Why is VR FPS so poor in comparison to monitor?

[toutenglisse]

You meant 4 cores and a global cpu usage at 25% can be 3 cores at 0% and 1 core at 100%.

And so/or, 25% usage on each 4 cores - due to windows bouncing a thread - can be (or is) in fact 1 core at 100%.

 

Still… I'm not sure… with a 1080P screen and DCS 2.5.0 I had monitored near 100% usage on Gpu and also at same time near 100% on each 4 cores during 40sec long successive periods - how can it be possible if DCS can't go over 25 or 35% global cpu usage.

(with actual OB in VR the summ of all cores is about 130% of 1 core - but can shortly go to near 400% / 100% global cpu when GPU is topping 100% for 10sec or more)

[Jabbers_]

You meant 4 cores and a global cpu usage at 25% can be 3 cores at 0% and 1 core at 100%.

And so/or, 25% usage on each 4 cores - due to windows bouncing a thread - can be (or is) in fact 1 core at 100%.

 

Still… I'm not sure… with a 1080P screen and DCS 2.5.0 I had monitored near 100% usage on Gpu and also at same time near 100% on each 4 cores during 40sec long successive periods - how can it be possible if DCS can't go over 25 or 35% global cpu usage.

(with actual OB in VR the summ of all cores is about 130% of 1 core - but can shortly go to near 400% / 100% global cpu when GPU is topping 100% for 10sec or more)

 

Yes, 1 core 100% or 4 cores ~25% (on dcs only)

 

100% on all cores could be other apps, i doubt it was 100% all just dcs. If you are using the windows CPU performance page to gather these stats it can be deceiving, but without showing exactly what you are talking about its hard to say.

 

If your GPU is hitting 100% chances are you are GPU bound, but you also want this over CPU typically. Most games are GPU bound. In VR its even easier to be GPU bound if you up your pixel density.

[Zoomer]

On the second mission of the red flag campaign, while stuck on the airfield for startup etc. with all the traffic and aircraft about, the fps in VR tanks to the mid 20`s/30`s. All while my 2080 gpu is running at ~35% usuage/load. My CPU is a i7 4790K OCed to 4.6ghz and it`s just bottlenecking due to the game engine`s single core limitation and poor optimization for use with modern multicore CPU`s. Regardless of what monster pc anyone builds, until the engine is overhauled then it`s like eating soup with a fork. Unfortunately most sim games are in the same boat, bring on Vulcan, VRS & DLSS to let the hardware do it`s job.

Obviously the AI in heavy populated examples eats into the performance of the hamstrung CPU which in turn limits it`s bandwidth to service the GPU leading to it`s complete underutilization.

Edited November 29, 2018 by Zoomer

[toutenglisse]

Jabbers here a msi monitoring of DCS VR (actual - screenshot from 1-2 days) when GPU is topping 100% during several seconds and CPU tends to follow/assist ...

 

[ATTACH]198971[/ATTACH]

 

 

And here cpu monitoring of DCS 2.5.0 with 1080p screen and unlimited fps (GPU don't appear but it was 100% when cpu was 100%...often) :

 

https://forums.eagle.ru/showpost.php?p=3578743&postcount=19

Edited November 29, 2018 by toutenglisse

[Jabbers_]

is it just looking at time from DCS or is it looking at CPU time from just your PC? I've never used it. Also was all of this during loading of a mission, or actual gameplay, or both? So many variables. The best way to test would be to replay a track and only bench during the track replay.

[toutenglisse]

This is only gameplay for the 2 monitoring pictures.

I Don't understand your first phrase.

EDIT : maybe I do ... you mean when I say 10-15 seconds ? I say that because 1 little square is 10 seconds in afterburner)

[Rymy]

So as latest news tells about Viggen maybe we soon finally start to see some use for other cores also if developers move to trend that some of the weapon/aircraft system simulations calculations are done of other different cores. Interesting ofcourse will be how they manage it to keep it fluent and not cause any issues for netcode/multiplayer environment.

[lemoen]

So as latest news tells about Viggen maybe we soon finally start to see some use for other cores also if developers move to trend that some of the weapon/aircraft system simulations calculations are done of other different cores. Interesting ofcourse will be how they manage it to keep it fluent and not cause any issues for netcode/multiplayer environment.

 

This won't really make the overall framerate better. It just stops it from getting worse. The game AI, path finding, rendering etc is all still done on one core only (and sound on another).

[Harlikwin]

So possibly a dumb question....

 

IF the main drag factor for VR is using the CPU to do 2 scenes, why not mulitcore/parallel that, it seems like that should be something parallel multicore procs should be able to do really easily/well (I was once a programmer long ago).

 

Also, is there a good way to log CPU/GPU use during play, and graph it later? I'm curious what the bottleneck is on my rather lower end system.

Edited November 30, 2018 by Harlikwin

[imacken]

So possibly a dumb question....

 

IF the main drag factor for VR is using the CPU to do 2 scenes, why not mulitcore/parallel that, it seems like that should be something parallel multicore procs should be able to do really easily/well (I was once a programmer long ago).

 

Also, is there a good way to log CPU/GPU use during play, and graph it later? I'm curious what the bottleneck is on my rather lower end system.

 

DCS using single core has been discussed many times before. Let’s hope something changes soon.

[Georgio]

My understanding of the DCS engine is that it’s old architecture doesn’t allow multiple cores except for sound.

 

To incorporate true multi-core optimisation would require a complete rewrite of the code which simply isn’t going to happen.

Maybe at some time in the future some enterprising publisher will stump up the money to fully optimise a flight sim for VR but probably not until VR technology matures.

 

Remember, this isn’t the 90’s, flight sims are a very select slice of the market with sales in the thousands at best not the millions that consoles enjoy, so consequently resources are extremely limited.

[hansangb]

Also, there's a latency dependency. The frames have to be drawn within a small tolerance (I want to say 20ms). But as the VR shader mod proves, there's a lot of room for VR optimization. And then there's the vulkan migration that I can't wait to see. So things are looking up. When DK1 was around, ED adopted it. and I'm glad they did!

[etherbattx]

Why is VR FPS so poor in comparison to monitor?

 

it’s not just DCS, most games can’t use multiple cores very well.

 

it’s actually really difficult to do.

the non progressional software engineers think it’s easy, but it is not.

 

it’s one of the reasons i’m not in favor of buying 6,8 and 12 core machines. they tend to be both expensive and utterly wasted.

Edited December 2, 2018 by etherbattx

[VampireNZ]

My understanding of the DCS engine is that it’s old architecture doesn’t allow multiple cores except for sound.

 

To incorporate true multi-core optimisation would require a complete rewrite of the code which simply isn’t going to happen.

Maybe at some time in the future some enterprising publisher will stump up the money to fully optimise a flight sim for VR but probably not until VR technology matures.

 

Remember, this isn’t the 90’s, flight sims are a very select slice of the market with sales in the thousands at best not the millions that consoles enjoy, so consequently resources are extremely limited.

 

 

Maybe some time in the future some CPU manufacturer will realise a CPU with a bazillion cores is a PITA for games, and actually designs a CPU with a few bloody fast cores for a change! Quad-core 20Ghz+ would do nicely for a start. I mean...it's not like it's friken nearly 2020 or anything! I thought we were supposed to be zipping around in jetsons cars like 18 years ago?

[lemoen]

The reason most games don't use multiple cores is because some things just cannot be calculated in parallel. So parts of DCS could be done in parallel (AI routefinding maybe, general AI decision making) but other things probably cannot (flight model calculations). Either way, you can not get DCS to run 400% faster on 4 cores even if you had a magic wand that could magically parallelize all the code that could be run in parallel. See https://en.wikipedia.org/wiki/Amdahl%27s_law

 

 

Basically, for most games, the speedup is not enough to warrant the extra complexity of writing and maintaining parallel code. DCS I think can use it in places but maybe the same applies here (not enough speedup for the effort).

[Rymy]

Actually it is not so difficult to put and assign different calculations to different cores if all those calculations would just stay in your computer. Problem is that when we are talking of multiplayer games which communicate via network to server sending data of your actions and also received data all what is happening around you is where problem comes.

 

You need to make sure that all calculations done by different cores are in sync and delay between results stay withing acceptable limit that it does not cause strange effects or what someone could describe as warping or lag. It doesnt help also that there are machines with lot of diffent harwares and amount of cores with different memory and speed amounts.

 

There has been a lot of discussions about this in iRacing forums where one of the developer has explained above problems.