Scaling interactions

[Pikey]

Hi, I am trying to understand the methods to tune VR scaling (everything works already) using OXR and HP Reverb 1. I need some help with some basics, specifically the interaction between:

OpenXR Tools for Windows Mixed Reality | Customised render scale (as per screenshot)

OpenXR Toolkit | Upscaling/Sharpening NIS | Size (as per screenshot)

 

QUESTIONS

1. Is the render scale on OpenXR Tools for WMR applied before the OpenXR NIS upscaling?

2. Which direction of render scale does what effect? I read on Reddit that 100% is auto adjusted based on GPU RAM but is this 100% of the GPU memory adjusted figure? i.e. is 100% the same resolution for everyone?

3. In VR headset the NIS upscaling "Size" setting requires the user to increase the number to reduce the screen resolution to improve the performance. Is this AFTER the Customised render scale or does it Overwrite that scale or something else?

4. I see NIS, FSR, CAS. CAS has no resolution size. What helps me pick one, what should be better for DCS? I can't really tell easily in game as I have to restart between and it seems no different.

5. Is the DCS VR setting, is Pixel Density applied after or before any of these and does any figure other than '1.0' help?

6. How will any of this be affected by Nvidia DLSS 2? Applied before, after, in the middle somewhere or not at all?

7. When either tool says to Restart VR to take effect, does this mean you need to close WM Portal mirror screen?

I'm sorry for these questions, good answers will help choose which slider to set before another and save hours of testing

 

Edited January 27, 2023 by Pikey
nullnullnull

[edmuss]

30 minutes ago, Pikey said:

1. Is the render scale on OpenXR Tools for WMR applied before the OpenXR NIS upscaling?

Yes.

30 minutes ago, Pikey said:

2. Which direction of render scale does what effect? I read on Reddit that 100% is auto adjusted based on GPU RAM but is this 100% of the GPU memory adjusted figure? i.e. is 100% the same resolution for everyone?

Default (as in override check box unchecked) will set a recommended render resolution based upon your VRAM.  If you override the resolution via either the desktop tools app or ingame with the toolkit then it will ignore the default.  You can see what actual resolution is being rendered on the second tab of the desktop tools.  100% depends on your headset, for my G2 it's 3160x3050 or something along those lines.

30 minutes ago, Pikey said:

3. In VR headset the NIS upscaling "Size" setting requires the user to increase the number to reduce the screen resolution to improve the performance. Is this AFTER the Customised render scale or does it Overwrite that scale or something else?

It takes the target resolution (what you have set by override or from the default as above) and reduces that by the value set in the upscaler and then upscales back to the target resolution.  The actual rendered resolution is shown in the toolkit when you adjust the upscaling value.  Note that values above and below 100% reduces the render resolution, pay attention to the actual rendered resolution as displayed by toolkit.  If you want to improve image quality, increase the target resolution above 100% with overrides and then apply the upscaling; it may take a couple of restarts to get the settings all nailed in to what you want.

30 minutes ago, Pikey said:

4. I see NIS, FSR, CAS. CAS has no resolution size. What helps me pick one, what should be better for DCS? I can't really tell easily in game as I have to restart between and it seems no different.

Currently the CAS upscaling component isn't implemented but I believe that it is planned or at least has been worked on.  Personal preference really, they handle colours and sharpening in different ways so it's up to you.  I must say that the CAS sharpening is superior to either NIS or FSR in my opinion.

30 minutes ago, Pikey said:

5. Is the DCS VR setting, is Pixel Density applied after or before any of these and does any figure other than '1.0' help?

I think it's applied before, I'd recomment leaving it at 1.0 as you have far more granular control on resolution overrides in the toolkit.

30 minutes ago, Pikey said:

6. How will any of this be affected by Nvidia DLSS 2? Applied before, after, in the middle somewhere or not at all?

Unsure, I would imagine that it would be before as it's a DCS integration rather than a seperate tool applied afterwards.

30 minutes ago, Pikey said:

7. When either tool says to Restart VR to take effect, does this mean you need to close WM Portal mirror screen?

Just need to restart the VR app, in this case DCS

 

Hope that helps

Edited January 27, 2023 by edmuss

[Glide]

1 hour ago, Pikey said:

1. Is the render scale on OpenXR Tools for WMR applied before the OpenXR NIS upscaling?

We really need @mbucchia to help with these answers. 

1 hour ago, Pikey said:

I see NIS, FSR, CAS. CAS has no resolution size. What helps me pick one, what should be better for DCS? I can't really tell easily in game as I have to restart between and it seems no different.

NIS only works on Nvidia cards, FSR is open source and works on any hardware, and I can't speak to CAS.

Edited January 27, 2023 by Glide

[Pikey]

Thank you for these answers! Yes, very helpful.

So typically I head to OpenXR tools for WMR and reduce the number there first to save the pixels being pushed which is the largest FPS gain, then make up the image quality in game with the upscaling toolkit. First one starts making the image jaggy, second blurs a bit? Is this roughly it for the layperson? Please correct me if I misunderstood! Cheers!

[edmuss]

4 minutes ago, Glide said:

We really need @mbucchia to help with these answers. 

NIS uses the custom Tensor cores on Nvidia cards, FSR is open source and works on any hardware, and I can't speak to CAS.

It's fairly straight forward with the first question, you adjust the override and it alters the upscaled render resolution.

NIS doesn't use tensor cores, that's DLSS which is limited to RTX only cards; NIS is driver based and is purely software only and will run on a GTX card as well.

[edmuss]

1 minute ago, Pikey said:

Thank you for these answers! Yes, very helpful.

So typically I head to OpenXR tools for WMR and reduce the number there first to save the pixels being pushed which is the largest FPS gain, then make up the image quality in game with the upscaling toolkit. First one starts making the image jaggy, second blurs a bit? Is this roughly it for the layperson? Please correct me if I misunderstood! Cheers!

My experience with upscaling in DCS has shown that if you start with a lower resolution then the upscaling will add a LOT of shimmer and aliasing.  You will get a healthy performance uplift but at the expense of a lot of image quality, however the uplift from upscaling is generally more than the uplift from simply reducing the resolution.

I run 3300 (around 105% on the G2) wide target resolution and use NIS 75% which reduces the render down to around 2600 which is than upscaled back up to the 3300.  I've found that this gives a very good balance of clarity, performance and lack of shimmer when coupled with MSAA2x.

[Burt]

@edmuss" use NIS 75% " ? Are sliding it to the right or left ?  Sorry for the lame question but I need to get this out and understand it. 

Edited January 27, 2023 by Burt

[edmuss]

It doesn't matter which way you slide it, the further you get from 100% (in either direction) the more it will reduce the render resolution.

The further you upscale, the more overhead it costs the GPU (measured by sclGPU on the toolkit developer overlay) and the more it will introduce aliasing and artifacts. For me to upscale to 3300 (overridden target resolution) using NIS 75% (which gives a render resolution of arrived 2500 I think?) costs me about 0.6ms GPU frametime.

[Burt]

Got it,  thx !