VR experiences IL-2 Korea

[huligan]

1 час назад, Antonovomatic сказал:

DLSS presets

All very bad!(((

[YoYo]

4 hours ago, Drewm3i-VR said:

How is the performance after .03?

I didn't have much time yesterday, but I didn't notice any noticeable improvement over Seoul. New buildings and elements have been added, but I still have a noticeable FPS drop. I've seen positive posts, but they're probably not about VR, but rather 2D.

[YoYo]

IPD work nows fpr Quest and OpenXR, but it's not just World Scale only like in the IL-2 GB or DCS. Here, it shifts the entire image for the left and right eyes. Well, that's how it should work, but it doesn't solve the problem of small cockpits for me. It looks best at "-6" for me, at least with the available settings. The problem is that it can't be precisely adjusted because the IPD increments are 3 degrees, not 1. Secondly, in the current situation, the image is set differently for the left and right eyes, creating the effect of strabismus and double vision (the closer to 30 degrees, the more noticeable it is, as each screen displays a shifted image). As a result, we have a problem with image convergence, which simply looks bad and is tiring for the eyes (it's impossible to fly with it; IPD 0 still looks natural and neutral, but the cockpits here are too small). So, the problem has been partially solved. Yesterday, out of curiosity, I fired up the IL-2 GB to check how it works, I have it set to 0 and the size of the cockpits are ok, and the IPD operation (LShift + NumEnter and Num+) works, it only changes the scale of the World, the convergence is always ok and it always looks good. So there is still no perfect solution here in IL-2 Korea. Changing the physical IPD in the headset does not help here by the way. In my opinion, the team 1CGS should look at this again and the IPD itself should work like in the IL-2 GB (the best way).

@Dash,Polder I remember you had some problem with convergence, but I think it was on Pimax, can this VectorXR program solve it? It seems that this matter has now affected Quests.

[Dash,Polder]

Here's the description, yes it can set the illusion of a larger pit.  None of these actually change the size of anything, it's just a trick to fool your brain into perception of a size change.  If one takes a screen shot of before and after there will be zero difference in the screen shot.  But here's the description from the app itself.

Have it at zero cause I was still tinkering with graphical settings from the new patch.  I can see the river currents now, they look very strong.  Some amazing stuff with a zillion settings and things to mess with.  My game is running awesome and spotting has never been so good.  Brand new 57 ppd Pimax, it's nearly perfect in Korea.  My nephew has the 50 ppd and struggles with VRAM always full and than can't crank up the candy, same systems GPU and CPU.

[Husar]

6 hours ago, Antonovomatic said:

That's just scene variance, all controlled testing has shown increased workload with L.  IL-2 is CPU intensive, and when GPU workload/VRAM is the properly isolated test variable the performance impact of DLSS presets are consistent and predictable.

I was testing 4 vs 8 AI skirmishes with only DLLA preset K,L,M without upscaling.

[=ILS=_Abu_27]

按这个设置，可以达到90帧稳定。 小派air 画质设为0.8

[Picchio]

4 hours ago, =ILS=_Abu_27 said:

With this setting, a stable 90 frames per second can be achieved. The Pimax Air's image quality is set to 0.8.

 

Screenshot contents don't get translated by the browser, just so you know.

[pr0x]

I think it's a bit naive to think that you can run a modern title in VR in high resolution without DLSS. And MSAA causes shimmering, which to me is a big no-no. I might as well run FXAA then.

When considering DLSS presets it's important to understand that output resolution also matters. DLSS presets are trained on upscaling from a certain input resolution to a certain output resolution. The further away from these resolutions you go, the less effective the model is going to be. So it's not at all surprising that two people see different results from M and L when the input and output resolutions are not given.

The K preset is the older generation which is more or less meant to sharpen an already high resolution image. M and L are the heavier models which are more aligned with NVIDIA's current philosophy of allowing AI to be more inventive. Both M and L are trained on low input resolutions and high output resolutions. L more so than M. If you're looking for a performance gain, it doesn't make much sense to run the M and L presets in Quality or DLAA mode.

The L preset combined with a big difference between input and output resolution is certainly a very interesting setting. I tried running Virtual Desktop Godlike (roughly 3.2kx3.2k pixels) and DLSS set to Performance and Performance+ in L mode. My assessment of this is the following:

1. It yields a massive performance boost. Flying in formation with 15+ planes in clouds, I'm seeing 110 fps with no fake frames

2. The image looks quite soft and natural while retaining sharpness where it matters. There is no shimmering

3. It works well with sharpening. VD sharpening set to 100% still looks good.

4. There is ghost trails, but not as much as with K. When I increased the DLSS setting to Balanced, ghosting was significantly reduced.

5. Long-range spotting is quite good, unlike M, which sometimes renders far-away planes invisible.

6. It doesn't smear close-up textures like M does

Obviously, running Performance+ is not ideal, but if you're desperate for FPS, even Performance+ looks decent with the L preset.

[GiftGruen]

11 minutes ago, reidarpeidar said:

I think it's a bit naive to think that you can run a modern title in VR in high resolution without DLSS. And MSAA causes shimmering, which to me is a big no-no. I might as well run FXAA then.

I'm not so deep inside DLSS. My first 'experiments' were rather disappointing, but I will for sure give it another try to find out how it works.

MSAA creates a lot of load - and therefore everyone using it would be quite happy to get rid of it, if only other AA algorithms would provide a similar ( or even better ) result with less GPU load. However, setting FXAA more or less equal to MSAA in my experience is simply wrong, especially when you talk about shimmering.

 

[Blitzen]

71 fps in exterior view- 31-36 inside cockpit, no matter what my settings are...just sayin'...🤓

[Husar]

33 minutes ago, reidarpeidar said:

doesn't make much sense to run the M and L presets in Quality or DLAA mode.

It does if one wants get ride of jagged lines and much of ghosting regardless of performance 

Edited July 22 by Husar

[pr0x]

46 minutes ago, Husar said:

It does if one wants get ride of jagged lines and much of ghosting regardless of performance 

Uh, yeah, but who in this thread wants anything regardless of performance? The premise that you conveniently left out of the quote was this:

IF YOU'RE LOOKING FOR A PERFORMANCE GAIN.

[=ILS=_Abu_27]

5 hours ago, Picchio said:

 

仅供你注意，浏览器不会翻译截图内容。

yeah i forgot and when i picture them its chinese version already , but i think its easy to compare .  the important thing is if it can help people  get a good performance of vr .

[OAE]

How do you change the "preset" in dlss? Dlss swapper doesn't work here..!? 

[Qcumber]

28 minutes ago, OAE said:

How do you change the "preset" in dlss? Dlss swapper doesn't work here..!? 

It's now built into the NVIDIA app. You can override the preset or go with the recommended one. 

[Dash,Polder]

My app only showed presets A and B, but they can be set using profile inspector, than they show up in NVIDIA app, but only the letter you currently have selected in inspector in addition to A and B, whatever those are. 

[Antonovomatic]

1 hour ago, Dash,Polder said:

My app only showed presets A and B, but they can be set using profile inspector, than they show up in NVIDIA app, but only the letter you currently have selected in inspector in addition to A and B, whatever those are. 

Those are the frame gen presets, super resolution should list A-M and ray reconstruction A-E

Strong recc for "NVPI-R" application, much cleaner and easier to read interface than the original NVPI

 

4 hours ago, GiftGruen said:

I'm not so deep inside DLSS. My first 'experiments' were rather disappointing, but I will for sure give it another try to find out how it works.

MSAA creates a lot of load - and therefore everyone using it would be quite happy to get rid of it, if only other AA algorithms would provide a similar ( or even better ) result with less GPU load. However, setting FXAA more or less equal to MSAA in my experience is simply wrong, especially when you talk about shimmering.

 

DLSS on any of the Transformer models (J, K, L, M) is already almost as intensive as MSAA unless you are using sufficently low res scaling factor (balanced for J/K, perf or ultra perf. for L/M).

[huligan]

19 часов назад, Drewm3i-VR сказал:

How is the performance after .03?

AND...

You can now binding HAT-buttons(POVs) in VectorXR 0.14.0- pivot (turn view to "6")

https://github.com/DienerTech/vectorxr/releases/download/v0.14.0/VectorXR-0.14.0-windows-x64-setup.exe

 

Edited July 22 by huligan

[huligan]

14 часов назад, YoYo сказал:

IPD work nows fpr Quest and OpenXR

dos't work from VD+Pico4+VDXR

[Dash,Polder]

If you guys want to try getting all those FPS back over those cities.

 

[huligan]

1 час назад, Dash,Polder сказал:

9950X3D with 4090

With such hardware, is anything else really needed?😄

 

[Dash,Polder]

6 hours ago, huligan said:

With such hardware, is anything else really needed?😄

 

More bleeding edge hardware one has the more the windows flaws show and put the brakes on your performance.   

[huligan]

VectorXR-pivot in KOREA It can be replaced with the dynamic camera mode built into the game.

pressed the button(LAlt+F1 or I binding on LCtrl +hat(POV) to the left)-accelerated rotation, pressed F1 (or I binding F1 on hat(POV) to the left)-a normal turn from the head

.. only there is a nuance (bug of the game) .After starting the game and the 1st start of the mission, this dos't work in VR .You need to end the mission, back into the editor, and restart the mission.Then the dynamic camera mode button is already working.

AND !!! You definitely need to return to the mission editor!!! And launch the mission from there!

...this is instead of the VectorXR-pivot program

 

[Jade_Monkey]

Cross posting this here from the other thread for awarenes:

TLDR: Delete your startup.cfg file and let the game create a new one. Then readjust your graphical settings to what you were using again.

Frametimes dropped and performance increased significantly.

 

Edited July 24 by Jade_Monkey

[huligan]

3 часа назад, Jade_Monkey сказал:

Cross posting this here from the other thread for awarenes:

TLDR: Delete your startup.cfg file and let the game create a new one. Then readjust your graphical settings to what you were using again.

Frametimes dropped and performance increased significantly.

is this magic? This doesn't make any sense ! how can such a hack work? 

Edited July 24 by huligan