Possible soluiton to stutters and FPS in MP (and subsequently SP)? Let me know

[mtd2811]

Hi all

 

I may have come across a potential solution for VR stutters and low/smooth FPS here. Forgive any typos, English is not my first language.

 

Can you disable NV Shader Cache in NVIDIA CP and delete everything from

 

C:\ProgramData\NVIDIA Corporation\NV_Cache

 

My setup is primeraly done for DCS. My win10 (latest 64 version) computer is clean of any bloatware, unecesserry montoring software etc. DCS is installed on a dedicated SSD with only DCS on. Other PC specs:

 

1. Intel 4790K OC to 4.8

2. 3080 , latest NV drivers

3. HP G2 set at 1.0PD in game and 100% SS

4. 32GB rame

5. Pagefile is set at 32,684 min and max, only to my SSD where DCS is on and no other drive (I have 2 SSDs and 2 m.2s.)

6. My gaming desktop is not connected to any microsoft home group or any other network (dont want windows to be pinging devices during MP sessions)

7. I connect via ethernet not wireless addapter

 

I assume you have done all the fixes flaoting in this subforum with regards to

 

1. shaders modes IC pass

2. 2.5.6.xxxx Adjusted shaders

3. CPU priority registry setting for DCS,VRServer,  VR Compositor,

4. shadow mods as noted by other users.

5. Windows GPU Accellerated graphics is off.

 

The above mods pass IC. Apologies i dont have direct links at this point,  but a bit of search in the mod section of these forums and User files should bring them up.

 

If you aplpy the shaders mods and after you tweak the NVIDIA Shaders mods be sure to delete FXO, Metashders in your user fodler of the game and the metashders in the DCS main folder (mods, terrain etc I assuem you know how to )

 

My NVIDIA CP settings follow Bignewy's set up posted here, but slighlty modified:

 

 

 

 

In game settings and generally whilist in main menum i am solid 90 fps with 1PD and 100SS at SteamVR

 

 

Online

I tried the TTI and 4YA servers , intentionally with 20+ people in and ALL maps.

 

On the ground i had solid 45-55 FPS 98% of the time with only very minor drops to 37-40 for a split second (usually when people log in or log out the server)  but at no point I had any stutters, which i had untill yesterday, before the I moved the pagefile as noted above and before i unchecked Shaders cache in NVCP.

 

Once airborne and about 3000ft my fps was jumping from 70-90 in the cockpit (F16 which is demanding) and 90 fps solid on external views.

 

Smooth as well, even when my fps was flactuating and no stutters.

 

The way I run the game.

 

1. Open WMR

2. Fire DCS.

3. Let steamvr load

4.Log in your prefered server.

5. Jump on a module

6. Give it a minute to load it fully and settle memory demands or something

7. In case you have the odd minor stutter and FPS drops,  alt-tab for 10-20 seconds (window out of focus) and your fps should be now fine at least on teh ground at half your HMDs refresh rate

 

PS: If the cockpit doesnt load wihtin 5-10secs give it a minute it will load.

 

Let me know if this works for you and Happy New Year to all. Good health above anything else and good landings.

 

 

 

 

 

[mtd2811]

I forgot to add:

I am in betas for both SteamVr and WMR for SteamVR.


Sent from my iPhone using Tapatalk

[Glide]

Nice work!  See my discussion of shader cache here:

 

[Glide]

Just some comments to help tune it up. 

 

FXAA is post-processing.  This means that it's applied AFTER DCS is done rendering.  You need to use Alt+Printscrn to capture FXAA.  Since you are (correctly) not adding any AA load to the shader cache in your DCS settings, it does not add anything to have AA Gamma Correction on in the NCP. 

 

Since you have high textures, you might as well have Anisotripic 16x and Trilinear Optimization on (as you do).  These help make great textures look even better.  High textures put a higher load on the Shader Cache, I find Medium Textures are a nice compromise because Low Textures make the cockpits ugly, but with Medium textures I turn Anisotropic Filtering down to 8x and turn TriLinear Optimization OFF because why waste the energy on lower textures.

 

Turn off Trees, Grass and Clutter.  These things have a lot of geometry and require a lot of shading making the Shader Cache work hard.  I know the WWII folks may not like this setting.

 

Texture Filtering -Negative LOD bias should be Allow. 

 

Finally, try turning off all Vsync/Gsync settings and let your card go as fast as it can.  This would mean adjusting the Low Latency Mode in your NCP.  I find with FXAA and Shader Cache OFF, you don't need Vsync to make it smooth.  I haven't seen a stutter in weeks.

 

Cheers, and happy flying.

Edited January 1, 2021 by glide

[mtd2811]

Thanks ; I have been following your post and from another couple people.

My main aim was to reduce and/or eliminate those annoying sudden stutters in VR, which my above settings seem to do that especially the shader cache and the pagefile on the DCS drive and maintain quality.

Having said that I will try your tips and see how it performs in MP


Sent from my iPhone using Tapatalk

[Glide]

Those sudden stutters, reduce your visibility range down to High, and reduce the preload radius to at least 3/5ths of the range.  The visibility range is like a bubble that surrounds the player as they fly, but only the part that's in the viewport gets rendered.  When you swing your head around, all those textures coming across the boundary need shading.  The larger the visibility range, the more pixels are crossing over the bubble boundary that will need shading too. 

 

Let me know how it goes!

[Glide]

Oh yes, shadows.  Flat shadows just stick to things.  They don't move around and call for things to be reshaded all the time.  Your jet will cast shadows around, but the things on the terrain and inside the cockpit will stay in place.  These things get reshaded as you change your angle to the sun or moon (the only two light sources in the game, i believe, I don't think muzzle flashes cast light.  This would require clustered shading or Vulcan?).  Flat shadows and no terrain shadows are the least stressful on the shader cache.

Edited January 1, 2021 by glide

[Glide]

This might be a question for ED: if the Sun and Moon are both up, are they fighting each other over the shader cache? Can you "turn off" the moon for testing?  I always try to have the moon around in my missions because it adds a nice cinematic touch.  Good way to test: take a (multiplayer) mission that lags, move it to a day that has no moon, and compare.

[mtd2811]

Sticking with my settings as it performs nicely and I have virtually no stutters in MP; and also smooth visuals even when the fps fluctuates (doesn’t go below 45 and up to 90).

I tested some of your other tips and only going with the textures to med and x8 anisotropic filtering. As you said the cockpits still look acceptable


Sent from my iPhone using Tapatalk

[Glide]

Nividia recommends Low Latency Mode = Ultra with Vsync = OFF.  Git that a try.  Seems to smooth my FPS out.

 

https://www.nvidia.com/en-us/geforce/guides/system-latency-optimization-guide/

[mtd2811]

Nividia recommends Low Latency Mode = Ultra with Vsync = OFF.  Git that a try.  Seems to smooth my FPS out.
 
https://www.nvidia.com/en-us/geforce/guides/system-latency-optimization-guide/

Thanks will try it although at fast seems to be fine so far and smooth.

Yeah thanks I think I had it at ultra already.

I have seen on another post a user mentioning steam vrserver and compositor where spiking the gpu creating stutters.

Not sure if indeed it has any effect but i added the above processes in the nvidia panel and set them to “no shaders” , and prerendered frame to “let the application decide” for each. I dis the same for WMR processes.

When I was launching wmr and steamvr I could see the NV cache filling up thats why i decided to set them to no shaders.

It is much much smoother and not once any steamvr processes “spiked” or anything all day today and the experience was amazing with even High shadows in MP!

45fps on the ground (TTI and all 4YA servers) and 45-75 airborne. External views above 3000ft 90solid


Sent from my iPhone using Tapatalki

[Glide]

I know, it's amazing dogfighting now!  Glad things are working.  I turned up all my quality settings with Ultra Low Latency.  BTW, if you have Reflex (my 1080ti doesn't), that's better than Ultra Low Latency according to that article.  

[mtd2811]

I know, it's amazing dogfighting now!  Glad things are working.  I turned up all my quality settings with Ultra Low Latency.  BTW, if you have Reflex (my 1080ti doesn't), that's better than Ultra Low Latency according to that article.  

Quite right! Fast turning heads and smooth as hell while spotting!

For reflex, I dont think DCS supports this?


Sent from my iPhone using Tapatalk

[Glide]

3 minutes ago, mtd2811 said:

For reflex, I dont think DCS supports this?

Could be, it's very new.  I can't test it with my card. 

 

My heart was literally pounding in my chest waiting for a stutter to happen for the first couple of days. 

[mtd2811]

Could be, it's very new.  I can't test it with my card. 
 
My heart was literally pounding in my chest waiting for a stutter to happen for the first couple of days. 

For me stutters were created because of the steamvr processes for some reason (probably 3000 drivers and the knows issues with steamvr). I could see on occasions steamvr “Waiting” in the background with its window popping up whilist in game and the game freezing for a few seconds - straight after that fps would deop and would be with stutters also.

Thats why i did the above. So far for me SP or MP not once I got the steamvr “waiting” window whilst in game so something must have worked


Sent from my iPhone using Tapatalk

[Glide]

I have TrackIR, and I could create a stutter on demand by turning my head past the sun.  I am guessing that if the shader cache hiccups then the handoff between VR/TrackIR and DCS gets interrupted. 

[mtd2811]

I have TrackIR, and I could create a stutter on demand by turning my head past the sun.  I am guessing that if the shader cache hiccups then the handoff between VR/TrackIR and DCS gets interrupted. 

Quite right same thing was happening to me in VR as well.

But stopped since the changes , so very happy now with how it performa.


Sent from my iPhone using Tapatalk

[Pikey]

I didn't 'notice' any positive gains following this, but its' fairly hard to tell because to introduce one of the many different types of stutter you have to introduce complexity to the mission which affects your overall performance - so very difficult if not impossible to say. The PD was three tenths larger, the SS much lower (this is contrary to the low PD, high SSS ratio that has generally been found to give back many frames.) I didnt understand the saving on AA which doesnt cost according to extensive testing. I also added Kegetys Shader mod at the same time, so again, very difficult to tell. I also didnt follow all of the prerequisities which were many, they required searching and I'd prefer not to guess at them.

 

90 FPS in the menu is normal for any settings of course. I can get 90 in single player in cockpit up high in many cases, where this stops is on the ground and in multiplayer and with mission complexity. 1400 unit server with scripting and lots of AI - 35/50 is really good for my 2080S and i5 8600.

 

The parts I think where this is interesting are the NV shader cache and low latency mode, which I believe is the point of this post. Maybe its doing something, maybe its intereacting with the slower IO of a disk, I dont know. Id like to say it did something but removing the windows virtual monitor for VR did way more in effect, so rather than say, 'it feels good', I will just say 'it doesn't feel bad'. The word "stutter" is overused and misused by many people, which unfortunately doesnt help the community progress and spreads misinformation. So I cannot say, I do or don't get stutters following this advice. I certainly get some pauses which last visibly a very long time inside the second, but I think the cause of the pauses are many. Texture loading is defintely one of them and this post does address in that ballpark.

Your settings on visibility are interesting. Medium. I run extreme. This is for one reason I cannot avoid. Object pop in, especially city night lights is tied to view distance. You cannot see from one side of a city to another at night with Medium view distance, despite the large savings. There are good savings in view distance cutting coupoled with preload radius. You also state medium textures but show High in the screenshot. Confusing. I believe the terrain texture doesnt impact or change much in testing, but the cockpit texture does impact performance, especially when using high end stuff like VR. These factors are important to some types of stutter which affect quickly changing your view direction in VR. You can reproduce that by switching on a TGP, it costs....'something'.

Overall its a worthy post but lacks decent empirical data to make it convincing and thus falls into the many tweak-guesses that VR users chase. By decent empirical data i mean before and after frame counts and stutter counts on a cited test mission. The OP is also devoid of the links that provide the backbone of the prerequisities so these all become uncertain. Also the screenshot disagrees with the words and doesnt explain all of the settings. A good post wont have follow up posts saying, "I forgot to mention ..." Also combine the third post with the second into the first, people don't read well! Please edit the original and keep it updated with any new information, this is complex stuff and we need good contributors to the VR performance scene in DCS!

Edited January 4, 2021 by Pikey

[mtd2811]

I didn't 'notice' any positive gains following this, but its' fairly hard to tell because to introduce one of the many different types of stutter you have to introduce complexity to the mission which affects your overall performance - so very difficult if not impossible to say. The PD was three tenths larger, the SS much lower (this is contrary to the low PD, high SSS ratio that has generally been found to give back many frames.) I didnt understand the saving on AA which doesnt cost according to extensive testing. I also added Kegetys Shader mod at the same time, so again, very difficult to tell. I also didnt follow all of the prerequisities which were many, they required searching and I'd prefer not to guess at them.
 
90 FPS in the menu is normal for any settings of course. I can get 90 in single player in cockpit up high in many cases, where this stops is on the ground and in multiplayer and with mission complexity. 1400 unit server with scripting and lots of AI - 35/50 is really good for my 2080S and i5 8600.
 
The parts I think where this is interesting are the NV shader cache and low latency mode, which I believe is the point of this post. Maybe its doing something, maybe its intereacting with the slower IO of a disk, I dont know. Id like to say it did something but removing the windows virtual monitor for VR did way more in effect, so rather than say, 'it feels good', I will just say 'it doesn't feel bad'. The word "stutter" is overused and misused by many people, which unfortunately doesnt help the community progress and spreads misinformation. So I cannot say, I do or don't get stutters following this advice. I certainly get some pauses which last visibly a very long time inside the second, but I think the cause of the pauses are many. Texture loading is defintely one of them and this post does address in that ballpark.

Your settings on visibility are interesting. Medium. I run extreme. This is for one reason I cannot avoid. Object pop in, especially city night lights is tied to view distance. You cannot see from one side of a city to another at night with Medium view distance, despite the large savings. There are good savings in view distance cutting coupoled with preload radius. You also state medium textures but show High in the screenshot. Confusing. I believe the terrain texture doesnt impact or change much in testing, but the cockpit texture does impact performance, especially when using high end stuff like VR. These factors are important to some types of stutter which affect quickly changing your view direction in VR. You can reproduce that by switching on a TGP, it costs....'something'.

Overall its a worthy post but lacks decent empirical data to make it convincing and thus falls into the many tweak-guesses that VR users chase. By decent empirical data i mean before and after frame counts and stutter counts on a cited test mission. The OP is also devoid of the links that provide the backbone of the prerequisities so these all become uncertain. Also the screenshot disagrees with the words and doesnt explain all of the settings. A good post wont have follow up posts saying, "I forgot to mention ..." Also combine the third post with the second into the first, people don't read well! Please edit the original and keep it updated with any new information, this is complex stuff and we need good contributors to the VR performance scene in DCS!

My intention was only to point out what uas worked for me based on my setup - in case it helps other people.

Apologies if my post doesnt look “proffessional” with “empirical data” as you have noted but I so t have the time in all honesty to write something more extensive.

I just pointed out key items which I changed and made a massive difference to my setup and maybe can help someone as they can be easily implemented.

At some point tonight I will edit the main post to see if I can add anything new from my discussions with glide afterwards


Sent from my iPhone using Tapatalk

[Pikey]

Dont take it personally, you might have many people read your post searching for VR improvements. If it worked for you, you should probably say 'what' worked. What is a "massive difference", for example? One less stutter per minute? Is it quantifiable? I'm feeding back because the time people put into VR tweaks is... let's say a non empirical word here.... "too much" There is also a lot of overlapping setting suggestions which causes duplication of reading, writing and loss of time. A few good posts would be better. Additionally i am compiling this data in case that can actually be done, but there lacks any empirical data to say what is worth doing, sadly

To me, I consider it better for the community to have more information in fewer places, rather than many posts with little tweaks here and there. Don't worry the post is helpful, it oculd be better. I tried to google "CPU priority registry setting for DCS,VRServer, VR Compositor site:eagle.ru" and the only hits were this post. So this is a prerequisite but its not referenced? That sort of thing woudl help to go from good to awesome! Thanks.

[mtd2811]

How can I not take it personally since you imply how “poor” this post is without extensive testing and empirical data. And why would I need to do extensive testing between changes and record them for posting. I am trying for days now to improve my set up and came across key items which work for me.

Right now my setup works perfectly in MP with my g2 and I am not willing to change any core settings just to provide before and after as it may mess something up.

I think my post says “no stutters” after doing these changes and clearly I have highlighted the Shaders cache and pagefile which have made a massive difference to my stutters. If people need to read follow up replies to see how the chat developed well thats the point of a post. Glide was decent enough to engage in constructive discussions to improve this furter. Some work some dont (for my setup)

For the cpu priority don’t search for a specific service like vrserver but instead maybe “how to include a service in regedit for cpu priority permanently “ . Thats what I did to find how it is added and particularly this

https://forums.battlefield.com/en-us/discussion/48138/performace-boost-how-to-set-process-priority-permanently

Hope this helps


Sent from my iPhone using Tapatalk

[Glide]

I believe we are trying to help our fellow flyers have a better experience, the same great experience we are having.  As for empirical data, I am sure ED has people who get paid to do that stuff.  I'm happy with my settings now, and I've shared them, hopefully for the betterment of everyone.  I find DCS is more fun without the FPS counter on.  Thanks for the feedback.  Cheers and fly safe.

[mtd2811]

I believe we are trying to help our fellow flyers have a better experience, the same great experience we are having.  As for empirical data, I am sure ED has people who get paid to do that stuff.  I'm happy with my settings now, and I've shared them, hopefully for the betterment of everyone.  I find DCS is more fun without the FPS counter on.  Thanks for the feedback.  Cheers and fly safe.

Couldn’t agree more.

At the end of the day, like so many others we are also undertaking “trial and error” and post both positives and negatives which hopefully may help some people


Sent from my iPhone using Tapatalk

[Pikey]

7 hours ago, mtd2811 said:

How can I not take it personally since you imply how “poor” this post is without extensive testing and empirical data. And why would I need to do extensive testing between changes and record them for posting. I am trying for days now to improve my set up and came across key items which work for me.

Right now my setup works perfectly in MP with my g2 and I am not willing to change any core settings just to provide before and after as it may mess something up.

I think my post says “no stutters” after doing these changes and clearly I have highlighted the Shaders cache and pagefile which have made a massive difference to my stutters. If people need to read follow up replies to see how the chat developed well thats the point of a post. Glide was decent enough to engage in constructive discussions to improve this furter. Some work some dont (for my setup)

For the cpu priority don’t search for a specific service like vrserver but instead maybe “how to include a service in regedit for cpu priority permanently “ . Thats what I did to find how it is added and particularly this

https://forums.battlefield.com/en-us/discussion/48138/performace-boost-how-to-set-process-priority-permanently

Hope this helps


Sent from my iPhone using Tapatalk

Don't read things into something that was never said. Read what I said, to quote myself, "This was a worthy post".
You specifically asked for feedback in the title. If you don't want feedback then don't ask for it!
I'm letting you know, I didn't see 'no stutters' myself. I also explained that stutters have multiple sources and configuration alone is unllikely to fix some sources (mission).

Have a good day!

[Pikey]

4 hours ago, glide said:

I believe we are trying to help our fellow flyers have a better experience, the same great experience we are having.  As for empirical data, I am sure ED has people who get paid to do that stuff.  I'm happy with my settings now, and I've shared them, hopefully for the betterment of everyone.  I find DCS is more fun without the FPS counter on.  Thanks for the feedback.  Cheers and fly safe.

 

Sorry to say, ED do not provide Empirical feedback, this is why people do this themselves. ED do read, and often they learn from the feedback, but they always ask us for tracks and empirical data because often the vague feedback doesnt help them refine the product. So, passing that on, an opportunity was missed to see what frame times, stutters, FPS changed with what settings. Thats all.