0.8 Rendering Scale && 300% Supersampling && AA

[Gomoto]

If you own a HP Reverb. Try! And dont tell ?

 

Do not wait a day like I did when I heard about the same trick in DCS.

 

We are getting somewhere with VR. I wanted my VR look like my monitor. Now I want my monitor look like my VR ?

[-332FG-Gordon200]

Are you using the non-beta WMR for SteamVR?

In the non-beta version 300% is equal to 124% SS in the beta version which is also the recommended setting for a 2080ti gpu.

https://docs.google.com/spreadsheets/d/1gJmnz_nVxI6_dG_UYNCCpZVK2-f8NBy-y1gia77Hu_k/edit#gid=1836884624

[SCG_motoadve]

How do I decrease rendering scale to 0.8?

 

[Gomoto]

300% Supersampling equals 3816x3740x2=28.543.680 Pixels in my SteamVR version.

 

In game rendering scale is called "Dynamic resolution factor" in the settings.

 

I did not believe this works, until I tried it in DCS where I read about this first. It also gives remarkable results in IL2, especially when combined with antialiasing.

 

You would think the 0.8 resolution factor would degrade the image quality, but the 300% supersampling produce a very nice and smooth combined result. You would further expect the 300% to totally crush your FPS, but in fact you achieve higher and smoother FPS, remarkable this is especially notable for CPU bound situations as well. I really recommend any SteamVR user to give this a try. 

 

 

Edited May 8, 2020 by Gomoto

[NiiranenVR]

i need ' for dummies' 

 

In beta I have my SS at 202 % for have the same as you 

The AA - how high 2 or 4 

 

I'll try it out 

[SCG_Fenris_Wolf]

Another glorious idea. Play at 40fps, enjoy 3008p. I don't get how people don't realize they are just applying multiplicators and not magic recipes.

 

Well, have fun!?

 

P.S. Also, enjoy your now-cut rendering bubble, and getting bounced by close to invisible aircrafts. ?

[Gomoto]

I am using SteamVR 1.11.11 (non beta SteamVR and Windows Mixed Reality for SteamVR beta) 

I use 2x AA

(I personally prefer to not use HDR, but high shadows)

 

I read about the rendering scale/supersampling combo here:

https://forums.eagle.ru/showthread.php?t=272325

Edited May 8, 2020 by Gomoto

[NiiranenVR]

Well - it's no success for me n it's make my game laggering ....so ......?

[Gomoto]

Perhaps it is very system related. I get a 20% to 50% Framerate increase in DCS, IL2 and Assetto Corsa Competizione.   

[Dutch2]

8 hours ago, Gomoto said:

300% Supersampling equals 3816x3740x2=28.543.680 Pixels in my SteamVR version.

 

In game rendering scale is called "Dynamic resolution factor" in the settings.

 

I did not believe this works, until I tried it in DCS where I read about this first. It also gives remarkable results in IL2, especially when combined with antialiasing.

 

You would think the 0.8 resolution factor would degrade the image quality, but the 300% supersampling produce a very nice and smooth combined result. You would further expect the 300% to totally crush your FPS, but in fact you achieve higher and smoother FPS, remarkable this is especially notable for CPU bound situations as well. I really recommend any SteamVR user to give this a try. 

 

 

 

 

DRF in BoX is working, to notice its effects: Shift dynamic resolution factor to lowest adjustments, start a flight check at your cockpit instruments. Now in the air turn your cockpit to the clouds check again you will notice the especially the instruments, are become very fuzzy. In my case I even thought it is the damp on my VR lenses. 

But 300% SS thats huge, you need to set the DRF to very low, the  motion reprojection capped to 45FPS and all the ingame adjustments to low.  But interesting I’m sure I will give it a go. 

edit: What CPU/GPU are you running? 

Edited May 8, 2020 by Dutch2

[Gomoto]

1080ti & 3600X

Like expected it also helps in multiplayer, same fps on ground like in single player, I am not used to this. ?

[VR-DriftaholiC]

8 hours ago, SCG_Fenris_Wolf said:

Another glorious idea. Play at 40fps, enjoy 3008p. I don't get how people don't realize they are just applying multiplicators and not magic recipes.

 

Well, have fun!?

 

P.S. Also, enjoy your now-cut rendering bubble, and getting bounced by close to invisible aircrafts. ?

 

If you render at 0.5 and super-sample at 200% the end result should be the same as rendering at 100% and not super sampling. I'm with Fenris on this, I don't think anything is happening here other then a little placebo.

[Dutch2]

2 hours ago, driftaholic said:

 

If you render at 0.5 and super-sample at 200% the end result should be the same as rendering at 100% and not super sampling. I'm with Fenris on this, I don't think anything is happening here other then a little placebo.

 

Think his approach is something different then a normal workaround we usually read here. I could be complete wrong inhere, but this is my theory in what he is trying to accomplish in BoX-VR. 

Lets say he is running 200 % ss, this should be no problem when everting is perfect, we have all periods during a flight everything is straight 90fps, GPU usage is showing low in fpsVR and yes then we could run 200%ss. If its performance must increase to maintain the 90FPS, then because of the 0,7 DYNAMIC resolution factor its resolution is decreasing to something that can be compared to 140%. So what he is doing is not setting up the adjustments for average on what we could expect, no he set the SS to max, and when performance is needed the DRF shifts this to the normal advised 140%. 

 

 From @SCG_Fenris_Wolf topic in case of an Odyssey: “Super Sampling 140% (GTX1080, 7.1M pxl @90Hz) “ you see as soon as with the 200% SS, it can not reach the 90FPS its shifts back to the advice resolution of Fenris. 

 

Sure I have to test this by myself to see if my theory is right. 

 

Edited May 8, 2020 by Dutch2

[dburne]

4 hours ago, driftaholic said:

 

If you render at 0.5 and super-sample at 200% the end result should be the same as rendering at 100% and not super sampling. I'm with Fenris on this, I don't think anything is happening here other then a little placebo.

 

This...

One should SS in one or the other, not both.

But of course everyone is free to adjust to anything they want.

Edited May 8, 2020 by dburne

[NiiranenVR]

I can  only tell about my pc , but when I tried the 200 % Ss and AA 2/4 and the DRF 0.5-0.8 it makes no good and the game was laggering , so I'm back to my normal setting 

 

SS 100 % - AA = 0 and DRF = 1 

 

I have had my Ss to 188 in my beta , but red some place that it could be more difficult to spot a plane / more SS  - and the difference is not big so I'm fine with 100% SS 

[WheelwrightPL]

So it was likely all a placebo. I hope the upcoming deferred engine will provide a sorely needed REAL boost (despite me having 2800TI) so I can finally run my Reverb without reprojection at a steady 90Hz/sec. That's probably unrealistic expectation though, so I will likely go back to the pancake screen. But I won't despair  because objectively speaking VR and Non-VR each have advantages/disadvantages, therefore currently I am of the opinion that there is no clear-cut winner. If you want better experience go for VR, if you want better game go for TrackIR. It  is about priorities or even how one feels at any given moment.

[Gomoto]

The idea is in CPU limited situations you sacrifice graphic fidelity, gain some frames because of lower CPU frametime. Because the situation is CPU limited your GPU has some headroom and you use this for supersampling to gain graphic fidelity.

 

(By 0.8 render scale the picture is not rendered at 1.0 and then downsampled by the GPU. It is rendered at the lower resolution in the first place with less strain on both GPU and CPU)

 

Overall you loose some detail by the 0.8 rendering scale that will not magically return. But the resulting picture will be very smooth and pleasing because of the 300% supersampling. The text markings on the instruments and switches look much cleaner. Detail is just one aspect of picture quality.

 

Definitly this is not a placebo. You destroy/sacrifice some picture information, beautify the remaining information to make the loss more acceptable and benefit from the smoother looks as side effect while striving for the cpu performance gain.

 

(Main CPU bound discomfort for me: Lower framerates on the ground, over buildings, many planes or vehicles around. And exactly here you experience an improvement. On the other hand if you are never CPU bound, but who is not in IL2.)

 

 

There is even a philosophical aspect. Some people hate to sacrifice something now to gain something later. Especially very young people aka children have to learn this first. Some never learn to let something go. ? 

Edited May 9, 2020 by Gomoto

[VR-DriftaholiC]

It makes a little more sense to allow dynamic resolution factor to activate when FPS drops low. In my testing the amount of pixels rendered were never much of a factor for my system and FPS. I can run 2460x2740 or 1400x1600 and get within a few FPS of either setting and still see drops into the 40's when on the ground or lots of complex objects like bombers or single player AI is working. So for me being able to dynamically lower the resolution doesn't help maintain frame rate as I'm CPU limited. When I run the frame time analyzer I see the GPU getting late starts in these situations because the frame wasn't ready for the GPU in time. 

[Gomoto]

That is a good idea i will try dynamic resolution factor to further improve my experience.

Target FPS seems not to work with VR. Or am I missing something?

[Dutch2]

My impression after two short test runs pure as an indication, two with and two without. 

It does work in the way I thought it could be, only problem as it seems:

• the DRF is more like a on/off switch and not something thats works dynamically in 0,1 increments, from 1 to 0,5. 
• It has no sense running 300% while 80% of your flying time is showing very low FPS, in this perspective its more for the 9900k/2080ti users then the midcap users like my 9700k/GTX1080 combo. The difference between 150% and 180%, is not wow factor. 
• Another point Steam SS can be adjusted to horizontal/vertical pixels while DRF seems to be calculating with the total amount of pixels. So its not 300% SS will lead to 150% if using 0,5 DRF. 
• You need an CPU/GPU logging system like fpsVR, otherwise you can not do any judgement inhere. 

 

Think the first step would be getting to the point were you can reach 50% of your flyingtime with the adjusted SteamVR SS. 

Then adjust the DRF to the normally SS that can be compared to 140-150% SteamVR SS. 

Shutdown the 45 reprojection if doing the testruns. 

 

Edit: btw do not use the benchmark as being published here as this is nothing more then a video recording, not for CPU. 

Edited May 10, 2020 by Dutch2

[chiliwili69]

I have always run with DRF=1, since I experimented at the beginning and I didn´t like that the system decrease the resolution and number of particles in effects (and possibly other details) while I am playing.

I prefered to run at lower settings and enter in the 40-80fps territory when things get crowded.

Yesterday I have tried DRF to 0.8 (and also 0.5) with 200% and also 300% with my Index. I didn´t improve the experience for me since the limiting factor is CPU.

I prefer "choppiness" over "blurriness", given that never go below 40fps. I tolerate quite well being in the 40-80 region for hot moments.

I don´t know how the code of DRF works, but I suppose will look just at fps, regardless of being GPU constrained or CPU constrained. I really don´t know.

 

I re-read your post with the remagen test:

https://forum.il2sturmovik.com/topic/56485-benchmark-for-cpuram-performance-remagen-4002-to-4005/?do=findComment&comment=872918

 

So if you run now at 300% res:3816x3740 (this is 3 times the pixels you run the remagen test at 100% res:2205x2160) your GPU will be always overload with DRF=1.

So, it a normal scenario (High clouds, High settings, no Motion Reprojection) you for sure are below 90fps, so the system is always applying the 0.8 factor to the resolution.

So it is like running with 240%SS (that is 300*0.8) and no DRF

 

I assumed that the 0.8 is a factor which is applied to the total number of pixels and not to each axis. If not the applied factor would be 300*0.8*0.8= 192%.

This is something I really don´t know. We should ask dev team how this factor is applied.

 

In any case, I think the DRF will only help to GPU constrained systems and this is not the case in most of the PCs and headsets we have.

 

I think  with the Reverb you will not gain too much detail beyond 130% (or even 100% at 2205x2160 which is 9.5 million pixels).

I experimented with SS% to see if I was getting more detail at different SS% (using previous SteamVR version, so % doesn´t match, so look just for pixels) for 5, 9.5 and 15 millions pixels:

https://forum.il2sturmovik.com/topic/53689-hp-reverb-supersampling-100vs188vs300-through-the-lens-pictures/

 

Don´t use your brain to judge if you get visual gain. Brains tend to tell you what you are expecting to watch...  ? 

 

Regarding GPU load with the Reverb for a 1080Ti (and no clouds) I made some testing at 5 and 9.5 million pixels and definitely the GPU was not constrained:

https://forum.il2sturmovik.com/topic/53603-valve-index-vs-hp-reverb-through-the-lens-pictures/?do=findComment&comment=815845

 

So I think that you system performance (with no clouds) is the same if you set SS to 100% (9.5 millions pixels) and set DRF ON or OFF. And probably the same if you increase SS to 130%.

 

Just a couple of additional comments:

- You pictures show no clouds. Clouds impact greatly the GPU load. Try to put High clouds with overcast clouds. You will need to apply DRF to 0.5.

- You don´t mention if you use WMR "Motion Reprojection". My experience was quite good with that in the Reverb.

 

 

[Gomoto]

Thanks for the detailed reply. 

You prefer choppiness over blurriness. I tend to prefer a smooth picture without jagged edges and try to reduce the resulting flimmering. I am sure picture quality is a very subjective matter. We do not have to decide what are the best settings for every person in all situations. I am just very pleased with the experience and would not have tried the combined 0.8/300 settings because they are counter intuitive without reading about it. So I just want to share the idea if somebody has fun experimenting with it.

 

Perhaps I try a systematic approach to compare 0.8 & 130%/200%/300% with the remagen test if I can stop flying these days. But when I just compared the experience, I simply enjoyed the 300% the most. With the through the lens pictures you can not capture the flimmering with pictures in motion. I tried to capture some pictures through the lense in the past myself but it is a little bit cumbersome and awkward. But I appreciate the through the lense pictures of you and others a lot. Such pictures helped me for example when I decided which HMD to get.

Edited May 10, 2020 by Gomoto

[Gomoto]

And a further remark for those with lower resolution headsets than the HP Reverb:

 

Rift S 1280x1440 per eye

Index 1440x1600 per eye

Reverb 2160x2160 per eye

 

That is why I can stand 0.8 PD much more than you:

2160*2160*0.8 > 1440*1600 

 

The 0.8 PD advice is only applicable to high resolution HMDs. 

[E69_Soec]

On 5/8/2020 at 12:34 PM, SCG_Fenris_Wolf said:

P.S. Also, enjoy your now-cut rendering bubble, and getting bounced by close to invisible aircrafts. ?

Could you please elaborate on this? how that 'rendering bubble' works? what parameters affect it?

 

I'm usually getting bounced by invisible aircrafts and the trucks render so so late when strafing with my reverb.

 

Thanks in advance

[SCG_Fenris_Wolf]

4 hours ago, E69_Soec said:

Could you please elaborate on this? how that 'rendering bubble' works? what parameters affect it?

 

I'm usually getting bounced by invisible aircrafts and the trucks render so so late when strafing with my reverb.

 

Thanks in advance

Sure,

the oversampling affects scaling, and leads to loss of information.

 

The higher you supersample, the harder spotting will become. If you exceed 120/130% SS to your native resolution, the information of pixels at range will only be a subpixel on your display. Hence, planes start to become invisible if you exceed that.

 

It also depends on the matrix. RGB matrix has more issues with 3 subpixels. AMOLED displays like Samsung Odyssey seem to do better.

Edited May 11, 2020 by SCG_Fenris_Wolf

[E69_Soec]

2 hours ago, SCG_Fenris_Wolf said:

Sure,

the oversampling affects scaling, and leads to loss of information.

 

The higher you supersample, the harder spotting will become. If you exceed 120/130% SS to your native resolution, the information of pixels at range will only be a subpixel on your display. Hence, planes start to become invisible if you exceed that.

 

It also depends on the matrix. RGB matrix has more issues with 3 subpixels. AMOLED displays like Samsung Odyssey seem to do better.

ok thanks, but does it have any link also with the dynamic resolution factor?

 

Edited May 11, 2020 by E69_Soec