VR experiences IL-2 Korea

[lefuneste]

Maybe already written here, but the default DLSS settings is "M", which is a fps killer even for my 5080. Changing it to "K" allowed me to set DLSS "quality", having DLAA and 90 fps on my test scenario on ground near Seoul. Otherwise it was less than 50 fps...Image quality is good enough for me, at least for ground attack.

Edited July 17 by lefuneste

[Jade_Monkey]

3 hours ago, lefuneste said:

Maybe already written here, but the default DLSS settings is "M", which is a fps killer even for my 5080. Changing it to "K" allowed me to set DLSS "quality", having DLAA and 90 fps on my test scenario on ground near Seoul. Otherwise it was less than 50 fps...Image quality is good enough for me, at least for ground attack.

 

It has been this way for a while and it's not any single party's fault, but it's exhausting how many different pieces of software need to be tweaked to get VR working well 😭

 

Pimax, SteamVR, toolkit, quad views, Nvidia. Sboys3, ..... 

 

What a mess.

Edited July 17 by Jade_Monkey

[YoYo]

Btw. Posted by @LuftManu

It sounds very positive and optimistic :)!

[Hotdognz]

On 7/17/2026 at 11:23 AM, Panzerlang said:

Distant Landscape, in my experience at least, has to be on X4. Since then I've had zero issues with the landscape looking crap. 🙂

It also hasn't noticeably impacted FPS in VR.

Even on 4x its looks a blurry mess in the distance, by my nest guess anything out side of 5km radius around my plane has absolutely no detail in VR

[Husar]

3 hours ago, Hotdognz said:

Even on 4x its looks a blurry mess in the distance, by my nest guess anything out side of 5km radius around my plane has absolutely no detail in VR

But contact spotting against that blurred textures is very easy 😂

[YoYo]

4 hours ago, Hotdognz said:

Even on 4x its looks a blurry mess in the distance, by my nest guess anything out side of 5km radius around my plane has absolutely no detail in VR

It doesn't look bad at x4. I also think the objects are generated just right for this distance. It was similar in the IL-2 GB. Keep in mind that increasing the distance will cause even more FPS drops, and no one wants that. Personally, I think it's fine 👍. The terrain is also sharp at x4 (although there are occasional areas that are less sharp, which sharpen after a while; something similar happened in the IL-2 GB). Below are screenshots from the headset and the x4 distance settings, 150km, landscape filter sharp.

[GiftGruen]

On 7/15/2026 at 9:58 PM, Mamakin said:

I wanted to wait before buying, but I decided to show my trust and support. However, following the official blog post, I would now like the option to get a refund. What I read really left me stunned. 

See my post in the Appreciation thread: Needed a lot of experiments, but now - I am happy. This engine is good, it's just not mature yet. You need some time to try out and optimize.

Some optimizations seem to be on the horizon. We'll see how much % they mean.

Based o my personal measurements I'd say that Korea runs approx 50% slower in VR in comparison to IL2 BoX. 

For my rig, I now covered this by reducing resolution by 10% vertical+horizontal ( which is 20% less pixels ) and lowering some settings ( e.g. MSAAx2 instead of MSAAx4 ).  50% seems to be a lot, but the graphical fidelity of the Korea engine, if you get it run stutter-free, is a really big step forward.

On the other hand Koreas seems to be more forgiving in terms of CPU loads and bigger formations. At lest that's my impression with Mig15 trainin mission #4 which uses quite a pack of planes (25? 30?)

 

[pr0x]

On 7/17/2026 at 1:11 PM, lefuneste said:

Maybe already written here, but the default DLSS settings is "M", which is a fps killer even for my 5080. Changing it to "K" allowed me to set DLSS "quality", having DLAA and 90 fps on my test scenario on ground near Seoul. Otherwise it was less than 50 fps...Image quality is good enough for me, at least for ground attack.

M is an fps killer only on balanced and higher. It is trained on fairly low input resolution unlike K, and is therefore more efficient at performance and performance+. The main advantage of M however, is that it gets rid of the ghosting trail behind the planes. K on the other hand, seems to make planes easier to spot far away.

[lefuneste]

2 hours ago, reidarpeidar said:

M is an fps killer only on balanced and higher. It is trained on fairly low input resolution unlike K, and is therefore more efficient at performance and performance+. The main advantage of M however, is that it gets rid of the ghosting trail behind the planes. K on the other hand, seems to make planes easier to spot far away.

I just confirm there is too much ghosting trail with K. I did not noticed it when playing last exams for F80, in which you are flying with a squad, but it was quite painfull on combat box yesterday. Maybe linked to cloud or haze ? I removed DLSS anyway. No AA plus ground blurring is my way to go to have good fps, image quality and good spotting, as not too much shimmering.

[YoYo]

30 minutes ago, lefuneste said:

too much ghosting

Even with M, ghosting is visible, I saw it in the IL-2 Korea as well, which is why I don't fly with any DLSS, but MSAA x2. Unfortunately, I can't stand that "Phoenix dust" trailing behind the plane. Even in MSFS2024 I can tolerate it (civilan flying there is not so dynamic), but in the coombat simulator it spoils the final effect for my :/.

[Husar]

3 hours ago, reidarpeidar said:

M is an fps killer only on balanced and higher. It is trained on fairly low input resolution unlike K, and is therefore more efficient at performance and performance+. The main advantage of M however, is that it gets rid of the ghosting trail behind the planes. K on the other hand, seems to make planes easier to spot far away.

Did you test prest L ? Less performance impact vs M and similar small amount of smearing, not that bat as K where it's hard to judge plane aspect.

[Antonovomatic]

On 7/16/2026 at 8:06 PM, huligan said:

That's exactly what I'm doing, but it still turns me upside down! (on vertically)

Make sure you don't have a per-game/app profile, otherwise I'm out of suggestions unfortunately.  A clean install  if you upgraded from an earlier release might be in order

[Antonovomatic]

On 7/17/2026 at 10:01 AM, Jade_Monkey said:

 

It has been this way for a while and it's not any single party's fault, but it's exhausting how many different pieces of software need to be tweaked to get VR working well 😭

 

Pimax, SteamVR, toolkit, quad views, Nvidia. Sboys3, ..... 

 

What a mess.

You shouldn't be using toolkit at all, least of all combined with quad views... which also isn't compatible with Korea. 

You can cut another two out by going with OpenXR which is far superior to SteamVR unless you absolutely NEED adjustable position offsets.  Just make sure your resolution multipliers are at truly x1.0/native HMD res in all the backend (Meta slider stated multipliers are untrue and arbitrary for ex.), and tweak with Korea's own in-game resolution slider.

13 hours ago, Husar said:

Did you test prest L ? Less performance impact vs M and similar small amount of smearing, not that bat as K where it's hard to judge plane aspect.

That's not correct, L is the slightly more intensive of the two 2nd Gen "Transformer" models.  It's a great option if M looks oversharpened, and handles foliage a bit better, especially at lower res scales.  In flat screen games I can often drop res scale significantly without fidelity loss by manually selecting L preset, it's my favorite esp. for games that can accept custom res. scale input ("Ultra Perf." is still too low for my tastes on 4k, running several games ~40% scaling and it's perfect)

On 7/17/2026 at 6:11 AM, lefuneste said:

Maybe already written here, but the default DLSS settings is "M", which is a fps killer even for my 5080. Changing it to "K" allowed me to set DLSS "quality", having DLAA and 90 fps on my test scenario on ground near Seoul. Otherwise it was less than 50 fps...Image quality is good enough for me, at least for ground attack.

Might as well go with 2x MSAA then, the VRAM hit and GPU loading is going to be equivalent but without the ghosting of DLSS.  DLAA is roughly equivalent to 4x MSAA, "Quality" preset to 2x MSAA for the modern (read: worth trying) DLSS SR presets.  Even on Model K, "Balanced" preset just barely claws back any performance increases over native res w/o any AA/scaling, anything above that is a net performance drop over native, and in VR usually visually worse too - depending on whether jaggies or blurring/ghosting bother you more.

Edited July 21 by Antonovomatic

[Husar]

5 minutes ago, Antonovomatic said:

That's not correct, L is the slightly more intensive of the two 2nd Gen "Transformer" models.  It's a great option if M looks oversharpened, and handles foliage a bit better, especially at lower res scales.  In flat screen games I can often drop res scale significantly without fidelity loss by manually selecting L preset, it's my favorite esp. for games that can accept custom res. scale input ("Ultra Perf." is still too low for my tastes on 4k, running several games ~40% scaling and it's perfect)

I know Nvidia paper on it , but I tested in Korea and had more  fps vs M

[YoYo]

IPD (RShift + and -) in OpenXR started working after 003. Unfortunately, I feel like it creates an additional squint effect. The world gets bigger, but at a certain point the image becomes essentially double. It looks good at -9 for me, but something's still wrong. Unfortunately, the configuration works in 3-point increments, not 1-point increments. So, 3, 6, 9, etc.

[GiftGruen]

Hmmmm. In my test scenery - Training Misson #4 Mig15 ‚combat‘ - at least I cannot see and/or measure ( I do this with fpsVR statistics )  performance benefits from Patchlevel 03.

Of course flying such a mission is not an exact measure, and also my settings maybe don‘t benefit - but for my setup it‘s for quite sure not a BIG step forward.

What‘s your first performance Impressions on PL03 at first sight?

Edited July 21 by GiftGruen

[huligan]

1 час назад, Antonovomatic сказал:

Make sure you don't have a per-game/app profile

no.

I use defolt profile only

1 час назад, Antonovomatic сказал:

clean install

game? or vectorXR ?

after update vectorXR to 0.13.7 + Korea 0.003 pitch disabled all ok.

But I can't binding POV on my VKB Gladiator Pro in pivot on reset view. VectorXR doesn't see my clicking POV buttons

Edited July 21 by huligan

[YoYo]

33 minutes ago, GiftGruen said:

Hmmmm. In my test scenery - Training Misson #4 Mig15 ‚combat‘ - at least I cannot see and/or measure ( I do this with fpsVR statistics )  performance benefits from Patchlevel 03.

Of course flying such a mission is not an exact measure, and also my settings maybe don‘t benefit - but for my setup it‘s for quite sure not a BIG step forward.

What‘s your first performance Impressions on PL03 at first sight?

I have the impression that over Seoul it is the same as it was. The city still has a significant FPS drop. The changelog didn't mention any optimizations in any specific aspect.

[huligan]

54 минуты назад, huligan сказал:

But I can't binding POV on my VKB Gladiator Pro in pivot on reset view. VectorXR doesn't see my clicking POV buttons

answer from github

Цитата

Collaborator

Thank you for bringing this to my attention at @huliqan . HAT binding capability will be added in the next release, slated for sometime this week.

 

[huligan]

Who knows how to turn off pi-pi-pi...morse code in the cockpit during flight? 

[YoYo]

1 minute ago, huligan said:

Who knows how to turn off pi-pi-pi...morse code in the cockpit during flight? 

Radio volume down or next, prev NDB station (it depends).

[Hotdognz]

On 7/19/2026 at 9:12 PM, YoYo said:

It doesn't look bad at x4. I also think the objects are generated just right for this distance. It was similar in the IL-2 GB. Keep in mind that increasing the distance will cause even more FPS drops, and no one wants that. Personally, I think it's fine 👍. The terrain is also sharp at x4 (although there are occasional areas that are less sharp, which sharpen after a while; something similar happened in the IL-2 GB). Below are screenshots from the headset and the x4 distance settings, 150km, landscape filter sharp.

What locations are the pics taken YoYo, ill liked to compare what i see as I have now copied your IL2 Kora graphic settings in game, as yours looks a lot less blurry than mine

[Antonovomatic]

3 hours ago, Husar said:

I know Nvidia paper on it , but I tested in Korea and had more  fps vs M

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.

[Drewm3i-VR]

How is the performance after .03?

Edited July 22 by Drewm3i-VR

[huligan]

2 часа назад, YoYo сказал:

Radio volume down or next, prev NDB station (it depends)

found 🙂

Edited July 22 by huligan