.)

Interestingly, three pairs of those values are very close to each other:
5800X3D – 4080 @125% (56 and 51)
5800X3D – 3080 @100% (60 and 51), and
5600X – 4080 @100% (55 and 49)

=> An average of ~50 - 60 FPS depending on sector, daytime, etc. can easily be achieved by changing either CPU or GPU alone. Swapping both allows increasing of SS to 125% without loosing too much performance. What to choose and what to go for, certainly is a question of personal preferences. Or just skip SS, upgrade both CPU and GPU and take 59 - 68 FPS @100%


Final thoughts:
The decision is still not easy as there really is no easy - and especially cheap - way.
But I think having upgraded to the 5800X3D now with working OpenXR really helps a lot to fully benefit from the NextGen GPU's, whichever it will be. So, it will stay with me , and the old 5600X will be sold.

Looking forward to the 7900 XTX next week now - and hoping that it will work with VR in general as well with my Pimax 8K-X in special: As far as I was able to understand from the OpenMR forums, AMD didn't confirm any compatibility of the new RDNA3 with common VR headsets so far, and even the old generation of 6800 and 6900 XT apparently did cause some issues with some HMD's not being recognized on DisplayPort, especially revised versions of the 8K-X. Lucky me that I've a got one from the very first batch, this should work. Hopefully.
If so, and if I'll be able to catch one on Dec 13th , I'll update my diagrams with further results for comparison purpose between 7900 XTX and RTX 4080 on my ATS benchmark routes


Maybe, these benchmarks might be of some help for some of you, too
In case of further questions, remarks, corrections or future proposals, please let me know.

Drive safely


Appendices:
General conditions:
* Ryzen 5600X / 5800X3D
- NVidia RTX 3080 / 4080
- 32 GB DDR4-3600
- MSI X570 Gaming Edge, BIOS 1.J0 AGESA 1.2.0.7
- Pimax 8K-X
- NVidia drivers: 527.37
- Pimax drivers: 1.01.284
- OpenXR 1.2.2 (@FSR 90% and sharpness 60%)
- PimaxXR 0.2.8
- Win10 Pro 64bit, 21H2

VR resolutions:
100% = OpenXR-FSR-90% = 3.416 x 2.844
=> 125% = 1.25 in PiTool with OpenXR-FSR-90% = 4.272 x 3.556

- Measurement of Avg FPS with RTSS plug-in for HWInfo64


Test route #1: Dallas - Lufkin

[ external image ]

- Start at daytime around 9am
- (Perhaps) important mods: Weather 2.3, Grime's Early Autumun, Daniel's Random Events
- I directly included both weight stations (waypoints 2 & 4) into routing since it's 60-80% sure I'll catch at least one - so better including both directly for better comparison of the results.
- Always driving according to traffic rules (preferably constant influence of total gaming / driving time for avg. FPS)


Test route #2: Sandpoint - Colville

[ external image ]

- Start at night around 1am
- (Perhaps) important mods: Weather 2.3, Grime's Early Autumun, Daniel's Random Events
- I directly included Spokane weight station (waypoint 2) into routing since it's 60-80% sure I'll catch it - so better including it directly for better comparison of the results.
- Always driving according to traffic rules (preferably constant influence of total gaming / driving time for avg. FPS)

[Shiva]

Thumbs up for the tests.