Performance in VR and FPS question

There is something I am not quite understanding re: Performance in DCS. I am just setting up my new VR machine (2080 Ti, i7-9700k, 32 MB memory, fast SSD etc.). The first 2 sims I installed are DCS and IL2-Great Battles. Both are looking great in VR and performance is silky smooth in both. In DCS I have PD set to 1.6 (so far) and the rest of the settings are balanced (some things on high, a few others still on medium and low) and I have no MSAA or SSAA.

IL2 is reporting in its in-game FPS counter that I am getting 90 fps almost all the time. DCS reports in its in-game fps counter that I am getting 30 fps all the time. It is 30 if I am alone in a clear sky or 30 if I am low in a city with lots of other planes and vehicles. In SteamVR the settings>>Video>>Application Resolution (global) defaults to 200% and everything else is default. So how can one be reporting 90 fps and the other 30 fps and both are performing equally well from a visual and "feel" point of view. Thanks for any input.

Can only say the DCS is getting some VR improvements soon, long story short there was a convergence of DCS versions last year and some changes under the bonnet which impacted VR performance.

Similar spec machine I run PD @ 1.2 and MSAA @ *4 and AF @ *16 and get pretty much constant 45FPS in game in built up areas close to the ground so you should be able to achieve that.

Maybe do some tweaking with settings and gauge CPU and GPU usage with taskmanager or something I aim to keep the GPU usage to average 80% it does ramp up and down. Also the maps have different overheads NTTR being the fastest least resource hungry and Normandy the most resource hungry.

As well different aircraft can impact on CPU/GPU usage.

HTH

There is something I am not quite understanding re: Performance in DCS. I am just setting up my new VR machine (2080 Ti, i7-9700k, 32 MB memory, fast SSD etc.). The first 2 sims I installed are DCS and IL2-Great Battles. Both are looking great in VR and performance is silky smooth in both. In DCS I have PD set to 1.6 (so far) and the rest of the settings are balanced (some things on high, a few others still on medium and low) and I have no MSAA or SSAA.

 

IL2 is reporting in its in-game FPS counter that I am getting 90 fps almost all the time. DCS reports in its in-game fps counter that I am getting 30 fps all the time. It is 30 if I am alone in a clear sky or 30 if I am low in a city with lots of other planes and vehicles. In SteamVR the settings>>Video>>Application Resolution (global) defaults to 200% and everything else is default. So how can one be reporting 90 fps and the other 30 fps and both are performing equally well from a visual and "feel" point of view. Thanks for any input.

If you’ve got 200% in SteamVR SS as well as PD 1.6 in DCS, that is a huge demand on your system. I use one or the other, but certainly not both together.

Thanks for the reply's to date. Yes, steam defaulted to 200% when I installed it (I guess because of the 2080 ti). Still not sure why I would want to reduce either the PD at 1.6 or the 200% when the performance and the graphics are so good (I am not complaining at all about DCS performance). If I had not turned on the fps counter in both games I just would have assumed they were both running at 90 fps. It also is still unclear to me why there is no difference visually (graphics or performance wise) when there is actually 60!! fps difference between the two. Something does not seem to add up here. Also should there not be some difference in fps in DCS given the graphic complexity in the scene (number of planes, number of objects, cities etc. etc.)?

OK, I get it now. Motion Smoothing is kicking in. SteamVr defaults to Smoothing ON globally and I can turn it on or off globally or per application. This is early days for me with this machine and I have only done a bit of testing but (so far) visuals are better with Motion Smoothing on with 200% Application Resolution in steamVR and a PD of 1.5 in the game than with 200% and a PD of 1.0. The only slight downside at the moment is a bit of propeller blur on the Yak-52 with Motion Smoothing but with jets it seems just fine. I actually did a few small setting changes and got Motion Smoothing to 45 FPS. So I can have 45 fps with it ON or 90 fps with it off when I drop the PD.

So either way I don't think I can lose. I can get good visuals and smooth performance without Motion Smoothing kicking in or I can get very good visuals and smooth performance with it on.

I can't really say as far as Steam VR goes as I haven't used it. Maybe someone with Steam VR experience could chime in.

But I would look at maybe turning SS etc off and use PD etc within DCS or vice versa and tweaking your settings so you can at least lock in 45FPS with minimal ASW or I guess Motion Smoothing.

I can't really say as far as Steam VR goes as I haven't used it. Maybe someone with Steam VR experience could chime in.

 

But I would look at maybe turning SS etc off and use PD etc within DCS or vice versa and tweaking your settings so you can at least lock in 45FPS with minimal ASW or I guess Motion Smoothing.

Ditto. Messing around with the oculus supersampling utility, it became clear you gotta turn on one or the other -- PD or SS -- but not both, they seem to be doing the same thing and are at least additive. Also, PD appears to have an exponential effect on GPU workload, backing that off a tad doesn't hurt your image nearly as much as it helps the GPU.

Some say the higher PD makes it harder to spot aircraft at a distance. (I can't see the durn things either way)

MSAA really helps text and aircraft images IMO. If I had your card, I'd start with a PD of 1.2, MSAA 4, I would turn off any external supersampling, and see what it looks like. Crank up the PD a little at a time till you like it.

Just posted this on "low end VR" settings

https://forums.eagle.ru/showthread.php?p=3755881#post3755881

I'm not using steam VR for DCS, but I am using pretty similar settings for IL-2 (Nvida settings) and most stuff set to high in IL-2 (except res, which is at minimum), but I'm getting stable 45fps at 120% Steam settings, which looks good.

DCS is getting some VR improvements soon

That may be overly optimistic .....but I hope your right ....I read they were coming but not the soon part :)

Oh c'mon surely soon, Wags has mentioned it, don't spoil the illusion. :D Although 2.5.4 seems better so far. :music_whistling:

Although 2.5.4 seems better so far. :music_whistling:

That is not a universally held opinion!

Oh c'mon surely soon, Wags has mentioned it, don't spoil the illusion. :D Although 2.5.4 seems better so far. :music_whistling:

LOL ....I'm with you! :D

That is not a universally held opinion!

True, 2.5.4 has been an interesting change to DCS World and I have observed different behaviour and performance and although I have totally muddied the waters by upgrading a GPU mid evaluation. Meh, over all for me it's getting a lot better. IMHO.

Better 45FPS less so PG and Normandy

Smoother

Less ASW

Shadows

Rift image quality getting better.

Can I has MSAA * 8?

:thumbup:

I'll qualify that,..

After tweaking for good performance and image quality on the Caucasus map (I've given up trying to make them all work kinda well) the other maps have some serious issues. Which is a shame as it's all the paid for maps that are not performing.*

So keeping settings constant.

NTTR is an aliasing nightmare.

PG and Normandy just gives constant stutters and frame rate drop.

Changing the 1080Ti to a 2080Ti just gave some more headroom/eye candy (4-Jan)

2.5.4 for my system in general can now use more CPU resources (yes I rolled back to 2.5.3 and confirmed it) and map dependent, DCS itself is faster to load feels more responsive and even logging onto MP servers seems faster. Initially after the update I could increase some GPU settings slightly with the 1080Ti over 2.5.3 however to my way of thinking it's the maps.

Now Nineline stated that there were no VR updates in 2.5.4 but in my experience and looking at the VR performance issue threads (even mine) something has affected overall performance and image quality.

*Map performance under 2.5.3 it was possible to get a reasonable compromise on all maps with careful tweaking but now I just give up.

I'll qualify that,..

 

After tweaking for good performance and image quality on the Caucasus map (I've given up trying to make them all work kinda well) the other maps have some serious issues. Which is a shame as it's all the paid for maps that are not performing.*

 

So keeping settings constant.

NTTR is an aliasing nightmare.

PG and Normandy just gives constant stutters and frame rate drop.

 

Changing the 1080Ti to a 2080Ti just gave some more headroom/eye candy (4-Jan)

 

2.5.4 for my system in general can now use more CPU resources (yes I rolled back to 2.5.3 and confirmed it) and map dependent, DCS itself is faster to load feels more responsive and even logging onto MP servers seems faster. Initially after the update I could increase some GPU settings slightly with the 1080Ti over 2.5.3 however to my way of thinking it's the maps.

 

Now Nineline stated that there were no VR updates in 2.5.4 but in my experience and looking at the VR performance issue threads (even mine) something has affected overall performance and image quality.

 

*Map performance under 2.5.3 it was possible to get a reasonable compromise on all maps with careful tweaking but now I just give up.

So what sort of settings are you running in VR (most things turned up I assume and PD at ??)

So what sort of settings are you running in VR (most things turned up I assume and PD at ??)

Yeah most things high, PD 1.4 works on Caucasus, but needs to be lowered for good performance on PG, Normandy.

Although NTTR PD can be and needs to be taken up to 1.8~2.0. just to bring the aliasing to something reasonable.

This seems to be a significant difference to 2.5.3

Having said that the Caucasus looks and preforms great I can get good performance form the other maps but that is with poorer visual quality in 2.5.4 compared to 2.5.3.

Now because there is such a spread in map performance rather than having a compromise "setting" I would need to have a setting for each map which seems odd. I do realise each map has it's own CPU/GPU usage requirements but these requirements have changed.

The PG map seems to be somewhat bugged with areas not even in built up areas where frame rates drop for no reason.

I'm sure it will get sorted by the publishers well maybe we are still waiting for Normandy to be fixed.

Maybe DCS needs some optimising but the other maps also seem to need some fixing as well.

Yeah most things high, PD 1.4 works on Caucasus, but needs to be lowered for good performance on PG, Normandy.

 

Although NTTR PD can be and needs to be taken up to 1.8~2.0. just to bring the aliasing to something reasonable.

 

This seems to be a significant difference to 2.5.3

 

Having said that the Caucasus looks and preforms great I can get good performance form the other maps but that is with poorer visual quality in 2.5.4 compared to 2.5.3.

 

Now because there is such a spread in map performance rather than having a compromise "setting" I would need to have a setting for each map which seems odd. I do realise each map has it's own CPU/GPU usage requirements but these requirements have changed.

 

The PG map seems to be somewhat bugged with areas not even in built up areas where frame rates drop for no reason.

 

I'm sure it will get sorted by the publishers well maybe we are still waiting for Normandy to be fixed.

 

Maybe DCS needs some optimising but the other maps also seem to need some fixing as well.

Man, thats nuts, I'm running a 980M (mobile 980, about like a desktop 970) with VR, A few things high (textures), a few things off (shadows/msaa), and PD 1.2... You have approximately 3-4x the rendering HP I do and it doesn't sound like your visuals are all that much better than what I get. I've cranked it to 1.4PD and it didn't look dramatically better (killed my frames tho) Are you getting the full 90 frames? I'm still on 2.5.3...

Man, thats nuts, I'm running a 980M (mobile 980, about like a desktop 970) with VR, A few things high (textures), a few things off (shadows/msaa), and PD 1.2... You have approximately 3-4x the rendering HP I do and it doesn't sound like your visuals are all that much better than what I get. I've cranked it to 1.4PD and it didn't look dramatically better (killed my frames tho) Are you getting the full 90 frames? I'm still on 2.5.3...

I don't know Caucasus looks pretty good. :D

Cough, 90FPS, I haven't seen 90FPS in game since 2.3.x with a 4GHz cpu and a 980Ti with shadows, that's a year now since deferred shading became a thing with 2.5. Indeed last night I was basking in the joy of having shadows on buildings and in the cockpit it's been a while but that has taken 2 GPU upgrades and O/Cing to achieve albeit now at 45FPS. :(

I'm pretty sure I could crack 90FPS but I'm not prepared to reduce resolution or fly at 20000ft or both to achieve it. I'm okay with 45FPS (been like that for a year now) as long as it's smooth and constant. :)

However the point I was making is for me it seems the other maps have taken a rather large performance hit in 2.5.4.

Thanks for the reply's to date. Yes, steam defaulted to 200% when I installed it (I guess because of the 2080 ti). Still not sure why I would want to reduce either the PD at 1.6 or the 200% when the performance and the graphics are so good (I am not complaining at all about DCS performance). If I had not turned on the fps counter in both games I just would have assumed they were both running at 90 fps. It also is still unclear to me why there is no difference visually (graphics or performance wise) when there is actually 60!! fps difference between the two. Something does not seem to add up here. Also should there not be some difference in fps in DCS given the graphic complexity in the scene (number of planes, number of objects, cities etc. etc.)?

If you’ve got 200% in SteamVR SS as well as PD 1.6 in DCS, that is a huge demand on your system. I use one or the other, but certainly not both together.

Hi, I came here specifically to explain and say the following.

If you have let's say an Oculus Rift CV1 and you are not using Super-Sampling (So Pixel-Density/PD set to default 1.0 and Steam set to 100% as well) your rig will have to spit out 1080×1200 pixels, times TWO (one for each eye).

-1080 X 1200 X 2 = 2,592,000 pixels total per frame generated for both eyes combined = 100% (Let's call this 100% since this is the default demand on PD 1.0, no Super Sampling)-

If you set P.D to 1.6 in DCS and keep Steam at 100%/no SS in Steam, then the total rendered SS resolution per frame, per eye, will be default resolution (written above) x 1.6:

1080 X 1.6 = 1728

1200 x 1.6 = 1920

~1728 X 1920 X 2 = 6,635,520 pixels in total (1.6 PD = 257% demand on your rig when compared to the 100%/default setting of 1.0 PD)~

When you also set Steam to 200% on top of your already 160% SS in DCS, you're essentially applying Super-Sampling on top of Super-Sampling (bad idea), before it is downsampled back to the native resolution of the two screens inside of your headset (1080 x 1200 each) and shown to you.

So in your case, 160% SS on top of 200% SS will yield:

~ 3456 x 3840 x 2 = 26,542,080 pixels total per frame (More than 1,000%- TEN TIMES the default 1.0 PD resources are required from your rig and mainly vidcard!!)~

Does it also provide you with a 10x better graphical experience? Cus it's using 10x as much computing power to deliver those frames.. My guess is not.

Ultimately it's up to you if this sacrifice is worth it or not, but if I were in your place I would definitely give up the SuperDuperSuperSampling for a smoother constant 90FPS experience with a lower ONE-LAYER PD boost :doh: :joystick:

The "Sweet Spot" in terms of PD in DCS is 1.8 to 2.0. Any perceived improvements above 2.0 is most likely placebo; so my advice to you (and anyone else reading this, with a beast GFX card in their rig) would be to use either SteamVR's Super-Sampling and set it to 180% to 200% or (better yet) use DCS' own PD slider and set it between 1.8 and 2.0.

1.8 PD =

1944 X 2160 X 2 = 8.398.080 Pixels (324% Compared to Default 1.0 PD)

2.0 PD =

2160 X 2400 X 2 = 10,368,00 Pixels (400% Compared to Default 1.0 PD)

I'm forced to do 1.0 (No super-sampling at all) and still it's not buttery smooth on my R9 390, 5820K & 16GB DDR4 :cry:

I wish I had even half of the graphical resources some of you have.

No more 45fps 'allday 'errday and no more choppy game-play with a sidewinder hot on my tail, or worse a flanker moving in for the gun kill :pilotfly:

Also, MSAA is very resource intensive (It's a form of SS also) and so are shadows in DCS VR (don't use high and up in settings). Be mindful of that also if FPS and smooth gameplay is what matters most to you.

Edit: The number of pixels in this post are referencing the Oulus Rift CV1, and will differ from headset to headset, but the (increase in) percentages will be the same, irrespective/regardless of what headset or resolution you are on.

Edit: PD = Super Sampling and Super Sampling = PD. The only real difference are the words themselves, Pixel Density & Super Sampling.

Last-Edit-i-Promise: As imacken will point out in the next page of this thread, it seems that SteamVR will now show your headset's native resolution at 50%, not 100%... Now... At 100%... You will be already SuperSampling quite heavily it seems. My advice is to just avoid playing around with Super Sampling in Oculus Debug Tool and Steam VR and keep SS strictly within DCS' own PD option.

Cheers!

IT--

oops. now you have done it.

facts and math brought to the discussion? tsk, tsk.

boys! get the pitchforks!!

:)

Hi, I came here specifically to explain and say the following.

 

If you have lets say an Oculus Rift CV1 and you are not using Super-Sampling (So PD set to default 1.0 and Steam set to 100% as well) your rig will have to spit out 1080×1200 pixels, times TWO (once for each eye).

-1080 X 1200 X 2 = 2,592,000 pixels total per frame generated (Both eyes combined)-

 

If you set P.D to 1.6 in DCS and keep Steam at 100%/no SS in Steam, then the total rendered SS resolution per frame, per eye, will be default resolution (written above) x 1.6:

1080 X 1.6 = 1728

1200 x 1.6 = 1920

-1728 X 1920 X 2 = 6,635,520 pixels in total- (257% Increase in pixels/demands on your gfx-card and system compared to default setting of 1.0)-

 

When you also set Steam to 200% on top of your already 160% in DCS, you essentially be applying Super-Sampling on top of Super-Sampling, before it is downsampled back to the native resolution of the two screens inside of your headset (1080 x 1200 each) and shown to you.

So in your case, 160% SS on top of 200% SS will yield:

- 3456 x 3840 x 2 = 26,542,080 pixels total per frame (More than 1,000% increase~ TEN TIMES the default 1.0 PD resources are required from your rig and mainly vidcard!!)-

Does it also provide you with a 10x better graphical experience? Cus it's using 10x as much computing power to deliver those frames.. My guess is not.

 

Ultimately it's up to you if this sacrifice is worth it or not, but if I were in your place I would definitely give up your Super-SuperSampling for a smoother constant 90FPS experience with a lower ONE-LAYER PD boost :doh: :joystick:

 

The "Sweet Spot" in terms of PD in DCS is 1.8 to 2.0. Any perceived improvements above 2.0 is most likely placebo; so my advice to you (and anyone else reading this, with a beast GFX card in their rig) would be to use either SteamVR's Super-Sampling and set it to 180% to 200% or (better yet) use DCS' own PD slider and set it between 1.8 and 2.0.

 

I'm forced to do 1.0 (No super-sampling at all) and still it's not buttery smooth on my R9 390, 5820K & 16GB DDR4 :cry:

 

I wish I had even half of the graphical resources you have.

No more 45fps 'allday 'errday and no more choppy game-play with a sidewinder hot on my tail, or worse a flanker moving in for the gun kill :pilotfly:

 

Also, MSAA is very resource intensive (It's a form of SS also) and so are shadows in VR. Be mindful of that also when FPS is what matters.

 

Cheers!

 

IT--

Thanks for that.

Let me just say a couple of things:

1) You are incorrect about SteamVR SS at 100% being no SS. 50% SteamVR SS is 'no SS'. At 100% SteamVR SS, on my Vive Pro, my output is 2016x2240, and 1426x1584 at 50%. Native res is 1440x1600. See attached screen grabs.

2) the days of 90fps disappeared with deferred shading, I'm sorry to say. A rock solid 45fps is Nirvana these days! Even with low DCS settings, PD set to 0.5 and SteamVR SS set to 20%, 90fps is nowhere to be seen!

After experimenting over a long period of time, there is no doubt in my mind that MSAA and PD are the only real critical factors in performance on a decent kit. Sure, shadows and visibility range affect fps a bit, but it's MSAA and PD that are the big boys!

Thanks for that.

Let me just say a couple of things:

1) You are incorrect about SteamVR SS at 100% being no SS. 50% SteamVR SS is 'no SS'. At 100% SteamVR SS, on my Vive Pro, my output is 2016x2240, and 1426x1584 at 50%. Native res is 1440x1600. See attached screen grabs.

2) the days of 90fps disappeared with deferred shading, I'm sorry to say. A rock solid 45fps is Nirvana these days! Even with low DCS settings, PD set to 0.5 and SteamVR SS set to 20%, 90fps is nowhere to be seen!

After experimenting over a long period of time, there is no doubt in my mind that MSAA and PD are the only real critical factors in performance on a decent kit. Sure, shadows and visibility range affect fps a bit, but it's MSAA and PD that are the big boys!

You're more than welcome, but I trust that this was not new info for you =)

I just checked out SteamVR on my CV1 and you are absolutely right. Haven't used SteamVR for some months now so I was completely unaware of this change.

It seems that after the update, the old 100% has become the new 50%.. I don't understand why they would do this though... It makes no sense to me; why label something 50% ("per eye") when it is in fact 100% (The native reso)?

Thank you for setting it straight mate! I will be sure to modify my post and add a disclaimer.

Silly Gaben doing silly stuff instead of giving us HL3... :P

Cheers! :smartass:

IT--

You're more than welcome, but I trust that this was not new info for you =)

I just checked out SteamVR on my CV1 and you are absolutely right. Haven't used SteamVR for some months now so I was completely unaware of this change.

 

It seems that after the update, the old 100% has become the new 50%.. I don't understand why they would do this though... It makes no sense to me; why label something 50% ("per eye") when it is in fact 100% (The native reso)?

 

Thank you for setting it straight mate! I will be sure to modify my post and add a disclaimer.

 

Silly Gaben doing silly stuff instead of giving us HL3... :P

 

Cheers! :smartass:

 

IT--

I know, it's absolutely crazy. The only possible reason I can think of is that 50% x 2 (eyes) gives you 100%. But, that's a stretch!

Other things that concern me about this slider are:

1) most people are probably not even aware of its existence, and

2) on install, it defaults to what it thinks is sensible for the GPU. In my case, that is 200%!

A lot of users could be stretching their system beyond sensible when running DCS.

Yeah but that 200% is probably fine for the vast majority of VR games. DCS needs some serious fixing with regard to VR.

Yeah, I'm not seeing huge increases in "quality"/clarity from pd1.0 to 1.2 to 1.4 ish which is the upper limit of what I can run and still pull 45 frames. MSAAx2 seems to help mainly be eliminating shimmer/artifacts. Recently got the shader mod, which freed up enough resources to go from PD1.2 to 1.3 and turn MSAAx2 on.