nVidia driver hotfix for SteamVR released

[Bailey]

I'm just chiming in. I didn't see a difference in performance. 

[imacken]

3 hours ago, Bailey said:

I'm just chiming in. I didn't see a difference in performance. 

+1. Me neither, but there again, I wasn't having any issues before this 'fix'.

[Dodge]

On 1/21/2021 at 9:06 AM, speed-of-heat said:

@Dodgesomething i really dont understand (and its not just you) is why people use both in game PD and super sampling to achieve their results ... is there an improved graphical outcome ?  I have never seen one but thats just me I'm weird... and whatever works for you works.

 

but just from a simple math POV you are currently driving a total of 6,381,582 pixels.

 

If you had left your custom resolution multiplier alone (e.g. 100% for DCS, or 150% of native) you would be driving 5m (ish)pixels.

 

By increasing your Steam super sampling to for DCS 126% you would get 6,333,344 pixels pretty much the same (128% would take you to just over 6.4m).

 

I can only see a case for lost cpu cycles and fidelity as a result of multiple super sampling (up and down and up again) operations... does one have different properties (I know the SS math is done differently for PS vs Steam SS, but ultimately that's a granularity issue not a visual outcome)?  

 

 Hi @speed-of-heat

 

Long week so forgive me.

 

Am I understanding your formula correctly....

 

Current settings as posted above:

 

SteamVR

• Native Resolution: 150%
•  Custom SS: 44%

DCS

• PD: 1.7

 

TOTAL SS: 66%

====

 

Your suggestion (am I reading the your reply correct?)

 

SteamVR

• Native Resolution: 150%
•  Custom SS: 126 or 128%%

DCS

• PD: 1.0

 

TOTAL SS@ 192%

 

I've just tried this and although the image seems similar to my settings, I have black edges appear when turning my head around (and the entire movement is somewhat laggy).

 

 

[Added after posting]

 

====

 

Your suggestion (am I reading the your reply correct?)

 

SteamVR

• Native Resolution: 150%
•  Custom SS: 100%

DCS

• PD: 1.0

 

TOTAL SS@ 150%

 

With fpsVR - It felt very 'jittery' on the outside

Without fpsVR - still 'jittery' but not as bad.

Notice the clarity wasn't as good (imo) as my original settings.

Edited January 22, 2021 by Dodge

[speed-of-heat]

yep, thats hould be the same number of pixels or close enough , make sure all the shaders have built though, if you are getting some lag 

 

 try  100% in custom ss and see what that gives you 

[Shrimp]

Nvidia GeForce 461.40 out.

 

Problem solved.

 

There’s still one rare dropped frame every now and then but the stuttering is gone. It is silky smooth and usable even at 30 FPS.

 

The only remaining issue is in Syria map where performance optimization is needed on the part of the developer. So an unrelated matter.

 

I tested SP on all maps with the Bf 109 and P/TF-51 with IA missions (mainly BFM and free flight in big cities) and MP on Storm of War server.

 

I can finally watch the F3 fly-by view in all its glory and identify bogies at the merge!!!

[Dodge]

Just to update this post from my previous posting (@speed-of-heat, I realised I had edited my last reply at around the same time of your last reply above. 

I have since upgraded my system to an AMD 5800x/X570 Motherboard, 64GB 3600MHz RAM, same 3070 GFX - using the latest NVIDIA drivers currently '461.40'), and the system does better than previously (higher FPS and somewhat more consistent). I'll continue to tweak the tweaking over the coming days/weeks and will try to reply here with the outcome that satisfies me

Edited February 2, 2021 by Dodge

[Notso]

On 1/20/2021 at 3:15 PM, airdog said:

I've installed the hotfix drivers and have noticed a huge improvement on my G2.

 

What video card are you and the others who have said this hotfix improves performance?

[imacken]

We’ve moved on from the hot fix driver now. 461.40 is out and see above for positive comments. 

[BMGZ06]

I was getting up to 90 fps in instant mission F18 cold and dark with 100%SS no MSAA no shadows with 16x AF before switching to 461.40. Then that got cut almost in half after. I tried to go back to my old driver and did a clean removal with DDU and reinstalled old driver and now it wont go above 63fps spikes in the same instant mission with same settings. My system also did update to latest win 10 release. It was forced and I could not stop it. 

 

Running HP G2 with 9900KS 5.1ghz 2080Ti FTW 3 oc'd and 32gb gskill royal and samsung 970 EVO

[Bailey]

On 1/21/2021 at 8:25 PM, Taz1004 said:

 

Game doesn't render twice.  It doesn't render once from DCS and another from SteamVR.  It takes the PD and SS values (as resolution) and multiply into final resolution to send to GPU.  And output once (per eye that is).  So both PD and SS does same thing.  Confusion comes from different scale as speed-of-heat mentioned.

 

And not only the different scale but 100% of SteamVR isn't even native resolution for many HMD.  Which adds even more confusion.

 

"Told ya"  

"DCS' Pixel Density, set in the VR tab in its settings, seems to disproportionally affect framerate performance with a worse picture, as compared to SteamVR's supersampling at equivalent and even lower resolutions."

[imacken]

That video gets the numbers all wrong. Like so many others, he misunderstands the relationship between PD numbers and Steam SS ones. 
it is infuriating!

[Taz1004]

2 hours ago, Bailey said:

"Told ya"  

"DCS' Pixel Density, set in the VR tab in its settings, seems to disproportionally affect framerate performance with a worse picture, as compared to SteamVR's supersampling at equivalent and even lower resolutions."

 

As imaken pointed, there are few errors on your numbers.  I'm assuming the SteamVR setting you showed at the end is what you used for benchmarks.  If so, then your SteamVR is not 500% nor 100%  Because your global was set at 140%  Per App setting does not "replace" Global.

 

Also, I'm sure your 100% SteamVR will not be your native resolultion.  The resolution specified at 140% on your video shows 2384x2648 which is 273% of native Index resolution of 1440x1600.  What resolution does it indicate at 100%?

 

Once again, calculate exact resolution you're running for each test.  Not % value.

Edited March 17, 2021 by Taz1004

[Bailey]

16 minutes ago, Taz1004 said:

Per App setting does not "replace" Global.

 

(That is not my video.)
How does it interact if "per app" does not replace "global"?

[Taz1004]

10 minutes ago, Bailey said:

(That is not my video.)
How does it interact if "per app" does not replace "global"?

 

Sorry, didn't realize it wasn't your video.

It is added.  140% Global + 140 Per App totals 180%

And SteamVR % value calculates pixel count.

DCS PD is linear dimension multiplier.

 

And once again, on top of that calculation mess, 100% SteamVR is not 100% native resolution.  It most of the time has multiplier already built in.  And it depends on headset.  So I don't even look at any guide that talks in %.  Always calculate what your final resolution is.

Edited March 17, 2021 by Taz1004

[Bailey]

2 minutes ago, Taz1004 said:

 

Sorry, didn't realize it wasn't your video.

It is added.  140% Global + 140 Per App totals 180%

And SteamVR % value calculates pixel count.

DCS PD is linear dimension multiplier.

 

And once again, on top of that calculation mess, 100% SteamVR is not 100% native resolution.  It most of the time has multiplier already built in.  And it depends on headset.  So I don't even look at any guide that talks in %.

So if you do 210% Global + 520% App, the result would be 530% total?
As for the video, I think that the "point" was that you can get out some frames with some interesting settings and still have similar visual quality.

[Taz1004]

2 minutes ago, Bailey said:

So if you do 210% Global + 520% App, the result would be 530% total?
As for the video, I think that the "point" was that you can get out some frames with some interesting settings and still have similar visual quality.

 

It would be 530% "increase".  So per SteamVR definition, it'd be 630%.  Because in SteamVR 140% means 40% increase in pixel count.  Please just don't talk in %.  SteamVR even gives you resolution number.  Use that instead.

I understood the point but the point itself is pointless unless you figure out the exact resolution because you may be sacrificing resolution to get performance which is nothing new.

[Bailey]

3 minutes ago, Taz1004 said:

 

It would be 530% "increase".  So per SteamVR definition, it'd be 630%.  Because in SteamVR 140% means 40% increase in pixel count.  Please just don't talk in %.  SteamVR even gives you resolution number.  Use that instead.

I understood the point but the point itself is pointless unless you figure out the exact resolution because you may be sacrificing resolution to get performance which is nothing new.

That makes more sense. But I believe that the point stands because it is what you actually "see", rather than the resolution. You brain may not notice that it is looking at a lower resolution, but it sees that it looks just as good as the other setting.

[Taz1004]

4 minutes ago, Bailey said:

That makes more sense. But I believe that the point stands because it is what you actually "see", rather than the resolution. You brain may not notice that it is looking at a lower resolution, but it sees that it looks just as good as the other setting.

 

Isn't that also known as "placebo"?

[Bailey]

3 minutes ago, Taz1004 said:

 

Isn't that also known as "placebo"?

No. More like "what you see is what you get". Not everyone is getting good results.

[speed-of-heat]

Look it’s very simple, in one your cpu is doing more work than the other (driving more pixels)... so it looks smoother in the one doing less work.

[ShaunX]

I'm going to try the PD 0.5 trick this morning, though I'm pretty skeptical

 

I currently have DCS PD at 1.0. In steam VR, I run Global SS at 100%, DCS specific SS at 100% which SteamVR calculates for my rig to be 3164 x 3096 (catering for barrel distortion)

 

So, if PD is a linear dimension and as a multiplier I should use it as 0.5 * 0.5 * SteamVR SS, then increasing SteamVR DCS SS to 400% should give a similar resulting resolution ? (Am I right up to now ?)

 

If I increase SteamVR DCS SS to 400% it calculates that to be 6332 * 6196 for my rig, which is pretty much exactly double to the 100% figures above ( err, shouldn't it be 4 times ? I'm confused now.)

 

Anyway, with my original settings, if I  jump in the Hornet on a hot start on a very busy supercarrier deck (which always stresses my rig so that I'm barely maintaining 45 FPS) , monitor the FPSVR figures, then quit the mission and save the track. Then replay the the track with the new settings and monitor FPSVR again, does that sound like a reasonable comparison ?

[ShaunX]

"which is pretty much exactly double to the 100% figures above ( err, shouldn't it be 4 times ? I'm confused now.) " - is it because it's a per-eye dimension and therefore actually 4 times over both eyes ?

 

[imacken]

This is crazy, and proves the point I was making!

it IS 4 times the number of pixels. 3000x3000=9000000 and 6000x6000=36000000, i.e. x4! (Approx numbers used)

[ShaunX]

8 minutes ago, imacken said:

This is crazy, and proves the point I was making!

it IS 4 times the number of pixels. 3000x3000=9000000 and 6000x6000=36000000, i.e. x4! (Approx numbers used)

 

I agree with you and I'm most skeptical about the trick. I will however give it a try this morning when I'm able with those settings.

 

I think the claim in the video is that the Steam VR advanced super sampling is somehow more advanced/efficient that DCS's rendering, such that, if you maintain the same absolute resolution and let Steam VR do the work instead of DCS, you'll get a better experience. I'll report back later

[imacken]

I was clarifying what you were asking, i.e. it's 4 times, not double.

IMO, all this stuff about reducing this, increasing that is clouded in mis-information and falsities.  The sensible approach has always been to leave PD alone and just use SteamVR SS to play around with.