Nvidia Sharpening and VR

[Chapa]

@Voight: Thanks you for your feedback. Effectively, we should say which plane we use to compare our settings because there are specific optimisation for VR to reduce bad effect like shimmering. (For example, adding around all textures a layer of full transparent pixels to reduce Aliasing). In my case, I test with Huey, P51 and F18. I bought yesterday the MiG21 and also noticed that shimmering is more noticeable in this plane and I pretty sure that I will reconsider my settings if I focus on this plane.

 

Concerning the AA settings in Nvidia Inspector, from my experience, they are not taken into account in DCS. One simple test is to select the more costly AA in Inspector, run DCS in no VR mode and see that the FPS is exactly the same than with just MSAAx2 or no MSAA. But if you have really the feeling that some AA settings works, please share with us your settings because it is a debate on other thread with nobody being able to prove it works.

 

At the time of DX9, the idea to improve some games was in a nutshell to:

 

1) Add negative LOD bias to increase sharpness but also shimmering unfortunately

2) Add transparency Sparse Grid SSAA to remove shimmering

3) Enhence/Override the application AA settings by adding MSSA from the Inspector with the same multiplicator than the Sparse Grid AA to remove the blur induced by the whole process

 

As you can see, surprisingly, the steps 1),2) is what we did. Keeping the LOD Bias Driver to On allow the driver to apply negative LOD bias. But because applying AA settings from the inspector seems to have no impact we can’t test the full approach .

 

 

P.S the AA settings that has no impact I am speaking about would be: AA mode. I tried to pair it with enhence and override application and try to apply the AA compatibility DX1x flags (all the possible combinations) with no success.

 

P.S.2 However FXAA works and have impact on the monitor. But in the HMD... I am not convinced FXAA works

Edited November 29, 2019 by Chapa

[Wicked.-]

I don’t think I was seeing a placebo effect tonight. I fly the 16 and aliasing in my index is really bad and has been since release. Two things make me think the AA setting you suggested are working. 1. After applying the settings in your post, I had to turn down my SteamVR custom resolution which I explained above. What I didn’t say was initially I set Custom resolution to 204% and all other super sampling to 1 or 100%. At 204% the frame rates were too low. Normally I can run that high all day with close to 45 FPS solid. I had to take it down to 188% to get the frame rates back up. 2. The aliasing was considerably reduced without using MSAA or super sampling of any kind other than Custom Resolution at 188%.

[Chapa]

Yes Wicked the transparency Sparse Grid Supersampling antialiasing works. It even cost at least 30 FPS.

 

I am speaking about other AA settings in Inspector: AA Mode that should override the MSAA setting in DCS. In fact we are lucky that the transparency setting works

 

For sure, it is not placebo effect you are seeing

[Wicked.-]

Well thank you for the settings! Now I need to figure out the watery look issue. Unfortunately the MFD’s still look like ass unless I zoom in. It’s worse now that we have red colored contacts with Link16. With white contacts I could lean in but with the red contacts I have to zoom to see their vectors.

[Chapa]

I don’t know the Index and if it has some similar feature than AWS for the Rift (reprojection maybe ?). The fact you are saying that the watery effect occurs on ground let me think it is an artifact caused by something triggering when the charge on the GPU increases.

 

My advice to track it would be to put all DCS settings to low first. Keep only Texture: High, your PD setting (decrease it a little bit could be good for testing) and the Inspector settings at they are now. If it disappear, it means that it could be:

 

1) some DCS settings

2) because you have free resources, the AWS like system don’t trig and no more artifact. To be sure, then increase a lot the PD setting. It will force its activation and see the problem occurs again. If it occurs, it means that the watery effect is caused by the AWS/reprojection process

[Supmua]

You're most welcome. That being said, the settings were recommended to me as well lol. Thanks for testing much further than I did, I tend to get a benefit and then fly for the next week instead going further lol.

 

 

@ Supmua, what are your SteamVR settings with the above NvidiaProfileInspector changes?

 

My SteamVR setting is custom res at 100%. I use PD 1.4 and MSAA 2x in DCS to counter shimmering with LOD sharpening effect. I only use -0.375 for LODs currently, not sure if there's any tangible difference above -0.5 though but below -1.0 it definitely induces grain and shimmering. This setting is ironically more for other games than DCS, lol.

Edited November 29, 2019 by Supmua

[Bob_Bushman]

Tested this out with the S tonight, and honestly find it vastly inferior to what I had from before.

 

These 'recommended' settings was as if someone smeared grease on my lenses with the bonus of heavy aliasing and shimmering.

 

Sure, ASW triggered a little less than my usual settings but only barely and even the Dashboard\desktop view in the rift took a nose dive in clarity, reverting back to defaults clear that back up and running at pd 1.3 and SSAA 1.5x is much much better in DCS.

 

Anyhoo that's my two cents opinion.

[Rosebud47]

@Chapa

 

Yeah, me too would say, that the graphical fidelity received depends also on the module. I´m gladly looking forward for the reworkd MiG21 cockpit, what should be more optimized for VR than the current one.

The new F-16 looks a bit more washed out in VR than other cockpits, but the F-16 I would consider as a work in progress also in terms VR graphical optimization.

I think the Tomcat could be considered as a standard in it´s graphical optimization and performancewise.

 

Beside this, the headset in use also makes some differences, like the P5K+ I´m using tend to shimmer more than others I think.

 

FXAA should work in VR, it is cheap in performance consuming, but makes the whole image more blurry, what would counteract any approach to sharpen the image. I wouldn´t use FXAA.

 

Now you´ve made me a bit unsettled regarding the use of AA with nvidia inspector. But I´m pretty sure, that the count of steps at thin edge line was significantly increased due to switchting from 2xAA in nvidia inspector to 16xAA. Anyway the steps were still visible and shimmering and not flatted out without shimmering. First time I tried some settings adviced in this thread the image was really good in the MiG cockpit. I´ll try to find back these settings.

[Wicked.-]

Voight, I went back to these settings but replaced sparse grid super sampling at 8x with AA MODE REPLAY MODE ALL., Negative LOD BIAS -0.625, MSAA and AF off in game. From here I’m adjusting my steamVR/Video/Custom Resolution slider. Currently it’s at 188%.

 

 

 

- Antialiasing - Transparency Supersampling: 8x Sparse Grid Supersampling

- Texture filtring - Driver Controlled LOD Bias: On

- Texture filtring - LOD Bias (DX): +0.000 / Automatic

- Texture filtring - LOD Bias (OGL): +0.000

- Texture filtring - Negative LOD Bias: Allow

 

I have also set the following anisotropic filtering settings and disable the AF in DCS (nvidia settings are taken into account):

 

- Anisotropic filtering mode: User-defined/Off

- Anisotropic filtering setting: 16x

 

The 16x AF from Nvidia inspector is better than the 16x provided by DCS.

 

 

And finally MSAAx2 in DCS settings.

[Chapa]

Tested this out with the S tonight, and honestly find it vastly inferior to what I had from before.

 

These 'recommended' settings was as if someone smeared grease on my lenses with the bonus of heavy aliasing and shimmering.

 

Sure, ASW triggered a little less than my usual settings but only barely and even the Dashboard\desktop view in the rift took a nose dive in clarity, reverting back to defaults clear that back up and running at pd 1.3 and SSAA 1.5x is much much better in DCS.

 

Anyhoo that's my two cents opinion.

 

I will try your settings because I always read that DCS SSAA is for monitor only. I have a rig similar to yours and it is clear that a PD of 1.3 tend to activate AWS most of the time.

 

The settings proposed in this thread is more fore people wanting to play at 80Hz but with good clarity at the cost of some aliasing and shimmering. With a Rift S, a 1080ti and AWS on it is possible to have near perfect view. Until you chase a warbird and you see 4 times the plane because the gosthing effect induced by AWS and you have hard time to know if it is an ally or not (not easy for me at least).

 

A PD of 1.2 is about the same number of pixels than a 4k screen, 1.4 is more than 11 million pixels and sometimes I read people playing with PD at 1.7. It can works with AWS but each time I try, seduced by the nice picture, I am hit by the hard reality that my 1080ti is tired or too lazy...

 

but I will try your settings. Thanks you for the alternative and feedback

[Chapa]

@Wicked: It appears that we are all converging toward a LOD Bias near -0.5

 

It the old days of DX 9, there was a rule saying that :

 

LOD -0.5 for Sparse Grid x2

LOD -1.0 for Sparse Grid x4

LOD -1.5 for Sparse Grid x8

 

this criteria was driven by a mathemical formula with a log, so just to say it was something more complicated than a rule of thumb formula.

 

I am playing from yesterday with the Sparse Grid x2 settings but to be honest, I don’t really know if sparse grid really improve the result instead MODE REPLAY ALL... However, LOD bias of -0.5 with minimum -0.75 seems to be the sweet spots to not induce tons of shimmering

[dburne]

There is an old Confucius saying I am pretty sure that goes something like this:

 

" He who tweaks lot, flies less".

 

:pilotfly:

 

Excellent words of wisdom right there.

[Wicked.-]

Tested this out with the S tonight, and honestly find it vastly inferior to what I had from before.

 

These 'recommended' settings was as if someone smeared grease on my lenses with the bonus of heavy aliasing and shimmering.

 

Sure, ASW triggered a little less than my usual settings but only barely and even the Dashboard\desktop view in the rift took a nose dive in clarity, reverting back to defaults clear that back up and running at pd 1.3 and SSAA 1.5x is much much better in DCS.

 

Anyhoo that's my two cents opinion.

 

 

 

 

While I appreciate the feedback, it doesn't tell me anything other than you like SSAA @ 1.5x and PD at 1.3. When you say "These 'recommended' settings", to which are you referring? You're the first S user I've heard of having a bad result with the settings I'm using. Below are my settings for the Valve Index. I have had my (VR using) squad mates apply these settings regardless of HMD. So far that's 3 HP Reverb's, 2 Index's and 1 Oculus S. In each case MSAA, SSAA and Anisotropic Filtering were turned off inside DCS but nothing else on the Settings/System page was altered. For the Oculus S we lowered the PD from 1.5 to 1.2 as an initial starting point to gauge FPS. The S is on an older system so we knew it would lose FPS. Immediately after making these changes for the S we flew a custom mission online that brought even my system to it's knees but the S remained playable throughout. He reports dramatic increases in visual clarity and performance.

 

The Reverb guys had to make changes to some file having to do with WMR and had to run SteamVR Stable along with WMR Beta to achieve the results we on the Oculus and Index are seeing. They report marked improvement in image quality and performance as well.

 

This thread does not make it clear which settings you should start with regardless of HMD. Here they are. Once you have these set, you should only need to turn off MSAA, SSAA (if you're using it) and Anisotropic Filtering within DCS. Then make whatever adjustments you're accustomed to to get your frame rates up. IE, for me I lowered my SteamVR Custom Resolution to 188% from 204%. So if you currently use PD to increase clarity, adjust it down until your frames come up.

 

 

 

Lastly, the smearing you refer to sounds like what happens when you select a positive LOD number instead of a negative number.

 

 

 

 

Edited November 30, 2019 by Wicked.-

[Wicked.-]

@Wicked: It appears that we are all converging toward a LOD Bias near -0.5

 

It the old days of DX 9, there was a rule saying that :

 

LOD -0.5 for Sparse Grid x2

LOD -1.0 for Sparse Grid x4

LOD -1.5 for Sparse Grid x8

 

this criteria was driven by a mathemical formula with a log, so just to say it was something more complicated than a rule of thumb formula.

 

I am playing from yesterday with the Sparse Grid x2 settings but to be honest, I don’t really know if sparse grid really improve the result instead MODE REPLAY ALL... However, LOD bias of -0.5 with minimum -0.75 seems to be the sweet spots to not induce tons of shimmering

 

 

I did not like the results from any of the Sparse Grid options. The MODE REPLAY ALL and -.6250 with SteamVR Custom Resolution set to 188% is my sweet spot and what I'm flying. I tried -.500 as well but it was too blurry. In the Viper, at -.6250 I can see the cage knob on the backup ADI and actually read it without leaning in. Before these settings I didn't realize it was modeled as an actual 3D knob lol.

[Chapa]

I agree with you Wicked, Sparse Grid is not the way to go... Not worth the 0.1 PD it cost at the end for less...

 

For info, SSAA has not effect (or I am not able to see it). I also think that he set positive LOD Bias. The importance is that he is happy with his stuff and settings.

 

Concerning Confusius, he will be happy to know that I am a patriot and the important thing is to have the settings ready for the release of the JF17.

[dburne]

I agree with you Wicked, Sparse Grid is not the way to go... Not worth the 0.1 PD it cost at the end for less...

 

For info, SSAA has not effect (or I am not able to see it). I also think that he set positive LOD Bias. The importance is that he is happy with his stuff and settings.

 

Concerning Confusius, he will be happy to know that I am a patriot and the important thing is to have the settings ready for the release of the JF17.

 

SSAA in DCS has no affect in VR HMD.

[Wicked.-]

For Bob_Bushman I would hate for him to miss out from an incorrect setting but I agree, if you're happy, it's all good.

 

 

In all my testing in the last two years SSAA has never made any difference in VR for me. Whether CV1 or Index. That's not to say it isn't working now with the new Nvidia drivers. It's obvious Nvidia is enabling adjustments to affect VR in the last few updates.

[VirusAM]

While I appreciate the feedback, it doesn't tell me anything other than you like SSAA @ 1.5x and PD at 1.3. When you say "These 'recommended' settings", to which are you referring? You're the first S user I've heard of having a bad result with the settings I'm using. Below are my settings for the Valve Index. I have had my (VR using) squad mates apply these settings regardless of HMD. So far that's 3 HP Reverb's, 2 Index's and 1 Oculus S. In each case MSAA, SSAA and Anisotropic Filtering were turned off inside DCS but nothing else on the Settings/System page was altered. For the Oculus S we lowered the PD from 1.5 to 1.2 as an initial starting point to gauge FPS. The S is on an older system so we knew it would lose FPS. Immediately after making these changes for the S we flew a custom mission online that brought even my system to it's knees but the S remained playable throughout. He reports dramatic increases in visual clarity and performance.

 

The Reverb guys had to make changes to some file having to do with WMR and had to run SteamVR Stable along with WMR Beta to achieve the results we on the Oculus and Index are seeing. They report marked improvement in image quality and performance as well.

 

This thread does not make it clear which settings you should start with regardless of HMD. Here they are. Once you have these set, you should only need to turn off MSAA, SSAA (if you're using it) and Anisotropic Filtering within DCS. Then make whatever adjustments you're accustomed to to get your frame rates up. IE, for me I lowered my SteamVR Custom Resolution to 188% from 204%. So if you currently use PD to increase clarity, adjust it down until your frames come up.

 

 

 

Lastly, the smearing you refer to sounds like what happens when you select a positive LOD number instead of a negative number.

 

 

 

 

 

 

 

I am also having bad results with rift s and any negative lod bias (but I don’t know what it means...and didn’t know the existence of such a setting before reading this thread) however i hope that experimenting more i will be able to find a sweet spot.

Thank you for your settings anyway

[Rosebud47]

Just realized, that I had set "Texture filtring Negative LOD Bias" to "clamp" instead "allow", what seems to be resulted in lots of shimmering again. With set to "allow" as you recommend shimmering is again so much reduced.

Now I need to fine tune sharpness ...

 

As far as I understood, SSAA in DCS is practically the same as PD, whereas SSAA is for the monitor output and PD for the VR headset output.

 

@dburne

latest research of transcriptions revealed, that these words of Confucius are misinterpreted. The correct translation says: "He, the wise man who goes along the green hills down to the water, has always tuned his fishing pole to catch the fish for the emperors men."

[dburne]

Just realized, that I had set "Texture filtring Negative LOD Bias" to "clamp" instead "allow", what seems to be resulted in lots of shimmering again. With set to "allow" as you recommend shimmering is again so much reduced.

Now I need to fine tune sharpness ...

 

As far as I understood, SSAA in DCS is practically the same as PD, whereas SSAA is for the monitor output and PD for the VR headset output.

 

@dburne

latest research of transcriptions revealed, that these words of Confucius are misinterpreted. The correct translation says: "He, the wise man who goes along the green hills down to the water, has always tuned his fishing pole to catch the fish for the emperors men."

 

LOL!

[Bob_Bushman]

I did some more A/B testing and indeed SSAA doesn't work at all.

So I'll admit to that one, ;)

 

Also been testing a little more but there it's as if the anisotropic setting doesn't kick in from the profile inspector, past 2 am here do getting to tired for vr now so will test more later.

 

So far the shimmering is far to bad for me with a lod bias of less than -0.375 and it's still a far cry in fidelity as what I had with supersampling of 1.3, I'm testing this with a PD setting of 1x

 

And regardless of of running with 1x PD, I drop to 40 when looking over a city scape or any other some what busy place.

So I don't really see much point in taking a drop in pq when I won't get any big performance increase when it matters.

 

And run PD at 1.3.

[fitness88]

For Bob_Bushman I would hate for him to miss out from an incorrect setting but I agree, if you're happy, it's all good.

 

 

In all my testing in the last two years SSAA has never made any difference in VR for me. Whether CV1 or Index. That's not to say it isn't working now with the new Nvidia drivers. It's obvious Nvidia is enabling adjustments to affect VR in the last few updates.

 

 

If Nvidia is enabling adjustments to affect VR in the last few updates why wouldn't they allow this to be done through the Nvidia control panel?

They keep telling me that Oculus is not currently designed to be allowed to be adjusted through the Nvidia CP, only 3rd party software may.

[Wicked.-]

If Nvidia is enabling adjustments to affect VR in the last few updates why wouldn't they allow this to be done through the Nvidia control panel?

They keep telling me that Oculus is not currently designed to be allowed to be adjusted through the Nvidia CP, only 3rd party software may.

 

I have no idea why they wouldn’t put it in the CP. I only know that access to negative LOD values can only be achieved with Nvidia Profile Inspector. AA MODE REPLAY MODE ALL can only be accessed there as well.

Lastly, Oculus can say what they want but making these changes in the Nvidia Profile Inspector directly effects the Oculus S.

[Wicked.-]

I did some more A/B testing and indeed SSAA doesn't work at all.

So I'll admit to that one, ;)

 

Also been testing a little more but there it's as if the anisotropic setting doesn't kick in from the profile inspector, past 2 am here do getting to tired for vr now so will test more later.

 

So far the shimmering is far to bad for me with a lod bias of less than -0.375 and it's still a far cry in fidelity as what I had with supersampling of 1.3, I'm testing this with a PD setting of 1x

 

And regardless of of running with 1x PD, I drop to 40 when looking over a city scape or any other some what busy place.

So I don't really see much point in taking a drop in pq when I won't get any big performance increase when it matters.

 

And run PD at 1.3.

 

That’s too bad it’s not working for you. It made a hell of a difference for me in the Viper which makes me think my Index is a CV1 sometimes. It did induce some shimmering but I can now read almost every label in the pit without having to shove my face against it. I have also been able to reduce my Custom Resolution setting to 178% from 204% without loss of clarity in the pit at all. This in turn has bumped my frame rates to a constant 45 FPS or above.

As other S users in my squad start experimenting I’ll post here if they find any other settings that help.

 

Cheers!

[Bob_Bushman]

It definitely works, no doubt of that.

 

And it's been a while since I did my comparing with custom res vs oculus pd, but from my guesstimate is 208% in Steam is approximately 1.5x in PD and 178% is in the neighborhood of 1.3.

 

Now that can't really compare with the Index all that well, I have never even seen one since Valve still refuses to sell hardware in the EEA.

 

And well my opinion and disappointment of the 5k+ is hardly any secret.

 

I also suspect desired PD is highly individual in the S since we can't adjust lenses in it, it varies between each anyone how each eye hits the lenses.

 

So far I have all only done my testing in the F18, I'll give the viper ago we well but so far I'm not having to much of an issue with performance and pd at 1.3.