2.8 update has 20 % fps loss on pimax8kx and 15 fps loss on Varjo Aero (RTX 4090)

[lj0076]

57 minutes ago, nachomaga said:

Opravdu? Přemýšlím, že se vrátím k 2.7, teď mám pochybnosti

Mimochodem, pamatuji si, že před lety to byla hrozná aktualizace VR a mapy Nevady ztratily 50 % výkonu (opět ve VR). Myslím, že oprava ED byla víceméně rychlá, pokud si pamatuji správně.

Problémem je tentokrát ztráta výkonu nejen ve VR. Doufám, že ED brzy vydá opravu.

Try it, maybe it will work for you. I don't know at all what is happening and why, the fact is that everything worked without problems until the 2.8 update.
I suspected the Nvidia drivers also updated on Friday night, but the only problem is again with DCS, the others (IL2, MSFS 2020) work without change.
The fact that I am frustrated is a very mild expression, I think that during more thorough testing before release, someone would have pointed it out. There are simply a lot of people with similar problems.

[BIGNEWY]

4 minutes ago, lj0076 said:

Try it, maybe it will work for you. I don't know at all what is happening and why, the fact is that everything worked without problems until the 2.8 update.
I suspected the Nvidia drivers also updated on Friday night, but the only problem is again with DCS, the others (IL2, MSFS 2020) work without change.
The fact that I am frustrated is a very mild expression, I think that during more thorough testing before release, someone would have pointed it out. There are simply a lot of people with similar problems.

Hi, 

we do test, personally I had a slight fps increase, but not by much. Others have also seen and increase, but also others have seen a decrease. We are investigating. 

thanks

[lj0076]

Před 4 minutami BIGNEWY řekl:

ahoj, 

testujeme, osobně jsem měl mírný nárůst fps, ale ne o moc. Ostatní také zaznamenali a zvýšili se, ale i ostatní zaznamenali pokles. Vyšetřujeme. 

dík

Thank you for looking into this. It's a really big hit, as I wrote before, in my case up to 30% performance in VR.

PS- I'm not sure, but from what I've read, I understand that it's mostly users of kits under WMR...? Just a feeling...

Edited October 31, 2022 by lj0076

[BIGNEWY]

If you are a windows mixed reality user on windows 11 try this 

open command prompt as admin 

type 

logman stop HolographicShell -ets

and press enter 

this stops windows logging your session while using VR and saves some frames 

[Willie Nelson]

5 minutes ago, BIGNEWY said:

If you are a windows mixed reality user on windows 11 try this 

open command prompt as admin 

type 

logman stop HolographicShell -ets

and press enter 

this stops windows logging your session while using VR and saves some frames 

While I’m sure ED is reluctant to ‘Recommend” either Windows 10 or 11, it would be great if we could get some Intel from the developers regarding their thoughts on which way to go especially for those of us that have a 12th Gen processor which theoretically should have greater ability to schedule tasks. 
 

So glad to hear these huge reductions in FPS many of are getting are already being looked at. 

[BIGNEWY]

Just now, Willie Nelson said:

While I’m sure ED is reluctant to ‘Recommend” either Windows 10 or 11, it would be great if we could get some Intel from the developers regarding their thoughts on which way to go especially for those of us that have a 12th Gen processor which theoretically should have greater ability to schedule tasks. 

windows 11 here, and when I made the change I had a FPS increase. 

[Hoirtel]

Oculus Q2 user, i9900ks + 3090. I believe a minor drop in performance mostly affects me low down. I still hit 40fps (reprojection) 95% of the time so I am mostly unaffected. When in rain I still get huge fps loss - impact like nothing else, this was the same with 2.7 not sure why. Rain drops effects make no difference and I can run the biggest clouds without rain fine.

Edited October 31, 2022 by Hoirtel

[zildac]

Just got the latest Varjo BETA and tried 2.8 again. Still seeing the circa 25FPS drop...but I do like the pumpkins at the edge of the taxiways at Caucasus ready on the ramp F18

 

 

[Supmua]

So far I've seen objective evidences of performance drop from people posting detailed graphs of CPU and GPU frametimes, but not the other way around.  The performance increase group seemed mostly subjective which doesn't tell me much. Personally my setup didn't take a big hit (around 2ms in both CPU and GPU frametimes) but I did have to reduce some settings to accommodate.  It's also hard to imagine that the new weather engine with moving clouds, lightings, rain effects, etc. would result in performance increase and not the other way around.  I'm guessing the performance might improve a bit once the shadows bug is fixed, but probably not by much.

[lj0076]

8 minutes ago, Supmua said:

Doposud jsem viděl objektivní důkazy o poklesu výkonu od lidí, kteří zveřejňovali podrobné grafy frametime CPU a GPU, ale ne naopak. Skupina zvýšení výkonu mi připadala převážně subjektivní, což mi moc neříká. Osobně moje nastavení nezaznamenalo velký zásah (kolem 2 ms v rámcích CPU i GPU), ale musel jsem snížit některá nastavení, abych se přizpůsobil. Je také těžké si představit, že by nový meteorologický engine s pohyblivými mraky, osvětlením, dešťovými efekty atd. vedl ke zvýšení výkonu a ne naopak. Hádám, že výkon by se mohl trochu zlepšit, jakmile bude opravena chyba stínů, ale pravděpodobně ne o mnoho.

I am on W10 and the modified weather system was the first thing that came to my mind and what is different against version 2.7. But even when I tried the same missions with completely blue and clear skies, the results were the same. Weather / clouds have only minimal impact in my case, I would almost say none..

[exil]

Okay, back from benching. Thanks to @mbucchia i was able to do it more properly.

Test scenarios were the following: F18 @ Caucasus, F16 @ Syria and AH-64 @ Marianas. 

Here are the results:

F18 Caucasus

CPU FT 2.7: 6.1, CPU FT 2.8: 6.4, Diffrenece: -4.8%

GPT FT 2.7: 13.4, GPU FT 2.8: 14.8, Differece: -10.3%

 

F16 Syria

CPU FT 2.7: 7.38, CPU FT 2.8: 7.35, Diffrenece: +0.43%

GPT FT 2.7: 14.62, GPU FT 2.8: 16.23, Differece: -11.64%

 

AH-64 Marianas

CPU FT 2.7: 9.82, CPU FT 2.8: 10.54, Diffrenece: -7.31%

GPT FT 2.7: 19.60, GPU FT 2.8: 21.34, Differece: -8.91%

 

It's worth to note that i left secondary shadows 'off' for a better comparison.

All in all it's roughly a 10% decrease in GPU Frametimes while CPU may vary. I'll provide the tracks for you guys to test as well if you want to.

F16 SYRIA BENCH.trkAH64 MARIANAS BENCH.trk

@BIGNEWY None of the people i know saw increases in performance and given the number of answears to this threat, i assume the vast majority of the people using DCS won't see any. Nevertheless, it's good to know that ED is looking into it. It's not about people like me who have lost around 10% (that would be perfectly okay for the visual improvements that came with the patch). It's about the people losing 15 FPS total under VR making the game unplayable for them under VR.

 

Couldn't attach the last track because of the 5MB limitation. So here it is:

F18 CAUCASUS BENCH.trk

Cheers!

Edited October 31, 2022 by exil
Attached last track file

[WipeUout]

Had similar drop, 7.5% less FPS in my test track (F-18 Caucasus) with increase in GPU frametime mainly the culprit. 

[slughead]

52 minutes ago, exil said:

All in all it's roughly a 10% decrease in GPU Frametimes while CPU may vary. I'll provide the tracks for you guys to test as well if you want to.

I think you meant an increase in frame time.

[exil]

1 hour ago, slughead said:

I think you meant an increase in frame time.

Yes, sorry I meant decrease in Gpu performance and increase in frametime. 

[darkman222]

@BIGNEWY here are my numbers from my high end gaming pc with a pimax 8kx.

Low preset - f16 free flight. Went down from almost 87 fps to 70 fps.

 

These are the specs from that machine.

 

I also posted the results from my "non VR PC " earlier in this thread. Also a comparable amount of fps decrease although not in VR just in 2D only.

Edited October 31, 2022 by darkman222

[BIGNEWY]

thank you 

[darkman222]

On 10/29/2022 at 4:44 PM, darkman222 said:

And even if its not VR. The frame drop from 2.7 to 2.8 is even in 2D mode.

I tried it on my other PC. ( weaker one with "only" a RTX 2070 instead for 2D playing)

2.7 runs in 2D mode with identical settings at 91-108 fps

2.8 runs in 2D with identical settings at 77 fps.

That would resemble the 20 fps drop in VR too. 20 fps is a lot, which causes to desync the VR headset. If you play in 2D mode you might not realize the drop at the first glance.

2.8

2.7

Although its "just" a 2070 in that PC the specs of it are decent too.

 

 

To keep it together, this is the result from my other PC ( the PC for gaming in 2D) I posted earlier.

Edited October 31, 2022 by darkman222

[BIGNEWY]

We will continue to look at the feedback and the data from 2.8 so thank you for the logs, tracks and results comparisons 

[Supmua]

2 hours ago, exil said:

Okay, back from benching. Thanks to @mbucchia i was able to do it more properly.

Test scenarios were the following: F18 @ Caucasus, F16 @ Syria and AH-64 @ Marianas. 

Here are the results:

F18 Caucasus

CPU FT 2.7: 6.1, CPU FT 2.8: 6.4, Diffrenece: -4.8%

GPT FT 2.7: 13.4, GPU FT 2.8: 14.8, Differece: -10.3%

 

F16 Syria

CPU FT 2.7: 7.38, CPU FT 2.8: 7.35, Diffrenece: +0.43%

GPT FT 2.7: 14.62, GPU FT 2.8: 16.23, Differece: -11.64%

 

AH-64 Marianas

CPU FT 2.7: 9.82, CPU FT 2.8: 10.54, Diffrenece: -7.31%

GPT FT 2.7: 19.60, GPU FT 2.8: 21.34, Differece: -8.91%

 

It's worth to note that i left secondary shadows 'off' for a better comparison.

All in all it's roughly a 10% decrease in GPU Frametimes while CPU may vary. I'll provide the tracks for you guys to test as well if you want to.

F16 SYRIA BENCH.trk 1.56 MB · 1 download AH64 MARIANAS BENCH.trk 3.08 MB · 1 download

@BIGNEWY None of the people i know saw increases in performance and given the number of answears to this threat, i assume the vast majority of the people using DCS won't see any. Nevertheless, it's good to know that ED is looking into it. It's not about people like me who have lost around 10% (that would be perfectly okay for the visual improvements that came with the patch). It's about the people losing 15 FPS total under VR making the game unplayable for them under VR.

 

Couldn't attach the last track because of the 5MB limitation. So here it is:

F18 CAUCASUS BENCH.trk 1.5 MB · 2 downloads

Cheers!

 

Good testing and mirrors my finding.

The CPU frametime on OXRTK is currently not accurate though as it only accounts for CPU rendering only, so it reports lower than the actual overall CPU frametime.  So don't rely on the absolute numbers but you can still rely on the changes.

[markturner1960]

3 hours ago, exil said:

Okay, back from benching. Thanks to @mbucchia i was able to do it more properly.

Test scenarios were the following: F18 @ Caucasus, F16 @ Syria and AH-64 @ Marianas. 

Here are the results:

F18 Caucasus

CPU FT 2.7: 6.1, CPU FT 2.8: 6.4, Diffrenece: -4.8%

GPT FT 2.7: 13.4, GPU FT 2.8: 14.8, Differece: -10.3%

 

F16 Syria

CPU FT 2.7: 7.38, CPU FT 2.8: 7.35, Diffrenece: +0.43%

GPT FT 2.7: 14.62, GPU FT 2.8: 16.23, Differece: -11.64%

 

AH-64 Marianas

CPU FT 2.7: 9.82, CPU FT 2.8: 10.54, Diffrenece: -7.31%

GPT FT 2.7: 19.60, GPU FT 2.8: 21.34, Differece: -8.91%

 

It's worth to note that i left secondary shadows 'off' for a better comparison.

All in all it's roughly a 10% decrease in GPU Frametimes while CPU may vary. I'll provide the tracks for you guys to test as well if you want to.

F16 SYRIA BENCH.trk 1.56 MB · 1 download AH64 MARIANAS BENCH.trk 3.08 MB · 1 download

@BIGNEWY None of the people i know saw increases in performance and given the number of answears to this threat, i assume the vast majority of the people using DCS won't see any. Nevertheless, it's good to know that ED is looking into it. It's not about people like me who have lost around 10% (that would be perfectly okay for the visual improvements that came with the patch). It's about the people losing 15 FPS total under VR making the game unplayable for them under VR.

 

Couldn't attach the last track because of the 5MB limitation. So here it is:

F18 CAUCASUS BENCH.trk 1.5 MB · 2 downloads

Cheers!

 

Although important to note that in all those tests, the weather is fairly benign and the performance is still degraded? ? I noticed that changing the weather from a sunny windy day preset, to a thunderstorm, I was losing perhaps 20 - 30 % performance as measured purely by FPS in 2D. It cant just be the weather eye candy that is the issue, ( although its a shame that these improvements all come at a price that negates significant outlays in hardware etc). I am curious as to why the devs were promising improvements in VR performance, when no one is seeing this and usually, a decrease? Were they seeing something different? Why advertise something they had not been able to achieve otherwise? They must have found some improvements and been excited enough about it to advertise it in advance of the patch........

[Werewolf_fs]

hace 5 horas, BIGNEWY dijo:

If you are a windows mixed reality user on windows 11 try this 

open command prompt as admin 

type 

logman stop HolographicShell -ets

and press enter 

this stops windows logging your session while using VR and saves some frames 

The problem with windows 11 is specifically with version 22H2 and WMR headset, but it was already before 2.8, and it is effectively solved by disabling it HolographicShell, but I already did it with said mod and there is a loss of 2-3 ms both CPU and GPU between 2.7 and 2.8. That's not the problem

Edited October 31, 2022 by Werewolf_fs

[zildac]

1 minute ago, markturner1960 said:

Although important to note that in all those tests, the weather is fairly benign and the performance is still degraded? ? I noticed that changing the weather from a sunny windy day preset, to a thunderstorm, I was losing perhaps 20 - 30 % performance as measured purely by FPS in 2D. It cant just be the weather eye candy that is the issue, ( although its a shame that these improvements all come at a price that negates significant outlays in hardware etc). I am curious as to why the devs were promising improvements in VR performance, when no one is seeing this and usually, a decrease? Were they seeing something different? Why advertise something they had not been able to achieve otherwise? They must have found some improvements and been excited enough about it to advertise it in advance of the patch........

This is my point, I’m using F18 Caucasus ready on the ramp, no weather, nowt And 2.7 86FPS 2.8 60FPS

[exil]

57 minutes ago, Supmua said:

Good testing and mirrors my finding.

The CPU frametime on OXRTK is currently not accurate though as it only accounts for CPU rendering only, so it reports lower than the actual overall CPU frametime.  So don't rely on the absolute numbers but you can still rely on the changes.

Thanks for the insights! I recently upgraded to the 5800X3D and was aware that it is a very potent CPU. 

But I was already wondering that it would be that good. 

But even if the numbers aren't too accurate, I'm still Gpu bound most of the time I guess. I am looking for a steady 45 FPS reprojection (which luckily still works after the patch). 

 

[Supmua]

23 minutes ago, exil said:

Thanks for the insights! I recently upgraded to the 5800X3D and was aware that it is a very potent CPU. 

But I was already wondering that it would be that good. 

But even if the numbers aren't too accurate, I'm still Gpu bound most of the time I guess. I am looking for a steady 45 FPS reprojection (which luckily still works after the patch). 

 

Yes, I found OXRTK’s CPU FT to be off by 4-5ms in MSFS where you can run both fpsVR and OXRTK at the same time via Steam OpenXR, no way to do this with DCS.

I just upgraded (the CPU) this weekend also, and I’ve been pretty much GPU bound in DCS >90% of the time since 2.8 beta.  Haven’t tried MP server yet though, mostly trying to mess around optimizing the PC since I’m new to the AMD system.

What would be really neat is for someone to add (unofficial) DLSS mod to DCS for significant GPU FT relief, if modders can do this to Skyrim/Fallout 4/Resident Evil they should be able to make it happen for DCS also.

[exil]

14 minutes ago, Supmua said:

Yes, I found OXRTK’s CPU FT to be off by 4-5ms in MSFS where you can run both fpsVR and OXRTK at the same time via Steam OpenXR, no way to do this with DCS.

I just upgraded (the CPU) this weekend also, and I’ve been pretty much GPU bound in DCS >90% of the time since 2.8 beta.  Haven’t tried MP server yet though, mostly trying to mess around optimizing the PC since I’m new to the AMD system.

What would be really neat is for someone to add (unofficial) DLSS mod to DCS for significant GPU FT relief, if modders can do this to Skyrim/Fallout 4/Resident Evil they should be able to make it happen for DCS also.

Yeah, that would be awesome. I am on DLSS (2) on MSFS. And oh boy, the improvements were huge! Really curious to see DLSS 3 in November (you can already test it within the beta, but I won't touch it for now). 

If DCS would bring DLSS to life, really a lot of the gpu performance issues would be solved! But I wouldn't hold my breath... 

Edited October 31, 2022 by exil