VR: Can't get 90fps with low settings, but am getting same fps at ultra settings?

[jamesp111]

Hi,

 

I can't work out why I can't get 90fps at low settings, when I am getting the same (maybe a touch less) fps at ultra settings? I thought I had got higher fps before, but have had an enforced break due to waiting for a replacement Rift.

This is with 5 AI planes in a quick mission by the way.

 

I've checked no other process like antivirus etc is consuming CPU. Also checked on Stalingrad map to confirm it's not something to do with Kuban.

 

My PC is an i5-2500k @ 4.2Ghz, 16GB ram and a GTX 1080. I'm using an Oculus Rift VR HMD.

 

I've tried two different GPU drivers (the latest 388.13 being much worse so I reinstalled 385.69). Is there a graphics config file I can delete to ensure it's definitely reset and I haven't turned up supersampling in the past or something?

 

Besides that, does anyone have any ideas?

 

Thanks for the help!

[Jade_Monkey]

The bottleneck is in the CPU, that's why the graphics settings are not making a difference.

[Mitthrawnuruodo]

I hate to say this, but I agree that the major problem is likely the CPU. The 2500K@4.2 has ~20% lower single-thread performance than recent Intel CPU models. 

[SCG_Fenris_Wolf]

Both Jade Monkey and mitthrawnnuruodo are right.

 

I'd prefer bending their statement a bit however - the issue is not your CPU being slow, but IL 2s threading management (unlikely to change) and its VR implementation (possible to change, e.g. single pass stereo via dx11 API anyone?).

 

I'd so like an enlightening response by Mister Williams to this one. Afaik the gfx card folks provide documentation and support to this. It would be fantastic to at least get a somewhat thorough response about it... :S

[dburne]

You might try doing a clean driver install, removing your drivers with DDU ( Display Driver Uninstaller ).

Install only the Driver and PhysX, none of that other stuff like GeForce Experience etc.

 

After clean driver install, be sure and set in Nvidia Control Panel for BoS to " prefer maximum performance". I believe the drivers set that at optimal by default, you don't want that. Leave all other settings at " application controlled".

 

See if that helps some. Also be sure and keep the HUD off in the game, it is a real performance killer.

[Herne]

the ASW seems to do a very good job of having smooth head translation at 45 FPS though. Of course 90 FPS would be better, but in spite of the gun sight flicker, its still pretty damn good.

[jamesp111]

the ASW seems to do a very good job of having smooth head translation at 45 FPS though. Of course 90 FPS would be better, but in spite of the gun sight flicker, its still pretty damn good.

Yep it's mostly the distortion of the gunsight, and targets behind the propellor that are the issue.

 

Besides that though i actually find this game the worst affected by ASW: Due to the fast rotation many of the aircraft offer, ASW can't keep up.

[marklar]

 

 

the ASW seems to do a very good job of having smooth head translation at 45 FPS though.

 

 

ASW is not a solution for cockpit games. I tried it in Assetto Corsa, War Thunder, DCS and BoX. It never feels smooth enough and causes a lot of graphical glitches.

 

OP> When I upgraded my i5-2500k @ 4.2Ghz to i7-6700K my fps in BoS jumped from 40 something to 80+. It was before dx11 was implemented and on a single 1080p screen.Now with dx11, CPU is less important than GPU but still it's what limits your fps in VR. I have 980ti which is less powerful than gtx1080 and I have usually 90fps in VR on Balanced settings. 

[chiliwili69]

If you take a look of this post:

 

https://forum.il2sturmovik.com/topic/29322-measuring-rig-performance-common-baseline/?p=478308

 

you will see that RR_Kraft was using am i5-2500K at 4.4Ghz and achieving 49fps avg in the old Balapan VR test track, which used High settings and SS=1.

 

The most important thing for BOS in VR is the Passmark Single Thread number that your CPU can achieve. In the case of the i5-2500K you have a public STmark of 1898, which is quite low for BOS in VR even for Low or Balanced.

 

The RAM speed is also important, What is your RAM speed?

 

Try to run the VR performance test with LOW/BALANCED/HIGH and report your average fps. The procedure is described here:

https://forum.il2sturmovik.com/topic/29322-measuring-rig-performance-common-baseline/

[Kraft]

I just ran the test. My results:

Frames: 2913 - Time: 60000ms - Avg: 48.550 - Min: 40 - Max: 83 

 

and with "Low" preset:

Frames: 4907 - Time: 60000ms - Avg: 81.783 - Min: 46 - Max: 91

SS in Steam is 1.7

I do not know what parameters are changing with the help of presets.

Edited November 6, 2017 by RR_Kraft

[chiliwili69]

Thanks for running the VR test. 

I am just missing what is your Passmark CPU and you Passmark Single Thread numbers as explained in the test procedure.

 

RAM at 1600MHz is a bit low, if you can upgrade to 2400 or 2600 you will see some improvement (not spectacular but nice to have)

 

Your test should give some guidance to the original poster since you are able to run at 81 fps at Low. So, if he run the test he should achieve similar results.

[TUS_Samuel]

The bottleneck is in the CPU, that's why the graphics settings are not making a difference.

 

For IL-2 in cases where only CPU is the bottleneck, graphics settings should make HUGE difference https://docs.google.com/spreadsheets/d/1gJmnz_nVxI6_dG_UYNCCpZVK2-f8NBy-y1gia77Hu_k/edit#gid=334824301&range=U20.

OP has something else limiting the performance.

[doog442]

Can anyone explain the difference between Low / Balanced / High settings ?

 

I get a super smooth experience on low but for the life of me cant see any difference in image quality between the settings other than a drop in FPS.- its still pretty good just higher FPS and very smooth.

 

I'm using VR SS at 1.9 , no sharpen, landscape detail x2, normal grass, low shadows.

 

i5 4690k @4.3 , 1070, 8gb of ram

[[SRH]Festa_VR_Noob]

If you look in the hanger on the different setting, you can see the grid on the floor to your left is lower poly.

 

I would imagine that rings true throughout the game, with vehicle/building/possible terrain poly`s being lower to improve performance.

 

I did some benchmarking (very unscientific), and found on low, i had FPS of min/max 43/70, balanced was 41/51 and high was 30/46, so while the lower end isn`t a huge difference, you can get some higher frame on lower settings.

 

You probably wont notice it, unless you do some super close formation flying, as you`re in the open air.

[polnoch_ksu]

Is it possible to get IL2 VR-ready configuration in the PC based in:

 

Xeon X5680 (3.3Ghz, 6 cores), slow ram: 1333Mhz, SSD?

 

I want to buy new Videocard for this PC, especially for the VR. And want to understand:

 

1) Is it possible to play in VR/IL2 using this CPU/RAM?

2) can I achieve good look? Maybe I can disable any cpu-related graphical features? Or use features which can be calculated in multi-thread mode?

3) which card is maximum for this PC(and in playing IL2/VR)? Probably, I'll not get any differences between GTX1070 and GTX1080, am I right?

4) can we expect performance optimization in the future (and I think reduce single thread performance requirements is a good idea, because there are lot of ryzen CPUs in the market which support a lot of computing threads) which can help to achive better perfomance in my CPU, load all cores? Should I buy GTX1070, or good idea to buy GTX 1080, which can be fully loaded by IL2 in my PC in the future?

[chiliwili69]

1) Is it possible to play in VR/IL2 using this CPU/RAM? 2) can I achieve good look? Maybe I can disable any cpu-related graphical features? Or use features which can be calculated in multi-thread mode? 3) which card is maximum for this PC(and in playing IL2/VR)? Probably, I'll not get any differences between GTX1070 and GTX1080, am I right? 4) can we expect performance optimization in the future (and I think reduce single thread performance requirements is a good idea, because there are lot of ryzen CPUs in the market which support a lot of computing threads) which can help to achive better perfomance in my CPU, load all cores? Should I buy GTX1070, or good idea to buy GTX 1080, which can be fully loaded by IL2 in my PC in the future?

 

1) This CPU has a SingleThread Passmark index around 1500 (https://www.cpubenchmark.net/cpu.php?cpu=Intel+Xeon+X5680+%40+3.33GHz&id=1312). This is very very low to play BOS in VR even with Low or Balanced you will not get more than 45 fps. People with Ryzen 1600 (with Singlethread passmark at 2000) and 2400MHz RAM was having 43-46 fps with Balanced .

So, no, this CPU is not good for VR/IL2 with Rift/Vive

 

2) Switching off all graphics details and using Low you will get very low fps (most likely no more than 50). IL-2 has several threads but there is one thread which is the main one and heavy . IL-2 doesn´t benefit too much from multi-core CPUs.

 

3) Since the CPU/RAM will be the bottleneck it doesn´t matter what GPU you use

 

4) Devs didn´t comment anything realted to optimize IL-2 in VR or benefit from multithread.

 

Don´t spend money in this rig if you want to enjoy IL-2 VR. If you already have the CPU with you try to sell it and buy another rig with better singlethread performance.

Someone can borrow you a 1080 or 1080Ti and you will see that your performance is very low.

 

I am sorry to give you bad news but it is better to tell you the real facts we found along this year. More at:

https://forum.il2sturmovik.com/topic/29322-measuring-rig-performance-common-baseline/

Edited November 15, 2017 by chiliwili69

[polnoch_ksu]

Thank you for answer. What happen if I try to overcloked cpu and/or memory? This cpu in the single thread not too bad in comparation with:

 

http://cpu.userbenchmark.com/Compare/Intel-Xeon-X5690-vs-Intel-Core-i5-2500K/m16752vs619

 

And in the several of tests uses overlocked i5-2500K. And results are better.

 

So, for my other tasks in this PC I do not want to use overclocing. The power of PC is enough. And I want to try make minimal-required overcloking. What is most required for the IL2? The RAM?

 

 

4) Devs didn´t comment anything realted to optimize IL-2 in VR or benefit from multithread.

 

 

 

Anybody asked they?

Edited November 15, 2017 by polnoch_ksu

[chiliwili69]

 

 

And in the several of tests uses overlocked i5-2500K. And results are better.

 

The i5-2500K is also not great for IL-2 VR, although it is better than the CPU Xeon you asked.

 

RR_Kraft was having an avg of 48fps with the test (https://docs.google.com/spreadsheets/d/1gJmnz_nVxI6_dG_UYNCCpZVK2-f8NBy-y1gia77Hu_k) using OC to 4.4, but he said he had about 80-90 with Low graphics :

 

https://forum.il2sturmovik.com/topic/31820-i5-2500k-generally-weak-game/

 

So, you can try to put a decent cooler, put a better RAM (1600 or more), OC the CPU to 4.4 or 4.5 and use a 1070GPU (it is more than enough for your CPU and you can keep it if you do a future upgrade) and see what you achieve , but I am not sure what you will achieve.

[TUS_Samuel]

4) Devs didn´t comment anything realted to optimize IL-2 in VR or benefit from multithread.

 

In august 2017 devs (AnPetrovich) said that multithreading of the simulations is in wish list but not planned yet. I'm personally not sure if the bottleneck is in simulation thread because we see this bad correlation even in flight records.

 

 

Il-2 users traditionally should wait (for multithreading, for ideal headsets, for comments from devs, for bugs fixed, for perfection), nothing new here.

 

[chiliwili69]

 

 

In august 2017 devs (AnPetrovich) said that multithreading of the simulations is in wish list but not planned yet. I'm personally not sure if the bottleneck is in simulation thread because we see this bad correlation even in flight records.

 

Yes, I knew that, but I think it was more a wish than a commitment. We will see.

 

A flight record is not different to a normal flight with regards to performance. The only difference with a recorded flight is that you don´t control the plane.

[dburne]

 

A flight record is not different to a normal flight with regards to performance. The only difference with a recorded flight is that you don´t control the plane.

 

There certainly was a big difference in my case.

 

Running that Balapan track was way different in performance to my actually flying the plane. 

I have played several scripted campaigns, and am about 170 missions into a PWCG campaign. 

The performance I am getting, even in heavy combat situations, is way better than the performance I saw whilst running that track. GPU Boost for my 1080 Ti would not even kick in with that track.

 

No comparison.

[chiliwili69]

 

 

There certainly was a big difference in my case.

 

Yeah, I remember I pushed to you a bit to run the test and you had 52 fps, but according to your system and ST Passmark you should achieve 66 fps, (14fps more).

 

Perhaps Fraps didn´t run well in your system or there might be something else affecting the test. 

 

But independently of the your test, when a recorded flight is played, BOS has to render everything again (including the rotation of the head) like if was a normal flight. So, in terms of performance should be the same.

 

You can try to record your own track from a QMB or PWCG and you will see that you obtain the same performance than when you flight that QMB/PWCG.

[dburne]

Yeah, I remember I pushed to you a bit to run the test and you had 52 fps, but according to your system and ST Passmark you should achieve 66 fps, (14fps more).

 

Perhaps Fraps didn´t run well in your system or there might be something else affecting the test. 

 

But independently of the your test, when a recorded flight is played, BOS has to render everything again (including the rotation of the head) like if was a normal flight. So, in terms of performance should be the same.

 

You can try to record your own track from a QMB or PWCG and you will see that you obtain the same performance than when you flight that QMB/PWCG.

 

Not even close to the same in terms of performance.

I am mostly at 90 fps during my SP flying even in PWCG campaign, dropping into maybe the 70's during heavy combat, 60's at the very worse. And even with ASW off, still very fluid and smooth.

 

All that matters to me is the performance I get in the game itself whilst actually playing, which is very good. 

My 1080 Ti boosts to full clock and as long as temps stay down which they typically do it pretty much stays there for the whole mission.