As you’ve experienced, the higher the OXR render scale, the better the image quality but the lower the performance is in MSFS DLSS. It’s really up to you where on this scale you are happy with clarity vs performance. For me, with a 4080, an OXR of 150% and DLSS quality gives me a clarity level I am happy with and performance in the 35-50 FPS band, which I am very happy with.
FYI, OXR 100% render scale is 3152 for the X axis. If you want to calculate what OXR render scale equivalent your override resolution is producing, just divide your X axis setting by 3152, square it then multiply by 100. eg. 4658 X axis equates to (4658 / 3152) ^ 2 * 100 = 218% OXR render resolution equivalent.
If you want to work out what your DLSS down sample resolution is, just mutiply your X axis resolution by 0.66 for DLSS Quality, 0.58 for Balanced, 0.5 for Performance and 0.33 for Ultra Performance. eg. with your 4568 X axis and DLSS quality, your down sample X resolution is 3074, which is pretty close to 100% OXR and TAA 100% of 3152.
I suspect you are confusing OXR (OpenXR DevTool) render scale (which I have set 1at 100% and I do not touch it) and OXRTK (OpenXR ToolKit) Resolution Override. I think the original value of OXRTK Resolution Override was around 4100 before I modified it the first time, but I’m not sure. This is more than OXR 100% resolution, if I understand correctly, this is because of this figure is taken before the barrel distortion correction (so not the final image), so it needs to be higher resolution to retain definition in the center of the image.
Normally there are optimal multipliers that give the least artifacts when upsampling or downsampling an image. Because I don’t know the original resolution in OXRTK and I don’t want to reset all settings to find out, I’m asking what it is. Then, this needs to be multiplied by something: ideally 200% but that’s too much to handle, even for 4090. So my second question is what would be the best multipliers to try for least artifacts…
BTW these 2 settings are exactly the same thing. OpenXR Toolkit override resolution and the OpenXR Tools for WMR Custom render scale do exactly the same.
The 2 key differences:
OpenXR Toolkit settings are saved per game, OpenXR Tools for WMR settings are system-wide
Because OpenXR Toolkit is independent of the headset, it doesn’t know what “100%” means, therefore you adjust the resolution itself and there is no percentage.
The first time you enable Override Resolution, it will just default to the current resolution… so the one from OpenXR Tools for WMR for example.
If you want to make up for the lack of percentage, what you can do is use OpenXR Tools for WMR Custom render scale slider to a given percentage, then on the middle tab of that tool, look at what values you see for recommendedImageRect (I think it’s called that?). You can then input that resolution in OpenXR Toolkit Override Resolution (and it will completely override whatever OpenXR Tools for WMR would set).
So why Override resolution is an option in OpenXR Toolkit?? If it does the same thing?
Well not all devices have nice sliders like OpenXR Tools for WMR. Varjo doesn’t to my knowledge. PimaxXR doesn’t either. So this option of OpenXR Toolkit closes that gap. Plus it also gives you per-app control (a la SteamVR) which is always nice.
There is no confusion. OXR at 100% for the 4090 (and 4080) with a Reverb G2 is 3152 X axis. As soon as you start using OXRTK resolution override, you are telling WMR to ignore this setting and use the resolution you have manually set instead.
What I am telling you is that if you were to turn off resolution override and set OXR to 218% (which is not exactly possible as it goes up in 5% increments, but you can round off to the nearest 5% ie. 220%), you will get approximately the same as what you were getting by manually setting X axis to 4658. ie. If you set 220% render resolution, MSFS will initially render at, using the inverse of the formula I gave, SQRT (220 / 100) * 3152 = 4675 on the X axis. You can verify this in MSFS DevMode FPS view or, as @mbucchia suggests here:
That is a very good question, the answer to which I do not know
The reason for the mention of the RTX 4000 is more about their VRAM amounts being sufficient for WMR to default to 100% OXR with the Reverb G2. The 3080 10GB I had before my 4080 defaulted to approx 80% because of VRAM amount I believe.
I’ve been learning how to set up a Pico 4 headset using Virtual desktop, which uses Steam VR, plus I have the OpenXR Toolkit running for things like colour adjustments. I have never used Steam VR much, so am bumbling around. I noticed that with override resolution on in the Toolkit I am limited to 3,200 or thereabouts in the settings and can go lower, but not higher. I am assuming that limitation is caused by Steam VR and I have to use their app setting for MSFS to adjust the resolution beyond 100%, not the Toolkit? Which is a pity as the Toolkit is easier to get at.
Also Steam VR has two potentially interacting resolution settings. Global and app specific, so where to set what is not immediately obvious.
After extensive testing I found out that there is no visible difference between even and uneven numbers.
With my 4090 I am using 5000x… resolution override, OXRTK CAS at 35 and MSFS FidelityFX at 140, all with DLSS Performance. No MR with frame rate cap 45 FPS in the Nvidia control panel.
In most situations my 4090 runs at 80 percent usage, so staying quiet and cool.
Blimey, that’s 251.6% according to ResetXPDR’s formula. With my G2 and a 4090 I’m using 4000x or 160% OXR and either DLSS Quality or Balanced and things look incredibly sharp in the G2’s sweet spot, 30fps with MR is stable.
With DLSS Performance this is a render resolution of 2500x… before the upsampling. This is nothing for a 4090.
To me 4000x… vs 5000x… is a huge difference, try it.
All I said is with Ultra graphics settings, except for TLOD at 150 (and 13900KF with 5.8 GHz All-Core), so the CPU can also go 45 FPS at 95% of the time.
It really depends on the individual chip. I have done my overclocking purely for MSFS, not for Cinebench scores.
In summary:
HT off, 5.8 GHz p-cores, 4.3GHz e-cores, fixed CPU voltage in the range of 1.42 to 1.46V (not an issue if you keep the current low), with Asus LLC6. I have set PL2 at 250W for 28secs, and PL1 at 200W. Without the watt limits you would see 300/350W in benchmarks like Cinebench, resulting in high current and high temperatures. But in MSFS you will never be near that limit, so it is just a protection for CPU-intense applications.
In benchmarks the watt limit will reduce the CPU clock to 5.4/5.5GHz, which is fine for me.
But let us stay on topic please and not discuss this any further.
This can be caused by the OpenXR runtime advertising this 3200 as the max resolution, and therefore OpenXR Toolkit refusing to let you go higher (as in "oh if you tell me this is max, I dont want the user to cause a crash or something). Varjo has this exact bug, so I had to add an exception “if Varjo, ignore the cap”.
That said, IIRC you are coming from WMR? And there is a nasty bug in SteamVR that IF any one of your installed drivers has a resolution cap, it will be enforced for all devices… try this: launch the WMR for SteamVR settings UI (if you don’t have your headset plugged in, you can go find a file called SettingsUX.exe in your Steam folder and run that). The look for an option called something “Allow resolution higher than recommend” and turn that on. Restart your SteamVR, and see if that worked.
Ah, thanks for that clarification. Makes sense now.
So, maybe I supersampled a bit too much. No need to overstress GPU either. OXR 200% = 4458x in OXRTK then. And OXR 150% you’re using is 3860x, unless I did the math wrong… Maybe I can do DLSS Balanced with that resolution…
Yes, that’s already justifies its existence. Also, I’d add another advantage to that: I can change it right in VR (with a quick off-on VR trigger that can be mapped to a voice command), no need to take it off and go to desktop - which for me means extracting myself from my motion rig, walking to PC and then inserting myself back.
Wow, that’s a lot! I wonder if that much supersampling can yield more sharpness than a lower settings, considering G2 resolution.
That sounds more reasonable, I’m going to try that.
Thanks for that. I popped my G2 headset on and found the setting in SteamVR. Will give it a go later today (my wife has just got a new iPad, which I have to set up and transfer her data from her old iPad, which should be fun…not. I hate Apple products with a vengeance )
Just to add, I have CAS sharpening off in MSFS menu and set to 100% in the Toolkit. Seems extreme but works for me, as I like to fly low and slow checking out the scenery, mostly in biplanes or choppers. The detail I get is just amazing. Screenshots giving an idea of what I get. Pompeii and Melbourne in the C172 I was using for comparing the G2 to the Pico 4. I used the fine lettering on the C172 compass as a test for clarity. In the Pico it was unreadable sitting back, but tack sharp in the G2.
I’m actually quite surprised how much more clarity I could get out of already hi-res G2 by supersampling. I initially rejected DLSS because of the blurry digit movement and blurry displays, but now with OXRTK flexible Resolution Overrise and CAS sharpening it allows us to supersample, which was quite unthinkable a little while ago…
Rather than faff about with SteamVR settings, I was then able to simply set Steam VR default resolutions and use the Toolkit resolution override, progressively cranking it up as high as reasonable frame rates allowed with various DLSS settings. That solved my issues getting a decent image out of the Pico 4. The instrument panel is now close to what I get in the sweet spot of the G2 (with some minor compression artifacts) and the scenery immediately below a low flying aircraft is much better, though again not as crisp as the G2’s sweet spot. Distant buildings however are still not as well rendered. But for flying higher it makes the Pico 4 a viable alternative, given its wider FOV and much wider sweet spot due to the pancake lenses.
Thanks again for your input and the ever excellent Toolkit.
I stumbled in to this purely by accident a few months ago and haven’t looked back. Got my 4090 and threw in a 5800x3D, I haven’t been on these forums as much because I hit a spot that has been so great, I’ve been busy flying.
In fact, I’m throwing somewhere around 260% in OXR for super sampling… it works out to somewhere around 5100x5000’ish in-game resolution with my Reverb G2 using DLSS Performance. I was using FSR for sharpening only (don’t need the upscaler, because already upscaling with DLSS).
The clarity is unreal. Target frame rate is 30fps with MR, and typically where I’m at most of the time. I hit 45fps in non populated areas or cruise altitude. And if the weather is bad or in populated areas it can lock at 22.5fps. Once I learned that Asus released a BIOS update for my board, I have PBO Curve Optimizer available, and can do a -30 all core CO… getting 4,450Mhz all core in MSFS with my 5800x3D. Was able to move TLOD up to 200 and still maintain pretty consistent frames.
I don’t tweak settings anymore. Which is crazy, because that’s all I did since April of last year. lol
)
