Reverb high CPU and GPU frame time - UNPLAYABLE

I would just set it to all the physical cores you have, and then let windows do it’s thing... from my own testing it needed at least 3 cores if you look here you will find out why, I can’t remember exactly the reason.

Hi mate, thanks for the info!

 

Out of curiosity, what is the third core required for by DCS? I understand by previous posts that one core is for the game and the other for the sound.

 

Also, I did a bit of googling and couldn't find exactly which cores are physical and which are logical for my specific CPU. If it is safe to assume that generally all multi-core CPUs structured from core 0->n where the first core is physical and the next is logical, would you recommend DCS to be set to cores 3, 5 and 7 in order to utilise three physical cores?

If you're going to use CPU affinity then don't select odd numbered cores for the game, as these are the HT cores and only offer around 35% of the performance of the actual physical core (even numbered).

I did try a similar approach myself for a while using Process Lasso, but found little to no improvement, so now just let Windows manage it, if you're still on Windows 7 it may offer an improvement.

Thanks for the advice guys.

I used Project Lasso to set the core affinity to 1, 3, 5, and 7 and found no noticeable difference between affinity set to all cores. Out of curiosity I then changed it to 0,2,4,6 and found the same...

Should there have been at least some difference one way or the other??

Thanks for the advice guys.

 

I used Project Lasso to set the core affinity to 1, 3, 5, and 7 and found no noticeable difference between affinity set to all cores. Out of curiosity I then changed it to 0,2,4,6 and found the same...

 

Should there have been at least some difference one way or the other??

Is your processor showing as 8 or 16 cores? I'm surprised that running on the odd numbered cores showed no difference to the even numbered cores!

Your results sort of confirm my own, in that Process Lasso didn't seem to make a noticeable difference to in-game performance.

From further reading on Intel cores it looks like each physical core is split into two logical cores, so that two threads can be run at once.

This suggests that running on 0, 2, 4, 6 will yield the same result as 1, 3, 5, 7, in that you are still running on four separate physical cores.

Copy from my post on fb:

"hi, i've proved different combinations to get the way to understand the variables of vr performance. mostly used fpsvr to check cpu and gpu frametime and the most perfirmance killer was the shadows. even in only flat, cpu goes down (with everything in low to minimize gpu usage). cpu graphic turn to red with >30 and fps drop to 27. gpu varies with other settings, but cpu show low usage but too slow. it does no change if mhz is 3600 or 4200. so, with shadows off, everything goes like charm, with both fpsvr graphics in yellow/green. is there any way to reduce shadows draw distamce to test?

specs: ryzen 3600, 16gb 3200, gtx 1070, rift s"

The only way i know of to reduce shadow draw distance is to reduce tree draw distance . Then you get pop-up trees so......

Thanks jalfrezi, that seems to be consistent with the results I got... Does this mean that this setup is as much I can get out of my system?

that seems to be consistent with the results I got... Does this mean that this setup is as much I can get out of my system?

Nope. There has to be something wrong, if you did lower all settings and it still dont get decent fps and cpu frame times.

Here is two videos, I shot these for another forum member that has stutter problems.

First I made videos with my normal settings( at 90 fps and sometimes 45 reprojection).

Those do nit come up smooth at youtube due to that filming 90fps with 60 will cause unsmooth playback.

In these, I did set the reverb to 60fps, smooth gaming mostly, It may be some reprojection causing 30 fps reprojected wich I find to unsmooth for enjoyable gaming.

My normal settings (90fps and reprojection when needed) feels smoother.

Shot with gopro trough one of the lenses in the Reverb:

Shot with nvidias tool (captures the desktop display):

We have similar hardware an got similar points on the VRMark.

We should get similar frames per second if we have the same setup on steamVR and in DCS etc if everything work as it should.

I didnt update to 2.5.6 BTW, because of all the problems. As most already sad, roll back to 2.5.5.

I have not changed any settings for processor cores( stock DCS setting ).

I use hyperthreading, as DCS isnt the only software I run. Virtually most other benefit from HT.

I havent messed with project lasso.

I havent messed with shaders mod.

All of these (except not disable HT) isnt done because I get performance good enough to enjoy gaming and I am happy as is.

Thanks @Gunnars Driver!

I reverted to 2.5.5 a few posts ago, I believe it has made significant improvement to my FPS and frametime in doing just that alone.

For info, this is what I have done so far:

Implement shdwp’s recommendations:

•Reduce visibility range to Medium - (post #2)

•Set shadows to Mid - (post #2)

•Set Terrain object shadows - (post #2)

Implement stuartaston’s recommendations:

•Copy stuartson’s personal settings - (post #4)

•Disable hyper threading in BIOS - (post #4)

•Revert to 2.5.5 - (post #4)

•Set processor affinity to physical cores (6 and 7 as per SkateZilla’s post #1 using the URL stuartaston provided in abovementioned post) - (post #4)

Implement Gunnars Driver’s recommendations:

•Reduce both global and DCS specific VRSS to 100% (post #9)

•Run VRmark (using URL Gunnars Driver’s provided) – (post #17)

•Reverted CPU affinity to all cores

Still to do:

Implement Gunnars Driver’s recommendations:

•Benchmark CPU/GPU/RAM – looking for a good free benchmarking tool (post #9)

•Reduce settings until reaching a playable game (post #9)

•Max out pre-load radius? (post #9)

•Reducing the tree visibility, clutter grass to a quite low value, terrain textures to low, lower visibility range to lower than ultra, shadows to ”flat”, smoke and chimney to lowest and anisotropic filtering to off or 2x. – (post #17)

A few more questions:

1. 
   Would you recommend playing with affinity mask via the game's lua file seeing as Project Lasso had no measurable impact?
    
   
2. Are you suggesting that I should be reaching something close to 90fps during gameplay?
    
   
3. How do you enable reprojection - is it automatic? I haven't been able to enable it using Thud's VR4DCS.
   

From your screenshots on the first page it definitely looks like it's your CPU that is the bottleneck, so I would look to drop Visibility to Medium/High and lower the Clutter/Grass setting <-- when I was carrying out testing I found this setting alone has a big hit on FPS even on the Tarawa - for no appreciable benefit other than some fake looking grass. I'd also ditch Process Lasso and re-enable HT, Windows 10 does a pretty good job of managing threads.

You do seem to have a lot of things running in your screenshots - Chrome, Outlook, Office apps, etc - I'd close as much as you can before playing DCS, particularly web browsers as they're massive resource hogs, particularly RAM.

In Nvidia control panel try turning on MFAA if you are using any anti-aliasing settings.

EDIT: only change one or two settings as a time, then test DCS and see if it makes things better or worse, then move on to the next settings. Also, try and find a consistent DCS test scenario like a ramp start that has a lot of moving vehicles.

A few more questions:

1. 
   Would you recommend playing with affinity mask via the game's lua file seeing as Project Lasso had no measurable impact?
    
   
2. Are you suggesting that I should be reaching something close to 90fps during gameplay?
    
   
3. How do you enable reprojection - is it automatic? I haven't been able to enable it using Thud's VR4DCS.
   

I havent used affinity mask/shaders mod as I get fine fps and decent picture( readable instruments and high or ultra in the distance setting).

I did consider it but a friend tried it (using Oculus Rift) and he had no or almost no gain. I was told that you have to re-do the shader mod after each update.

No/low gain and ‘lot of work’ made it a nogo for me.

I also wouldnt like to risk to trash a very good working setup.

I think DCS should get the shader mod implemented by Eagle Dynamics, it seems like an easy way to lower the demands for VR.

I can lower settings to get 90fps for more % of the time, but as reprojection work as good as it does, I mainly use the settings to ”never” go below 45fps = means I have smooth gaming all the time.

I think you should lower the setting just to see that the CPU isnt loaded too much for comparable settings/comparable hardware( ie processor times below 11ms possible). This to find out if it is only too high or wrong settings or if there is a major c**k up in your DCS installation or somewhere else.

When fixed or when knowing there is no problem, you should work with the settings to find your personal setting level, ie the compromise between high fps and picture quality you want. You can not get constant 90fps with high settings - it is not possible.

For me, the goal is to always stay at or above 45fps(never see it drop below). This ensures smooth gaming.

Reprojection- I use the automatic setting, this give me 90fps when the computer has the power for this and 45 fps when it dont. I dont really see any other numbers(except for the brief transition between 90 and 45). The fixed 90 or 45 is a receipt for the automatic switch.

I could help via discord or skype to try make the correct settings and make sure reprojection work as it should.

Arriving at home tonight and I will be off work for one week so I think I have the time.

Thanks guys!

I will reduce everything to pretty much barebones as suggested to see if I can get the 90/45 FPS and sub-11ms frametime. I will kill as many resource hogging tasks as possible make the changes a couple at a time. I will also re-enable hyper threading as suggested.

@Gunnars Driver

I agree with the shaders mod, it's a lot of work for minimal gain which needs to be redone with each update.

I had a bit of a run with the game last night (I spent all this time lowering settings and chasing FPS and frametiems without actually checking what the game looks like when I put on the goggles) and the results were pretty disappointing, I have to say...

There was a lot of flickering in the picture, text looked like it was tearing as I was looking at it and moving my head even slowly and I noticed that if I tilt my head (not turn but tilt to either side) the clouds would deform/change shape??

Thank you for offering your time to help via Discord or Skype, that's amazing! I am free this weekend and if you'd be happy to, we can arrange a time to discuss?

Thanks again!

oh if skype is installed and running in the background that can impact performance...

Hi stuartaston, I definitely haven't installed Skype on this recent installation of Windows 10 (certainly not manually anyways). Looking through my Task Manager I did not see it running as a process either.

cool

The guys i play with did use skype earlier and I did never notice a problem. We changed to Discord a couple of years ago so for the last years I dont know.

Good tip though: All possible issues should be looked into, to be able to find the issue and get it running.

I did make some screenshots for reference.

I did lower the CPU Clock to 5.1, just saw that you dukman seems to run at 5.0Ghz, not much of a difference so I didnt retake them.

Settings: SteamVR at 100%, rendering resolution 2204*2160:

[/img]

Framte rates and processor loads:

F18 Ready on the ramp caukasus:

This is just after boot, flying give me GPU frame time around 9ms and CPU around 8-9.

After flyging for a few minutes the average fps is closing towards 90fps.

[/img]

Not fun at all to fly like this, and it is not the meaning either. It it to check wheter it is possible to get the CPU load and CPU frame times down. If not, there is something we need to fix.

After check, and possible problem fixed, the settings should be a lot higher :-)

I disagree with that. 150% is only equivalent to a PD of 1.2. Not that much really, especially with his hardware.

 

I run with 150% and have no issues with 45 fps rock solid on 2.5.5 on pretty high settings.

I have the same CPU, GPU and same amount of ram. Also dual nVME m2’s.

After the setting test shown above I reverted to my normal settings( including SS 120%)

I had smooth gaming and never below 45fps.

When rising it to 150% I get a lot of stutter and it isnt a smooth experience any more at least at low level flight.

I notice stutter even when fpsVR and/or dcs shows 45 fps. This in the same F-18 mission ready on the ramp. Multiplayer etc is harder so for me 150% is a clear nogo.

My 9900KS and RTX2080ti 11Gb (ASUS ROG STRIX OC) isnt up to deliver enough power at SS150%.

I have the same CPU, GPU and same amount of ram. Also dual nVME m2’s.

 

After the setting test shown above I reverted to my normal settings( including SS 120%)

I had smooth gaming and never below 45fps.

 

When rising it to 150% I get a lot of stutter and it isnt a smooth experience any more at least at low level flight.

I notice stutter even when fpsVR and/or dcs shows 45 fps. This in the same F-18 mission ready on the ramp. Multiplayer etc is harder so for me 150% is a clear nogo.

 

My 9900KS and RTX2080ti 11Gb (ASUS ROG STRIX OC) isnt up to deliver enough power at SS150%.

It may a no go for you, but not for me and many others. No stuttering whatsoever here.

It may a no go for you, but not for me and many others. No stuttering whatsoever here.

It is funny - or not so funny depending on the individual's results - how there will be very different results in performance with the same headset and similar rigs.

For instance I have seen some with the Rift S, who after getting 2.5.6 experience much worse performance, yet here I am and I saw no degradation whatsoever in performance with my Rift S, in fact it almost seemed like mine went up a little with 2.5.6.

I guess with PC's it is anyone's guess as to the why's.

It is funny - or not so funny depending on the individual's results - how there will be very different results in performance with the same headset and similar rigs.

 

For instance I have seen some with the Rift S, who after getting 2.5.6 experience much worse performance, yet here I am and I saw no degradation whatsoever in performance with my Rift S, in fact it almost seemed like mine went up a little with 2.5.6.

 

I guess with PC's it is anyone's guess as to the why's.

Yep, it is strange. I’m dismissing 2.5.6 performance as ED are fully aware of how appalling it is - at times - for many users and are working hard to fix it. Fingers crossed for next week!

It may a no go for you, but not for me and many others. No stuttering whatsoever here.

Yes, it might be either a difference in computer performance or a matter of personal preferences*.

—> This makes my case; for the matter of this thread 150% is/can be a quite high setting and therefore we should not tell the OP that its a low setting that cant give you problems.

I have learned from 20 years of Home Cinema hobby that there is a big difference in how people percept the stutter of movie not running in perfect frame rate. I am above average sensitive to non smooth gaming, or non smooth movies. Probably a result of 20 years of home theater PC tuning :doh:

Anyway, the settings also matters. I wouldnt like to lower the settings and try to make a low settings picture sharper with SS.

Yes, it might be either a difference in computer performance or a matter of personal preferences*.

—> This makes my case; for the matter of this thread 150% is/can be a quite high setting and therefore we should not tell the OP that its a low setting that cant give you problems.

 

I have learned from 20 years of Home Cinema hobby that there is a big difference in how people percept the stutter of movie not running in perfect frame rate. I am above average sensitive to non smooth gaming, or non smooth movies. Probably a result of 20 years of home theater PC tuning :doh:

 

Anyway, the settings also matters. I wouldnt like to lower the settings and try to make a low settings picture sharper with SS.

The OP has got other issues going on that no one can figure out. A 1.2 PD equivalent setting for SteamVR SS, i.e. 150%, is not a high setting by any stretch.