2080Ti vs RTX 3090 for VR

[speed-of-heat]

yep

[Gunnars Driver]

Minor correction guys, motion smoothing doesn’t force 45Hz or 90Hz, it forces 45 or 90 fps at 90Hz.

 

If we need to be really correct in ”auto” we have no motionsmoothing or forcing at 90fps.

If 90fps, motionsmoothing is off and we keep the vsynced 90.

below 90fps but not lower than 45, we force 45fps, and motion smoothing is on.

 

 

[imacken]

Really not going to get into that!

[dburne]

Technically I think it is really Motion Reprojection.

But whadda I know, I am still new to all this WMR/Steam VR stuff...

 

[Gunnars Driver]

Really not going to get into that!

 

:)

 

 

[Gunnars Driver]

Technically I think it is really Motion Reprojection.

But whadda I know, I am still new to all this WMR/Steam VR stuff...

 

If you use the WMR version its ”motion reprojection”.

If you use stemVR version its ”motion smoothing”.

If you have oculus its ”ASW”.

 

Different name, same basics but not exatly the same. Some are better some are worse.

 

 

[imacken]

 

If you use the WMR version its ”motion reprojection”.

If you use stemVR version its ”motion smoothing”.

If you have oculus its ”ASW”.

 

Different name, same basics but not exatly the same. Some are better some are worse.

 

 

We are talking about Reverb/WMR. WMR headsets don't use SteamVR's motion smoothing but within SteamVR the WMR settings tab says the attached.

[Gunnars Driver]

 

We are talking about Reverb/WMR. WMR headsets don't use SteamVR's motion smoothing but within SteamVR the WMR settings tab says the attached.

 

[Edit]Delited, just found out StramVR motion smoothing is in fact motion reprojection.

Didnt know, or had forgotten...

 

 

 

 

 

[imacken]

 

[Edit]Delited, just found out StramVR motion smoothing is in fact motion reprojection.

Didnt know, or had forgotten...

 

 

 

 

 

SteamVR Motion Smoothing is not supported by WMR headsets.

[Gunnars Driver]

 

SteamVR Motion Smoothing is not supported by WMR headsets.

 

Yep, just saw that. Maybe knew but had forgotten.

[Supmua]

My Asus TUF OC 3090 came in, haven't done any DCS VR testing just overclocking and some 3DMarks benchmarking to test stability. It is very quiet compared to my Asus Dual 2080Ti which requires high fan speed to maintain good temperature, also seems to run cooler as well (65C max temp with mild-moderate overclocking). Not planning to do any serious testing til weekend, just going to play and play.

 

Edit: Initial impression is very positive. Free flying missions with F/A-18C in Caucasus, PG, Syria with default VR preset (PD 1.0, Steam SS 200%) I'm getting 80 fps in 80Hz mode on my Valve Index which means reprojection free (sharp hard turn at low altitude in Syria would still drop my FPS to 40 for a few seconds). This is with motion smoothing on so 80 is likely not my max fps. This kind of framerate never happens with my 2080Ti. The only mission I tested that gave me trouble is the Supercarrier launch, in external view F2 spin the camera toward the tower and fps would just bottom out to 40 or even lower. In MSFS, I could run high-end preset with 90% render scale and maintain 4K60 fps in downtown NYC with Cessna 152 (had to drop to 80% scale for night flying); and the developer fps windows only showed main thread limited now instead of GPU. So, with this card I'm no longer GPU bound in these tests. Deep dive testing will come when time permits.

[Sn8ke_iis]

My Asus TUF OC 3090 came in, haven't done any DCS VR testing just overclocking and some 3DMarks benchmarking to test stability. It is very quiet compared to my Asus Dual 2080Ti which requires high fan speed to maintain good temperature, also seems to run cooler as well (65C max temp with mild-moderate overclocking). Not planning to do any serious testing til weekend, just going to play and play.

 

Edit: Initial impression is very positive. Free flying missions with F/A-18C in Caucasus, PG, Syria with default VR preset (PD 1.0, Steam SS 200%) I'm getting 80 fps in 80Hz mode on my Valve Index which means reprojection free (sharp hard turn at low altitude in Syria would still drop my FPS to 40 for a few seconds). This is with motion smoothing on so 80 is likely not my max fps. This kind of framerate never happens with my 2080Ti. The only mission I tested that gave me trouble is the Supercarrier launch, in external view F2 spin the camera toward the tower and fps would just bottom out to 40 or even lower. In MSFS, I could run high-end preset with 90% render scale and maintain 4K60 fps in downtown NYC with Cessna 152 (had to drop to 80% scale for night flying); and the developer fps windows only showed main thread limited now instead of GPU. So, with this card I'm no longer GPU bound in these tests. Deep dive testing will come when time permits.

 

Thanks Supmua, this is good info. Looking forward to your data.

[javelina1]

 

Thanks Supmua, this is good info. Looking forward to your data.

 

me too, I have a 3090FE inbound. :)

[RogueSqdn]

I got my new system up and running last night and posted my first impressions in my cockpit thread:

 

Well, I just got my first few flights in instant action with the new system. This is still with the Rift S, the G2 will be here Monday.

 

I can turn settings way up, use OTT to force ASW off, and rarely get below 60 fps, and most of the time peg 80. This is with 1.5 supersampling in OTT and none in DCS, the IC shaders mod, and visibility range to ultra. Before I averaged around 30 fps in the air and 15 on the ground, with most settings turned down.

 

So the FPS and increased settings are the good thing. The bad thing is resolution is still limited since it's the Rift S, and there won't be anything I can do about that for a few days. :)

 

I'll see in the morning how it does with a mission with a lot of objects in it.

 

Specs for both the old system (Mach V) and the new one (Talon) are in my sig.

[Nahemoth]

Hi,

 

For those that already have a 3090, it has been a while since the first impressions, and I assume that it is somehow consolidated the experience... I have read about really good improvements with the 3090, but does it worth the upgrade from a 2080Ti?

 

I'm using a Reverb (the G2 is comming), and with my current setup (9700K + 2080Ti) I experience some stuttering with the detail settings not so high (MMSA off), and always bellow 45 FPS.

 

Thanks!

[Gunnars Driver]

Hi,

 

For those that already have a 3090, it has been a while since the first impressions, and I assume that it is somehow consolidated the experience... I have read about really good improvements with the 3090, but does it worth the upgrade from a 2080Ti?

 

I'm using a Reverb (the G2 is comming), and with my current setup (9700K + 2080Ti) I experience some stuttering with the detail settings not so high (MMSA off), and always bellow 45 FPS.

 

Thanks!

 

I have the Reverb Pro(”G1) and while I did not have any troubles with the 2080ti, tuned for smooth ride at 45 fps reprojection I see a huge increase in performance with the 3090. My settings was tined with the 2080 ti so I more or less never see any stuttet( I dont like gaming with a non smooth experience).

 

For me, the 2080ti was always the limiting factor and the 3090 made a quite big difference and now the performance gap is less between the CPU(9900KS@5.2) and the GPU. Still almost always higher frame times for the GPU.

 

Havent tuned the settings to max for the Reverb, as I did get the 3090 mostly for pimax 8KX where the 3090 is indeed needed in DCS.

 

I would say that you should be able to get it running well with the 2080ti + the Reverb. There shouldnt be any real difference in hardware demands between G1 and G2 as the use the same resolution.

So my conclusion is that you should be able to get it running well with the 2080ti, if all computer hardware/software is well tuned. If tou have the money for the 3090 it will be a big lift and you should be able to use higher settings for a better picture/ gaming experience. But with the weong settings Im sure it is possible to bog down the 3090 as well.

 

Also, many here report that they are CPU bond. I havent seen this for my self, but one poster recently deleted mods and after that his computer delivered the about same frame times as I have.

 

If you by some reason is highly CPU bond, you might not see that big improvement with the 3090. In my case the CPU frame times decreased slightly so it might help in that case also.

 

 

 

 

[speed-of-heat]

So the only instance I remain CPU bound is in the super carrier cold start instant action mission, when the sheer number of AI planes on the deck, over 16 or so hornets, which makes for a crowded but CPU bound deck i have been using:

 

https://forums.eagle.ru/showthread.php?t=269320

 

options.graphics.stereo_mode_use_shared_parser = true

 

this is very experimental but it does have benefits that out weigh the experimental nature IMO (now) which gets me a regular 45fps on the deck of the carrier even when "looking" at the superstructure on a crowded deck.

 

The mods i think were a red herring (at least for me), as they are now all re-installed and i get regularly 90FPS in flight/combat /missions usually more often than i get 45fps, but i never see a drop below 45FPS

[RogueSqdn]

Reposted from the DCS Liberation thread:

 

So I've received my new system and Reverb G2 and have spent the past several days tuning the settings in for the best performance/visuals balance, with a 5950X and 3090FE card. In sparse areas I get 90 fps, but in crowded areas like Dubai, or the FEBA during a Liberation campaign, it's between 45-55 fps.

 

I spent time yesterday wiping my old MachV clean, setting up file sharing and RDC, and installing DCS OpenBeta server. I managed to get it working where I could copy a Liberation mission over, but couldn't get it to write state.json until I saw this little tidbit on GitHub:

 

For instance, if on your computer, the DCS Liberation directory is in : D:\DCS Liberation 2.0RC9, make sure to create the same directory on the server. (If you have the possibility to do so.)

 

After creating that on the server, state.json is written and I'm able to copy it back to the Talon system to update Liberation's status. :thumbup:​

 

My FPS test is the FEBA of the Golan Heights campaign, just north of the Sea of Galilee, and by playing it as a multiplayer mission, my FPS went from 45-55 to 55-65. Now I just have to decide if the hassle and the loss of time acceleration is worth it. ;)​

 

This is with the parsing turned off again, I'll have to try activating it and seeing what happens (though it feels good to have normal cockpit lighting back again). Frametimes over the FEBA are around 17-18 in fpsVR with the mission being server hosted. I'd like to get that down a little bit but I'm not sure what to do without sacrificing quality.

 

 

 

 

 

Thoughts?

 

 

[Gunnars Driver]

So the only instance I remain CPU bound is in the super carrier cold start instant action mission, when the sheer number of AI planes on the deck, over 16 or so hornets, which makes for a crowded but CPU bound deck i have been using:

 

I'm still not CPU bond on that [Mission qualification -Supercarrier cold start]. Down below, datalogging from mission start with manual engine start and taxi - catapult - take of.

I did use motion reprojection as I most always do, without the mean fram rate had been higher as Im either close to 90 fps(frame times about 12ms) or I get 90fps.

In this test I did not stare at the tower, and did not use F2 and circel around the tower. Just engine start as I would do and taxi -take off.

 

I can produce CPU frame time about 23-24ms by starring at the tower. But for all gaming inside aircraft I never go below 45, and mostly get 90 fps.

( SS 100% with the reverb G1, as I also use Pimax 8KX and there 150% is a bit high. I dont think SteamVR SS loads the CPU noticable so it might not make a big diff.)

 

 

fpsVR Report:

App: DCS HMD: HP Reverb VR Headset VR1000-2xxx0 (90.020 Hz, IPD 65.5)

GPU: NVIDIA GeForce RTX 3090 (27.21.14.5730) CPU: Intel® Core i9-9900KS CPU @ 4.00GHz

Delivered fps: 52.57 Duration: 5.8min.

GPU Frametimes:

Median: 10.5 ms

99th percentile: 14.3 ms

99.9th percentile: 19.2 ms

frametime <11.1ms(vsync): 74.6%

CPU frametime:

Median: 9 ms

99th percentile: 16.4 ms

99.9th percentile: 17.6 ms

frametime <11.1ms(vsync): 65.4%

Max. SteamVR SS: 100%

Render resolution per eye: 2208x2160(by SteamVR settings, Max.) (HMD driver recommended: 2208x2162)

 

 

 

[Gunnars Driver]

Reposted from the DCS Liberation thread:

 

 

 

This is with the parsing turned off again, I'll have to try activating it and seeing what happens (though it feels good to have normal cockpit lighting back again). Frametimes over the FEBA are around 17-18 in fpsVR with the mission being server hosted. I'd like to get that down a little bit but I'm not sure what to do without sacrificing quality.

 

[ATTACH=JSON]{"data-align":"none","data-size":"custom","height":"392","title":"Screenshot 2020-11-28 060447.png","width":"647","data-attachmentid":7155781}[/ATTACH]

 

[ATTACH=JSON]{"data-align":"none","data-size":"custom","height":"389","title":"Screenshot 2020-11-28 060601.png","width":"644","data-attachmentid":7155782}[/ATTACH]

 

[ATTACH=JSON]{"data-align":"none","data-size":"custom","height":"502","title":"Screenshot 2020-11-28 060830.png","width":"641","data-attachmentid":7155783}[/ATTACH]

 

 

Thoughts?

 

 

All settings will always be a compromise between quality and fps, in DCS. In other games you might be able to use all settings to the max.

 

I suggest to start with using no supersampling. That will be looking at the resolution and not the percentage. Go as close as possible to 2160x2160. This is to get the fps running and when it does you can tune the SS up step by step.

The about 2700 x2700 you run now is actually about 160%. Which is a bit high if you combine it with other GPU-killing settings.

The supersampling settings multiply so you need to keep control of the universal settings also. If you set 150% SS in the universal and 150% in per application setting/ DCS you end up with 225% !

 

I have always used PD 1.0 in DCS togheter with Reverb/SteamVR SS. At least in this case you shouldnt need to set it lower than 1.0

 

I use shadows : Flat, instead of low.

MSAA, actually Supersampling does about the same thing when it coems to jagged lines and combined its a hard hit on the GPU.

Start with MSAA OFF and test if you like 150% Supersampling(about 2645x2645) better than MSAA. You probably need to choose, or go low on both. I did choose to use high SS and no MSAA witht he 2080ti I had earlier.

 

Clutter/grass has a impact, suggest lower it a quite much for starters, try 30%.

Trees visibility also have a impact but as we do not want to see the threes grow close to the nose, suggest start with 50%, and then increase is, as one of the last things to increase.

Smoke/chimney density: do you need to see this, otherwise set it to lowest.

 

 

 

 

[RogueSqdn]

I'll try the supersampling... but what should I leave the universal settings at? 2160?

[speed-of-heat]

@Gunnars Driver , they only time I am cpu bound is when I stare at the tower...otherwise I have again similar results tomyou

[Gunnars Driver]

I'll try the supersampling... but what should I leave the universal settings at? 2160?

 

I think setting both to 2160x2160(or as clos as possible) would be best. am a little bit unsure, because of the SteamVR settings for the G2.

 

Perhaps best to use fpsvr to get a reading of the reported SS, and recalculate from Steam vr 100% resolution setting.

What resolution is reported at 100% on the slider with the G2 ?

 

I dont know if installing the beta version of steamVr fix this for the G2, it did that when Reverb G1 was new.

[RogueSqdn]

I installed the SteamVR beta, FWIW.

 

That said, if I leave the universal at 50% and put the app setting at 200%, fpsVR reports 99%, and my frame times over the FEBA go down to around 10-11, but are still around 18 in cities. This is in a Liberation mission over the Golan Heights, hosted by my old PC, with smoke turned off on the front (which also made a big difference).

 

I don’t think hosting Liberation on the old PC is helping much. CPU frame times improve, but only 5-10ms, and frame rate doesn’t change.

 

if I run the mission locally, CPU usage is only around 9% and system RAM usage is about 24/64 GB.

 

GPU is using 20/24 GB over the FEBA.

 

I’d like to get rid of shimmering distant buildings, but 4x MSAA and 16x AF doesn’t seem to help much. The biggest culprit that shimmers are small buildings. If I use VR zoom on them the shimmering goes away.

[Lurker]

So I went ahead and ordered a 3090, yeah I splurged a little bit I hope I don't have buyers remorse lol. 

 

In any case, any tips for installing the graphics card and drivers so I have the least possibilities of problems? Im coming from an 2060 so I should see a pretty big improvement in any case. Should I first download the latest Nvidia drivers and then install the card, or the other way around?