Jump to content

SYN_Vander BENCHMARK v6 to measure IL-2 performance in monitor & VR


Recommended Posts

18 minutes ago, chiliwili69 said:

 

That´s a good question. I have never used on per-application tab in SteamVR (in fact IL-2 doesn´t appear there as it is not IL-2 from Steam).

I only use the General or Video tab which are the same.

 

I believe that you should leave on per application setting at 100%, so the value specified in the General tab is not affected when running IL-2. I mean 54%*100% is 54%.

 

For the G2, set global to app 50% to get as close to native as can ( current bug with G2 and Steam VR)

Per application settings you can set to 100% it is accurate.

Also note, you can add non Steam games to the Steam library, so the game does not have to be running to be able to access the per game setting in Steam VR.

Link to post
Share on other sites
On 11/27/2020 at 12:26 AM, Jaegerschnitzel said:

VR Test with Reverb G2:

Frames: 3199, Time 60000, Min: 44, Max: 67, Avg: 53.17

 

Yesterday I got in contact with a friendly guy who got his G2 already on friday (pre-ordered directly from German HP Store on 15th July). He wasn´t happy with the field of view and sold me his G2 for a fair amount of money in between the reduced pre-order-price and the now regular price. So it was a win-win-situation.

 

I added the VR results to my original post. It seems that my VR result is bottlenecked by the CPU as chiliwili got a nearly identical result with a similar powerful CPU and a slightly less powerfull GPU.

 

I´ll write my first impressions on the G2 in the G2 sub-forum today in the evening. But don´t take my opinions to serious as this is my first own VR headset, the only comparision I have is the Index that I tried a couple of times at a friend of mine. Spoiler alert: Fliying Il-2 in VR with the reverb G2 is awesome 🙂

 

Edited by Jaegerschnitzel
Link to post
Share on other sites

@Jaegerschnitzel Stell unbedingt sicher, dass du die Rechnung mit dabei hast, und ein bisschen Kontakt behälst. Auf Reddit häufen sich die Beschwerden über Ausfälle, tote Pixel, sterbende Kopfhörer, und ähnliche Probleme. Nur damit du sicher bist! Viel Spaß damit, meine kommt Montag :))

 

 

@dburne According to HP now, we would be supposed to use 100% resolution in SteamVR so that the distortion profile would get mapped properly in the framebuffer, so to experience edge-to-edge clarity. This would be coming directly from an HP Engineer.

 

So, the VR load we need to test are not 9Mio pixels, but 19.5Mio. Or as high as we can get it :rofl:

Link to post
Share on other sites
4 minutes ago, SCG_Fenris_Wolf said:

@Jaegerschnitzel Stell unbedingt sicher, dass du die Rechnung mit dabei hast, und ein bisschen Kontakt behälst. Auf Reddit häufen sich die Beschwerden über Ausfälle, tote Pixel, sterbende Kopfhörer, und ähnliche Probleme. Nur damit du sicher bist! Viel Spaß damit, meine kommt Montag :))

 

 

@dburne According to HP now, we would be supposed to use 100% resolution in SteamVR so that the distortion profile would get mapped properly in the framebuffer, so to experience edge-to-edge clarity. This would be coming directly from an HP Engineer.

 

So, the VR load we need to test are not 9Mio pixels, but 19.5Mio. Or as high as we can get it :rofl:

 

 

Seems odd to me with the headset listing native res at 2160x2160, and 100% in Steam VR giving much higher.

I will have to give it a try and see how it looks and performs. Most common belief was it was a bug. 

At 50% I can't complain with the sweetspot, seems at least as good as my Rift S was if not slightly better.

Link to post
Share on other sites
2 hours ago, dburne said:

 

 

Seems odd to me with the headset listing native res at 2160x2160, and 100% in Steam VR giving much higher.

I will have to give it a try and see how it looks and performs. Most common belief was it was a bug. 

At 50% I can't complain with the sweetspot, seems at least as good as my Rift S was if not slightly better.

 

The only comparision I have is the Index and the size of the sweetspot of the G2 seems a little smaller to me. If I had to make a guess, I´d say that the absolutely sharp area covers about  30% of the vertical and 25 % of the horizontal FOV. If I keep my head straight and move my eyes away from the center, the image starts to become blurry outside of that 30%-25%-area slowly, but it doesn´t bother me much while playing. I tried supersampling of 50 and 100 % in SteamVR and I honestly can´t see any changes in the size of that central sharp area. I will most likely stay at 50% to keep acceptable performance... But please keep in mind that this is my first own headset and I might have missed some hidden settings.

 

@SCG_Fenris_Wolf Ich habe die Rechnung bekommen und der Verkäufer hat mir beim gehen noch hinterhergesagt, dass ich mich gerne nochmal melden kann, falls es irgendwelche Probleme gibt. Ich war ne lange Zeit da und durfte das Headset ganz in Ruhe ausprobieren etc. In der Hinsicht wird es denke ich keine Probleme geben. Ich bin mal gespannt, was du in den nächsten Tagen zum Sweetspot zu sagen hast. Ich bin nicht wirklich enttäuscht, hätte mir aber nach dem ganzen Hype bei MRTV einen etwas größeren "klaren Bildanteil" erwartet. Alles andere hat mich aber absolut vom Hocker gehauen, vor allem natürlich das astrein scharfe Bild in der Mitte des Sehbereichs! Viel Spaß ab morgen mit deinem Headset!

Link to post
Share on other sites

For the G2 owners, What are the pixels per eye values shown by SteamVR when you set the SS slide bar at 50% and 100%? (please provide screenshoot if you can)

 

If you look to the column I in this spreadsheet you will see the "absolute SS ratio" which every device is using.

The Index and Vive Pro were using 1.96 (it is 196% over the native display resolution).

 

According to the number of reddit, the G2 would be using 2.09 (209% over the native resolution). So, given that they use the same lens technology than Index it could be not a bug, but the intended resolution.

 

Let´s see if it is really needed to have 100%SS to have an optimum clarity. 

Link to post
Share on other sites

Here's my updated VR test:

Reverb G2 at 90Hz, 50% SteamVR (2236 x 2188 x 2 = 9,784,736 pixels) in SteamVR Settings -> General tab (and repeated on Video tab) "Resolution per eye". Also, SteamVR Settings -> Video -> Per-Application Video Settings -> "Custom Resolution Multiplier" set at 100%.

 

2020-12-06 11:02:40 - Il-2
Frames: 4851 - Time: 56062ms - Avg: 86.529 - Min: 88 - Max: 91

 

(For some reason SteamVR settings for 54% show as 2324 x 2272. SteamVR also won't offer me an option that replicates dburne's 2205 x 2165.)

 

Computer specs repeated below from before, no changes:

Motherboard: MSI MPG Z490 Gaming Carbon WIFI
CPU: i7 10700K
CPU Freq: 5.1 Ghz
L3 cache: 16 MB
Cores: 8
Threads: 8
RAM type: DDR4
RAM size: 32Gb (4x8GB)
Uncore Frequency: 4801.2 MHz
RAM Freq: 4000 MHz
RAM timings: 17-17-17-37
GPU: EVGA RTX 3090 FTW3 Ultra (thank you, Mrs. Charlo)
 

SteamVRVideo.jpg

SteamVRPer-ApplicationIL2.jpg

Link to post
Share on other sites
2 hours ago, chiliwili69 said:

For the G2 owners, What are the pixels per eye values shown by SteamVR when you set the SS slide bar at 50% and 100%? (please provide screenshoot if you can)

 

If you look to the column I in this spreadsheet you will see the "absolute SS ratio" which every device is using.

The Index and Vive Pro were using 1.96 (it is 196% over the native display resolution).

 

According to the number of reddit, the G2 would be using 2.09 (209% over the native resolution). So, given that they use the same lens technology than Index it could be not a bug, but the intended resolution.

 

Let´s see if it is really needed to have 100%SS to have an optimum clarity. 

 

It certainly does not jive with what the original Reverb shows, which is has the same resolution so does not make much sense to me.

Will see if Steam makes any changes there in the near future.

Link to post
Share on other sites
3 hours ago, Charlo-VR said:

Here's my updated VR test:

Reverb G2 at 90Hz, 50% SteamVR (2236 x 2188 x 2 = 9,784,736 pixels) in SteamVR Settings -> General tab (and repeated on Video tab) "Resolution per eye". Also, SteamVR Settings -> Video -> Per-Application Video Settings -> "Custom Resolution Multiplier" set at 100%.

 

2020-12-06 11:02:40 - Il-2
Frames: 4851 - Time: 56062ms - Avg: 86.529 - Min: 88 - Max: 91

 

(For some reason SteamVR settings for 54% show as 2324 x 2272. SteamVR also won't offer me an option that replicates dburne's 2205 x 2165.)

 

Computer specs repeated below from before, no changes:

Motherboard: MSI MPG Z490 Gaming Carbon WIFI
CPU: i7 10700K
CPU Freq: 5.1 Ghz
L3 cache: 16 MB
Cores: 8
Threads: 8
RAM type: DDR4
RAM size: 32Gb (4x8GB)
Uncore Frequency: 4801.2 MHz
RAM Freq: 4000 MHz
RAM timings: 17-17-17-37
GPU: EVGA RTX 3090 FTW3 Ultra (thank you, Mrs. Charlo)
 

SteamVRVideo.jpg

SteamVRPer-ApplicationIL2.jpg

Awesome, can you repeat the test using 80% supersampling please?

I want to know how your amazing 3090 performs with full clarity as intended by HP. I got a 3080 here and am thinking about switching it, I may run the 6900XT in the test as well if I get one, and then decide between that and the 3090. I've read the 3090 scale much better at high resolutions like these and slaughter the AMD cards.

 

Btw here is the confirmation that the target resolution is indeed the one that sits at 100% https://www.reddit.com/r/HPReverb/comments/k82pnv/definitive_answer_for_the_100_resolution/?utm_medium=android_app&utm_source=share

  • Upvote 1
Link to post
Share on other sites
25 minutes ago, SCG_Fenris_Wolf said:

Awesome, can you repeat the test using 80% supersampling please?

I want to know how your amazing 3090 performs with full clarity as intended by HP. I got a 3080 here and am thinking about switching it, I may run the 6900XT in the test as well if I get one, and then decide between that and the 3090. I've read the 3090 scale much better at high resolutions like these and slaughter the AMD cards.

 

Btw here is the confirmation that the target resolution is indeed the one that sits at 100% https://www.reddit.com/r/HPReverb/comments/k82pnv/definitive_answer_for_the_100_resolution/?utm_medium=android_app&utm_source=share

 

So is that an official statement by HP?

If so wonder why the difference over the first gen Reverb.

 

Edited by dburne
Link to post
Share on other sites
1 hour ago, SCG_Fenris_Wolf said:

Awesome, can you repeat the test using 80% supersampling please?

I want to know how your amazing 3090 performs with full clarity as intended by HP. I got a 3080 here and am thinking about switching it, I may run the 6900XT in the test as well if I get one, and then decide between that and the 3090. I've read the 3090 scale much better at high resolutions like these and slaughter the AMD cards.

 

Btw here is the confirmation that the target resolution is indeed the one that sits at 100% https://www.reddit.com/r/HPReverb/comments/k82pnv/definitive_answer_for_the_100_resolution/?utm_medium=android_app&utm_source=share

 

Sure! Which setting do I change to 80% - Video settings "Resolution per eye" or in the Per-Application Video Settings -> "Custom Resolution Multiplier"?  It may be best to tell me what number you want for both so I can run the benchmark with the settings you want 🤓

Link to post
Share on other sites

This post is for the bare metal CPU test results with the settings mentioned in the guide.

 

OS: Windows 10 Professional

CPU: 5900X

CPU AIO: EK 360

GPU: 3090 EVGA ftw3 ultra, Base Core clock +200 (2000MHz)

RAM: 3600 18-22-22-42-85 1T

Motherboard: X570 Aorus Master

 

PBO +200MHz OC on an EK 360 AIO didn't really make much of a difference for me, only 1 frame higher on average and max framerate compared to all my other test results. The lows fluctuate for me, even with PBO off I had around 76fps minimum in one test. Might be after the benchmark goes to a black screen and transitions to the main menu.

 

Sick of retesting the CPU portion want to move onto GPU passthrough and VR those are more interesting.

I will just disable PBO for future tests until PBO 2 is available. Can't be stuffed tuning DRAM, I won't gain much for all the hassle.

 

Test Without PBO (GPU clock was +150MHz shouldn't impact the result much)

07-12-2020, 15:56:51 Il-2.exe benchmark completed, 6987 frames rendered in 58.968 s
                     Average framerate  :  118.4 FPS
                     Minimum framerate  :  100.8 FPS
                     Maximum framerate  :  153.1 FPS

 

Test with PBO +200MHz (GPU clock +200MHz):

07-12-2020, 16:27:06 Il-2.exe benchmark completed, 7029 frames rendered in 59.000 s
                     Average framerate  :  119.1 FPS
                     Minimum framerate  :   86.5 FPS
                     Maximum framerate  :  153.9 FPS

 

I attached a hardware monitoring Riva tuner log file, if you have MSI afterburner you can view the charts for each log entry alternatively the data can be read with a text editor.

Ignore GPU2 it is a 980 that I am using on the Host OS when doing GPU passthrough, GPU1 is the 3090 that I am testing.

 

If viewing the charts for reference:

 

Spoiler

 

  • GPU Core + 150, Mem 2100MHz, failed mem OC
    • 07-12-2020, 15:37:43 Il-2.exe benchmark completed, 6543 frames rendered in 61.438 s
      Average framerate  :  106.4 FPS
      Minimum framerate  :   72.4 FPS
      Maximum framerate  :  148.1 FPS
      1% low framerate   :   37.9 FPS
      0.1% low framerate :    2.7 FPS
  • GPU Core + 150, Mem 3600MHz, 18-22-22-42-85 1T
    • 07-12-2020, 15:56:51 Il-2.exe benchmark completed, 6987 frames rendered in 58.968 s
      Average framerate  :  118.4 FPS
      Minimum framerate  :  100.8 FPS
      Maximum framerate  :  153.1 FPS
      1% low framerate   :   56.5 FPS
      0.1% low framerate :   40.3 FPS
  • GPU Core +200, Mem 3600MHz, 18-22-22-42-85 1T
    • 07-12-2020, 16:04:29 Il-2.exe benchmark completed, 6900 frames rendered in 60.515 s
      Average framerate  :  114.0 FPS
      Minimum framerate  :   76.9 FPS
      Maximum framerate  :  153.2 FPS
      1% low framerate   :   45.5 FPS
      0.1% low framerate :    3.1 FPS
  • PBO on +200MHz, GPU Core +200, Mem 3600MHz, 18-22-22-42-85 1T
    • 07-12-2020, 16:27:06 Il-2.exe benchmark completed, 7029 frames rendered in 59.000 s
      Average framerate  :  119.1 FPS
      Minimum framerate  :   86.5 FPS
      Maximum framerate  :  153.9 FPS
      1% low framerate   :   47.2 FPS
      0.1% low framerate :    3.3 FPS

 

 

 

 

HardwareMonitoring-cpu-bare metal.zip

Edited by sunnyB
  • Like 1
Link to post
Share on other sites
14 hours ago, Charlo-VR said:

50% SteamVR (2236 x 2188 x 2 = 9,784,736 pixels)

Thank you Charlo for runing again the VR test. You achieve a good number as well.

To measure all devices with the same GPU load I was making the instructions for all VR headset to be as much as close to 9.54 Mpixels. So you did it well. The closest number is 9.78 Million using 50%. I will modify the instructions (since 54% was wrong).

 

The values shown for 100% are slightly different for every user (maybe due to IPD):

Jaegerschnitzel    3176 x 3104
reddit                      3160 x 3092
reddit                      3170 x 3100
Charlo                    3162  x 3093

 

So using 50% for the test is OK to have a common basis with other VR devices, but we could request in the instructions to run a second VR test with higher number of pixels (this is info will useful for G2 or Pimax users), for example 19.5 million pixels. Then GPU is bottleneck for sure.

Link to post
Share on other sites
15 hours ago, Charlo-VR said:

Frames: 4851 - Time: 56062ms - Avg: 86.529 - Min: 88 - Max: 91

 

Something is not correct in this test, the Time shoud show 60000ms, not 56062. I don´t know why. (The P to unpasue IL-2 should be the same than the P to make fraps to start the 60 seconds test). Your previous VR test showed 60000ms.

14 hours ago, dburne said:

It certainly does not jive with what the original Reverb shows, which is has the same resolution so does not make much sense to me

 

The previous G1 uses a different types of lenses. They suffered from an small Sweet spot and also chromatic aberrations in the edges. 

Here the G2 uses the same lense technology that Valve patented.

This lenses are used also in the Index, and the Index needed a high target multiplier x1.4 over the native display. That´s why Index physical display is 1440x1600 and 100%SS is 2016x2240.

So you can not extrapolate numbers of the G1 over the G2. They use totally different technology although the native panel has the same resolution 2160x2160. (Also native panels of G1 and G2 are not the same models).

According to HP it seems that the G2 need a x1.5 multiplier, this should should be around 105%SS in the G2.

10 hours ago, Charlo-VR said:

Sure! Which setting do I change to 80% - Video settings "Resolution per eye" or in the Per-Application Video Settings -> "Custom Resolution Multiplier"?  It may be best to tell me what number you want for both so I can run the benchmark with the settings you want

 

He refers to touch only the General/Video settings, from 50% to 80% (or also 100%). So the per aaplication settings remains always untouched at 100%.

Edited by chiliwili69
Link to post
Share on other sites
19 minutes ago, chiliwili69 said:

The previous G1 uses a different types of lenses. They suffered from an small Sweet spot and also chromatic aberrations in the edges. 

Here the G2 uses the same lense technology that Valve patented.

This lenses are used also in the Index, and the Index needed a high target multiplier x1.4 over the native display. That´s why Index physical display is 1440x1600 and 100%SS is 2016x2240.

So you can not extrapolate numbers of the G1 over the G2. They use totally different technology although the native panel has the same resolution 2160x2160. (Also native panels of G1 and G2 are not the same models).

According to HP it seems that the G2 need a x1.5 multiplier, this should should be around 105%SS in the G2.

 

Does this mean HP stating the G2 does not require more computing power than the G1 was false?

Link to post
Share on other sites
5 hours ago, sunnyB said:

Test with PBO +200MHz

Thank you for your test. The Rivatuner provides a much rich set of info, but fraps is much quicker to report and test.

I believe the Rivatuner doesn´t differenciate between GPU vs CPU frametimes, like fpsVR does.

I believe that fpsVR provides a hotkey to start/stop logging both CPU and GPU frametimes, but it would complicate the test in VR. (in addition fpsVR is a paid tool but worth)

 

I don´t know how PBO works exactly, but I understand the CPU freq was mantained at the stock freq + 200. right?

Link to post
Share on other sites
1 hour ago, HunDread said:

 

Does this mean HP stating the G2 does not require more computing power than the G1 was false?

 

Good question. I find this all rather confusing myself.

 

Edit:

From an HP MS update for the G2 three days ago:

 

Quote

 


WMR render target in SteamVR? This is always changing as SteamVR and WMR update themselves. SteamVR reports the total framebuffer, which in VR is larger than the panel resolution to allow for last millisecond adjustments to the displayed frame to match any movement of the user’s head from when the frame was rendered. So, no need to update it manually.
 

 

 

Not sure I understand completely that statement, but according to HP for native res it should not be adjusted over what it already shows at 100%. At least that is how I read it, and of course anyone is free to adjust the resolution to achieve the performance and quality they prefer. No one says what is right for the individual. I will probably have a go with it later today or more likely tomorrow and see how mine does around 100% res.

Full update:

https://www.reddit.com/r/HPReverb/comments/k6504n/hp_microsoft_reverb_g2_update/

 

Edited by dburne
Link to post
Share on other sites

Bare metal windows 10 pimax 8kx and 3090 test results, might do some light testing for a VM with GPU passthrough later. I am just so over testing, funnily enough my job is also to test stuff all day I was supposed to relax on my day off :help:

 

Summary & thoughts:

  • performance penalty for all AA is high so I never bothered testing with it on
  • FoV normal seemed to add a decent amount of fps.
  • Dropping the preset from ultra to high did add a good chunk of performance for the same settings.
  • Everything else did not seem to contribute a sizeable amount but they all add up
  • Enabling smart smoothing and lowering settings that I don't notice on ultra to lock me at 45fps sounds tempting granted I don't get distracted by any artifacts
  • High seems like a good place to start, need to A/B test what settings I will notice easily then go from there

 

I compiled some test results on the 8kx starting with the recommended settings then going berserk with ultra preset and pitool, then knocking them down here and there until preset was dropped to high, after which I dropped the settings to lowest then increased some settings here and there. I didn't test all combinations.

 

unfortunately I could not test fixed foveated rendering, no matter what I did (stock clocks, restart steamVR, reboot) I kept crashing at the start of the mission when any FFR mode other than off("close") was set.

I seemed to be capped at the display refresh rate despite fps target in game being set to unlimited. I checked nvidia control panel and steamVR for any vsync/frame limiting settings but none were enabled.

 

And most importantly there was no method, only madness.

 

hardware, software, and test environment related info:

Spoiler

pitool version: V1.0.1.266

HMD firmware version: V2.1.255.294

OS: Windows 10 Professional

CPU: 5900X

CPU AIO: EK 360

GPU: 3090 EVGA ftw3 ultra, Base Core clock +140 (1940MHz)

RAM: 3600 18-22-22-42-85 1T

Motherboard: X570 Aorus Master

Ambient temperature range 20~30 degrees Celsius

 

steamVR advanced supersample filtering was always off

All benchmarks were conducted with the SYN_Vader benchmarkv6 mission in the OP, MSI afterburner was used to capture the results.

 

Average/Max/Min Framerate tests with descriptions.

if you have MSI afterburner you can view the attached log file as charts and use the description text + timestamps here as reference, unfortunately not all the data was captured it must have cleared the monitoring log file when I cleared history in the monitoring tool as well, what a waste I should've just carried on testing without resetting the chart after each test run or closed and reopened MSI afterburner.

 

test results:

Spoiler

forum settings, GPU core +140, capped at panel refresh rate for reasons unkown target fps was set to off
07-12-2020, 18:11:37 Il-2.exe benchmark completed, 3720 frames rendered in 49.734 s
Average framerate  :   74.7 FPS
Minimum framerate  :   72.7 FPS
Maximum framerate  :   75.9 FPS
1% low framerate   :   46.3 FPS
0.1% low framerate :   31.8 FPS

 

forum settings, GPU core +140, rerun. Same issue
07-12-2020, 18:18:54 Il-2.exe benchmark completed, 4843 frames rendered in 65.156 s
Average framerate  :   74.3 FPS
Minimum framerate  :   54.3 FPS
Maximum framerate  :   75.9 FPS
1% low framerate   :   35.8 FPS
0.1% low framerate :    3.6 FPS

 

forum settings, GPU core +140, rerun after a reboot. Same issue
07-12-2020, 18:34:08 Il-2.exe benchmark completed, 4584 frames rendered in 61.578 s
Average framerate  :   74.4 FPS
Minimum framerate  :   47.4 FPS
Maximum framerate  :   75.9 FPS
1% low framerate   :   37.7 FPS
0.1% low framerate :    2.7 FPS

 

FoV Large, GPU core +140, other settings same as metioned in OP
07-12-2020, 18:51:12 Il-2.exe benchmark completed, 4386 frames rendered in 59.078 s
Average framerate  :   74.2 FPS
Minimum framerate  :   49.8 FPS
Maximum framerate  :   75.8 FPS
1% low framerate   :   35.8 FPS
0.1% low framerate :    3.0 FPS

 

 GPU core +140, SS 100%, everything else same as OP (FoV normal)
07-12-2020, 18:57:15 Il-2.exe benchmark completed, 4571 frames rendered in 61.578 s
Average framerate  :   74.2 FPS
Minimum framerate  :   50.7 FPS
Maximum framerate  :   76.3 FPS
1% low framerate   :   36.7 FPS
0.1% low framerate :    3.1 FPS

 

 GPU core +140, FXAA set to off, ultra all settings maxed out, map scenery unlimited, highest sound quality, FoV normal, 100%SS, render quality 1
07-12-2020, 19:45:16 Il-2.exe benchmark completed, 3182 frames rendered in 62.609 s
Average framerate  :   50.8 FPS
Minimum framerate  :   35.9 FPS
Maximum framerate  :   70.3 FPS
1% low framerate   :   18.9 FPS
0.1% low framerate :    3.0 FPS

 

repeat - GPU core +140, FXAA set to off, ultra all settings maxed out, map scenery unlimited, highest sound quality, FoV large, 100%SS, render quality 1
07-12-2020, 19:48:48 Il-2.exe benchmark completed, 3159 frames rendered in 62.157 s
Average framerate  :   50.8 FPS
Minimum framerate  :   32.8 FPS
Maximum framerate  :   74.8 FPS
1% low framerate   :   24.9 FPS
0.1% low framerate :    2.5 FPS

 

repeat - GPU core +140, FXAA set to off, ultra all settings maxed out, FixedFoveatedRendering=off, large FoV
07-12-2020, 19:53:07 Il-2.exe benchmark completed, 3100 frames rendered in 61.219 s
Average framerate  :   50.6 FPS
Minimum framerate  :   31.8 FPS
Maximum framerate  :   53.9 FPS
1% low framerate   :   27.7 FPS
0.1% low framerate :    2.4 FPS

 

repeat -  GPU core +140, FXAA set to off, ultra all settings maxed out, FixedFoveatedRendering=off, large FoV
07-12-2020, 19:55:50 Il-2.exe benchmark completed, 3085 frames rendered in 60.797 s
Average framerate  :   50.7 FPS
Minimum framerate  :   35.8 FPS
Maximum framerate  :   53.9 FPS
1% low framerate   :   29.3 FPS
0.1% low framerate :    3.0 FPS

 

skipped this test - FFR=Conservative crashing the moment I unpause even after reverting to factory clocks and restarting all VR related services and applications

skipped this test - FFR=Balanced crashing the moment I unpause

skipped this test - FFR=Aggressive crashing the moment I unpause

 

GPU core +140, FXAA set to off, ultra all settings maxed out, FixedFoveatedRendering=off, small FoV (I suspect the settings did not apply)
07-12-2020, 20:34:57 Il-2.exe benchmark completed, 2777 frames rendered in 63.937 s
Average framerate  :   43.4 FPS
Minimum framerate  :   27.5 FPS
Maximum framerate  :   70.2 FPS
1% low framerate   :   18.4 FPS
0.1% low framerate :    2.5 FPS

 

restarted steamVR and repeated - GPU core +140, FXAA set to off, ultra all settings maxed out, FixedFoveatedRendering=off, small FoV
07-12-2020, 20:46:37 Il-2.exe benchmark completed, 3609 frames rendered in 62.250 s
Average framerate  :   57.9 FPS
Minimum framerate  :   38.8 FPS
Maximum framerate  :   74.9 FPS
1% low framerate   :   27.2 FPS
0.1% low framerate :    2.6 FPS

 

GPU core +140, FXAA set to off, ultra all settings maxed out, FixedFoveatedRendering=off, FoV=potato, apply success! save success! much wow, much success (I suspect the settings did not apply)
07-12-2020, 20:52:00 Il-2.exe benchmark completed, 3640 frames rendered in 62.422 s
Average framerate  :   58.3 FPS
Minimum framerate  :   42.1 FPS
Maximum framerate  :   63.2 FPS
1% low framerate   :   26.1 FPS
0.1% low framerate :    3.0 FPS

 

restart steamVR and repeat - GPU core +140, FXAA set to off, ultra all settings maxed out, FixedFoveatedRendering=off, FoV potato
07-12-2020, 20:55:30 Il-2.exe benchmark completed, 3899 frames rendered in 61.938 s
Average framerate  :   62.9 FPS
Minimum framerate  :   41.7 FPS
Maximum framerate  :   74.4 FPS
1% low framerate   :   26.7 FPS
0.1% low framerate :    2.6 FPS

 

Clouds dropped from ultra to high - restarted steamVR set FoV to large, everything else maxed out as before. SS100%, pitool render 1, etc...
07-12-2020, 21:02:03 Il-2.exe benchmark completed, 3603 frames rendered in 72.656 s
Average framerate  :   49.5 FPS
Minimum framerate  :   30.6 FPS
Maximum framerate  :   75.6 FPS
1% low framerate   :   17.8 FPS
0.1% low framerate :    2.6 FPS

 

retest - retest of the previos test with cloud quality high
07-12-2020, 21:07:09 Il-2.exe benchmark completed, 2877 frames rendered in 62.297 s
Average framerate  :   46.1 FPS
Minimum framerate  :   36.5 FPS
Maximum framerate  :   62.3 FPS
1% low framerate   :   26.2 FPS
0.1% low framerate :    3.9 FPS

 

dropped horizon draw distance to 100km from the previous settings (was 150Km)
07-12-2020, 21:15:42 Il-2.exe benchmark completed, 3344 frames rendered in 68.954 s
Average framerate  :   48.4 FPS
Minimum framerate  :   30.3 FPS
Maximum framerate  :   75.4 FPS
1% low framerate   :   20.4 FPS
0.1% low framerate :    2.5 FPS

 

dropped map scenary distance in Game menu from unlimited to 50Km
07-12-2020, 21:20:23 Il-2.exe benchmark completed, 2925 frames rendered in 62.922 s
Average framerate  :   46.4 FPS
Minimum framerate  :   35.8 FPS
Maximum framerate  :   75.4 FPS
1% low framerate   :   22.1 FPS
0.1% low framerate :    3.5 FPS

 

dropped grass quality to distant from previous settings
07-12-2020, 21:24:26 Il-2.exe benchmark completed, 3254 frames rendered in 66.516 s
Average framerate  :   48.9 FPS
Minimum framerate  :   34.6 FPS
Maximum framerate  :   75.2 FPS
1% low framerate   :   21.5 FPS
0.1% low framerate :    3.2 FPS

 

dropped grass quality to normal from previous settings
07-12-2020, 21:28:29 Il-2.exe benchmark completed, 2946 frames rendered in 61.313 s
Average framerate  :   48.0 FPS
Minimum framerate  :   35.3 FPS
Maximum framerate  :   51.9 FPS
1% low framerate   :   22.9 FPS
0.1% low framerate :    3.1 FPS

 

canopy reflections dropped to normal
07-12-2020, 21:32:09 Il-2.exe benchmark completed, 2983 frames rendered in 61.859 s
Average framerate  :   48.2 FPS
Minimum framerate  :   34.4 FPS
Maximum framerate  :   55.3 FPS
1% low framerate   :   24.3 FPS
0.1% low framerate :    2.9 FPS

 

shadows dropped to high
07-12-2020, 21:37:28 Il-2.exe benchmark completed, 2917 frames rendered in 60.781 s
Average framerate  :   47.9 FPS
Minimum framerate  :   31.8 FPS
Maximum framerate  :   51.9 FPS
1% low framerate   :   24.3 FPS
0.1% low framerate :    2.5 FPS

 

mirrors dropped to medium
07-12-2020, 21:41:41 Il-2.exe benchmark completed, 3015 frames rendered in 61.797 s
Average framerate  :   48.7 FPS
Minimum framerate  :   35.3 FPS
Maximum framerate  :   70.2 FPS
1% low framerate   :   22.6 FPS
0.1% low framerate :    3.0 FPS

 

grass off
07-12-2020, 21:45:22 Il-2.exe benchmark completed, 3171 frames rendered in 63.922 s
Average framerate  :   49.6 FPS
Minimum framerate  :   32.7 FPS
Maximum framerate  :   75.7 FPS
1% low framerate   :   22.0 FPS
0.1% low framerate :    2.6 FPS

 

horizon draw distance dropped to 70km
07-12-2020, 21:49:10 Il-2.exe benchmark completed, 3144 frames rendered in 63.469 s
Average framerate  :   49.5 FPS
Minimum framerate  :   37.7 FPS
Maximum framerate  :   75.0 FPS
1% low framerate   :   22.9 FPS
0.1% low framerate :    3.6 FPS

 

mirrors off
07-12-2020, 21:53:39 Il-2.exe benchmark completed, 3606 frames rendered in 69.484 s
Average framerate  :   51.8 FPS
Minimum framerate  :   36.6 FPS
Maximum framerate  :   75.1 FPS
1% low framerate   :   20.5 FPS
0.1% low framerate :    3.1 FPS

 

shadows quality medium
07-12-2020, 21:57:41 Il-2.exe benchmark completed, 3081 frames rendered in 62.110 s
Average framerate  :   49.6 FPS
Minimum framerate  :   33.4 FPS
Maximum framerate  :   62.8 FPS
1% low framerate   :   18.2 FPS
0.1% low framerate :    2.5 FPS

 

set FoV=normal and restarted steamVR
07-12-2020, 22:02:51 Il-2.exe benchmark completed, 3659 frames rendered in 61.922 s
Average framerate  :   59.0 FPS
Minimum framerate  :   45.2 FPS
Maximum framerate  :   64.1 FPS
1% low framerate   :   26.2 FPS
0.1% low framerate :    3.5 FPS

 

dropped graphics preset from ultra to high with the same settings from the previous test
07-12-2020, 22:07:46 Il-2.exe benchmark completed, 4327 frames rendered in 67.500 s
Average framerate  :   64.1 FPS
Minimum framerate  :   45.8 FPS
Maximum framerate  :   75.6 FPS
1% low framerate   :   21.6 FPS
0.1% low framerate :    2.9 FPS

 

maxing out everything in the high preset and set FoV=large (except FXAA set that to off) any SS or render quality remains 1:1 or 100%
07-12-2020, 22:14:03 Il-2.exe benchmark completed, 2672 frames rendered in 61.859 s
Average framerate  :   43.1 FPS
Minimum framerate  :   31.3 FPS
Maximum framerate  :   74.6 FPS
1% low framerate   :   18.2 FPS
0.1% low framerate :    3.0 FPS

 

dropping all settings with the exception of any supersampling or render quality those were left at 1:1 or 100% (includes sound and map scenery in the settings >> game menu) dynamic resolution full, and FoV set to Large
07-12-2020, 22:20:15 Il-2.exe benchmark completed, 4307 frames rendered in 60.063 s
Average framerate  :   71.7 FPS
Minimum framerate  :   42.8 FPS
Maximum framerate  :   75.8 FPS
1% low framerate   :   31.1 FPS
0.1% low framerate :    2.5 FPS

 

increased distant landscape detail to x4 - same as previous settings (everything else was lowered)
07-12-2020, 22:24:51 Il-2.exe benchmark completed, 4343 frames rendered in 60.328 s
Average framerate  :   71.9 FPS
Minimum framerate  :   43.6 FPS
Maximum framerate  :   75.7 FPS
1% low framerate   :   33.2 FPS
0.1% low framerate :    2.5 FPS

 

shadows increased to ultra
07-12-2020, 22:28:01 Il-2.exe benchmark completed, 3476 frames rendered in 51.813 s
Average framerate  :   67.0 FPS
Minimum framerate  :   45.4 FPS
Maximum framerate  :   74.1 FPS
1% low framerate   :   31.4 FPS
0.1% low framerate :    3.0 FPS

 

SSAO enabled
07-12-2020, 22:34:28 Il-2.exe benchmark completed, 4095 frames rendered in 62.063 s
Average framerate  :   65.9 FPS
Minimum framerate  :   51.5 FPS
Maximum framerate  :   72.1 FPS
1% low framerate   :   30.7 FPS
0.1% low framerate :    4.3 FPS

 

hdr, sharpen, 4k textures, and distant buildings were all enabled
07-12-2020, 22:38:08 Il-2.exe benchmark completed, 3961 frames rendered in 62.906 s
Average framerate  :   62.9 FPS
Minimum framerate  :   44.6 FPS
Maximum framerate  :   75.6 FPS
1% low framerate   :   24.8 FPS
0.1% low framerate :    2.9 FPS

 

HardwareMonitoring.zip

@chiliwili69

I haven't added the cpu clock speeds to the chart, there probably is a setting in there to add it but for 1 fps if even that (already within margin of error) I wouldn't bother.

Link to post
Share on other sites

VM with GPU passthrough benchmarks, windows is run in a VM inside of linux, IL2 is run inside the VM on a GPU passed through to the VM.

Anyways I am knocking off I don't want to see any benchmarks for good long while.

 

Summary & thoughts:

  • Significant hit on the framerate compared to bare metal when running the CPU test, roughly 20% performance hit when 6 cores are pinned from the same chiplet.
  • Similar performance hit in VR when running the VR1 test mentioned in the OP.
  • VM might need more fine tuning.

 

hardware, software, and test environment related info:

Spoiler

 

pitool version: V1.0.1.266

HMD firmware version: V2.1.255.294

OS: Windows 10 Professional

CPU: 5900X

CPU AIO: EK 360

GPU: 3090 EVGA ftw3 ultra, Base Core clock +140 (1940MHz)

RAM: 3600 18-22-22-42-85 1T

Motherboard: X570 Aorus Master - PCIE gen4 is set in Bios

Ambient temperature range 17~20 degrees Celsius

 

Host:

  • OS: Arch Linux with i3 as the window manger
  • GPU: GTX980 mounted in pcie slot 2
  • CPU (core ids 8 to 13 are used by the host OS, core ids 6 and 7 are ccx I think)
  • hypervisor: KVM + QEMU

 

Guest:

  • OS: Windows 10 Professional
  • CPU: core ids 0 to 5 pinned to Guest (1 6 core chiplet dedicated to VM, host will not use these cpus until the VM is shutdown)
  • RAM: 24GB statically allocated with 1GB pagefile size
  • GPU: 3090 with Core Clock OC +140 - nvidia code 43 error workaround applied - GPU mounted into pcie slot 1
  • not using looking glass
  • not using spice for mouse+KB, a PCIE USB controller is passed through to the VM and all USB inputs are connected to it including the pimax 8kx.

 

 

VM - CPU test 1080p benchmark with the settings as mentioned in the OP:

Spoiler

first run was with OSD overlay enabled by mistake
07-12-2020, 14:59:29 Il-2.exe benchmark completed, 5027 frames rendered in 60.422 s
Average framerate  :   83.1 FPS
Minimum framerate  :   53.9 FPS
Maximum framerate  :  138.8 FPS
1% low framerate   :    2.6 FPS
0.1% low framerate :    1.9 FPS

 

second run without OSD overlay
07-12-2020, 15:03:58 Il-2.exe benchmark completed, 6054 frames rendered in 62.750 s
Average framerate  :   96.4 FPS
Minimum framerate  :   56.5 FPS
Maximum framerate  :  139.9 FPS
1% low framerate   :    3.7 FPS
0.1% low framerate :    2.6 FPS

 

Bare metal results (windows run on bare metal) from my previous post for the sake of comparison:

Test Without PBO (GPU clock was +150MHz shouldn't impact the result much)

07-12-2020, 15:56:51 Il-2.exe benchmark completed, 6987 frames rendered in 58.968 s
                     Average framerate  :  118.4 FPS
                     Minimum framerate  :  100.8 FPS
                     Maximum framerate  :  153.1 FPS

 

Test with PBO +200MHz (GPU clock +200MHz):

07-12-2020, 16:27:06 Il-2.exe benchmark completed, 7029 frames rendered in 59.000 s
                     Average framerate  :  119.1 FPS
                     Minimum framerate  :   86.5 FPS
                     Maximum framerate  :  153.9 FPS

 

VM - Pimax 8KX benchmarks:

Spoiler

vr test 1 settings from forum 78%SS pitool render 1 normal FoV
07-12-2020, 15:49:58 Il-2.exe benchmark completed, 3585 frames rendered in 61.297 s
Average framerate  :   58.4 FPS
Minimum framerate  :   32.6 FPS
Maximum framerate  :   72.9 FPS
1% low framerate   :    2.2 FPS
0.1% low framerate :    1.8 FPS

 

rerun of the previous test

07-12-2020, 15:54:12 Il-2.exe benchmark completed, 3665 frames rendered in 61.093 s
                     Average framerate  :   59.9 FPS
                     Minimum framerate  :   36.6 FPS
                     Maximum framerate  :   70.9 FPS
                     1% low framerate   :   30.1 FPS
                     0.1% low framerate :    2.7 FPS

 

Windows 10 running on bare metal result for comparison (from previous post):

forum settings, GPU core +140, capped at panel refresh rate for reasons unknown target fps was set to off
07-12-2020, 18:11:37 Il-2.exe benchmark completed, 3720 frames rendered in 49.734 s
Average framerate  :   74.7 FPS
Minimum framerate  :   72.7 FPS
Maximum framerate  :   75.9 FPS
1% low framerate   :   46.3 FPS
0.1% low framerate :   31.8 FPS

 

HardwareMonitoring.zip

Edited by sunnyB
Link to post
Share on other sites
6 hours ago, sunnyB said:

forum settings, GPU core +140, capped at panel refresh rate for reasons unkown target fps was set to off
07-12-2020, 18:11:37 Il-2.exe benchmark completed, 3720 frames rendered in 49.734 s
Average framerate  :   74.7 FPS
Minimum framerate  :   72.7 FPS
Maximum framerate  :   75.9 FPS
1% low framerate   :   46.3 FPS
0.1% low framerate :   31.8 FPS

 

Wow!  I see you had a good testing session. The system you have 5900X+3090+Pimax8KX is the most extreme case in all senses, CPU power, GPU power and VR performance demand (pixels and FOV).

The tests performed with the Pimax8KX (and other Pimaxes) can not be compared apple to apple with other VR devices. The first reason is the FOV. Even with the Normal FOV, the number of objects that are included in the view is much higher than in a regular VR headset or 16:9 monitor, so the CPU has to draw the scene of those objects and this load more the CPU.

In the other hand, the native refresh rate of the Pimax8KX (not upscaling) is 75Hz, so the secene has to be re-drawn less frequently requiring less CPU load and GPU load.

 

Since your frequency is 75Hz you will never see fps higher than 75. This also happen with all VR devices. The fps are capped to the panel frequency (either 75, 80, 90 or whatever is avaiable in the options of the VR device). So you run is OK and is telling us that a 3090 is capable of running the Pimax8KX (78%SS) with 9.64 million pixels at 75Hz. This is something expected for a 3090 as seen with others devices.

4 hours ago, sunnyB said:

second run without OSD overlay
07-12-2020, 15:03:58 Il-2.exe benchmark completed, 6054 frames rendered in 62.750 s
Average framerate  :   96.4 FPS

 

Thanks. This is an interesting comparison. The VM thing, based on my experience with other CPU single-core intensive applications, has a cost of about 20% of the performance. That´s why I always prefer to avoid VMs although they are very handy.

Link to post
Share on other sites
9 hours ago, chiliwili69 said:

The previous G1 uses a different types of lenses. They suffered from an small Sweet spot and also chromatic aberrations in the edges. 

Here the G2 uses the same lense technology that Valve patented.

This lenses are used also in the Index, and the Index needed a high target multiplier x1.4 over the native display. That´s why Index physical display is 1440x1600 and 100%SS is 2016x2240.

So you can not extrapolate numbers of the G1 over the G2. They use totally different technology although the native panel has the same resolution 2160x2160. (Also native panels of G1 and G2 are not the same models).

According to HP it seems that the G2 need a x1.5 multiplier, this should should be around 105%SS in the G2.

 

Hmmm. I thought the main reason the Index needs lots more pixels rendered is due to the canted displays and the fact that most applications assume parallel projection (screens flat, not canted). Maybe there are some extra pixels there for the distortion correction too, then.

 

This is disappointing. I thought I was going to be able to easily power a G2 based on the fact that it's less physical pixels than the Index's inflated render target. I'll go set the thing to 19.5 million pixels and see how it performs. What I will say is that players running that other simulator are going to be screwed with a 3080 10GB, you will run out of framebuffer instantly. You'll need a 16GB AMD card or a 3090 24GB or a (hypothetical) 3080ti 20GB to run it in multiplayer.

Link to post
Share on other sites
21 minutes ago, Alonzo said:

Hmmm. I thought the main reason the Index needs lots more pixels rendered is due to the canted displays and the fact that most applications assume parallel projection (screens flat, not canted). Maybe there are some extra pixels there for the distortion correction too, then.

 

Others previous VR headsets with no canted displays like Rift CV1 or VivePro has to render more pixels than physical display. This has been like that from the very begining.

In fact, only few of them didn´t do that, like the G1. But in the G1 the edges get blurry (small sweet spot).

You will see in yellow the "target multiplier" which is the multiplication factor that every device applies to each axis to get the rendered pixels at 100%SS.

 

So the 1.45 factor used by the G2 is not too far from the 1.4 factor used by Vive, VivePro and Index.

target.thumb.png.28670923f790be135e0fb26eb299ecf4.png

 

target2.thumb.png.81c5cc877b76795e0fe9ade73eafafad.png

 

Link to post
Share on other sites
11 hours ago, chiliwili69 said:

 

Something is not correct in this test, the Time shoud show 60000ms, not 56062. I don´t know why. (The P to unpasue IL-2 should be the same than the P to make fraps to start the 60 seconds test). Your previous VR test showed 60000ms.

 

According to HP it seems that the G2 need a x1.5 multiplier, this should should be around 105%SS in the G2.
 

@SCG_Fenris_Wolf requested:

Awesome, can you repeat the test using 80% supersampling please?


@chiliwili69 explained:

He refers to touch only the General/Video settings, from 50% to 80% (or also 100%). So the per aaplication settings remains always untouched at 100%.

 

Here are my new VR tests:


VR test with Reverb G2 at 90Hz, 50% SteamVR (2236 x 2188 x 2 = 9,784,736 pixels) in SteamVR Settings -> General tab (and repeated on Video tab) "Resolution per eye".
2020-12-07 12:59:53 - Il-2
Frames: 4738 - Time: 60000ms - Avg: 78.967 - Min: 60 - Max: 91

 

VR test with Reverb G2 at 90Hz, 80% SteamVR (2828 x 2768 x 2 = 15,655,808 pixels) in SteamVR Settings -> General tab (and repeated on Video tab) "Resolution per eye".
2020-12-07 12:49:13 - Il-2
Frames: 4458 - Time: 60000ms - Avg: 74.300 - Min: 55 - Max: 91

 

VR test with Reverb G2 at 90Hz, 100% SteamVR (3164 x 3092 x 2 = 19,566,176 pixels) in SteamVR Settings -> General tab (and repeated on Video tab) "Resolution per eye".
2020-12-07 12:40:40 - Il-2
Frames: 4507 - Time: 60000ms - Avg: 75.117 - Min: 54 - Max: 91

 

For all the above SteamVR Settings -> Video -> Per-Application Video Settings -> "Custom Resolution Multiplier" set at 100%.

 

My computer specs again, unchanged from earlier:

 

Motherboard: MSI MPG Z490 Gaming Carbon WIFI
CPU: i7 10700K
CPU Freq: 5.1 Ghz
L3 cache: 16 MB
Cores: 8
Threads: 8
RAM type: DDR4
RAM size: 32Gb (4x8GB)
Uncore Frequency: 4801.2 MHz
RAM Freq: 4000 MHz
RAM timings: 17-17-17-37
GPU: EVGA RTX 3090 FTW3 Ultra

  • Like 1
  • Thanks 1
Link to post
Share on other sites

Awesome, thank you.

 

The 50% is an outlier, because 80 and 100 are identical results and imply a CPU bottleneck, and not a GPU bottleneck.

 

Could you retest 50% once more please, then we can confirm that.

 

 

@Chiliwili69, we may have to modify the test. The VR test bottlenecks the CPUs first. Which means that any test in which the CPU is bottleneck, we cannot make any assumption about the GPU. We need a vr CPU and a VR GPU one, to differentiate.

Link to post
Share on other sites

OK, here's a new test:

 

VR test with Reverb G2 at 90Hz, 50% SteamVR (2236 x 2188 x 2 = 9,784,736 pixels) in SteamVR Settings -> General tab (and repeated on Video tab) "Resolution per eye".

2020-12-08 00:53:41 - Il-2
Frames: 4447 - Time: 60000ms - Avg: 74.117 - Min: 55 - Max: 92

 

As before, the above SteamVR Settings -> Video -> Per-Application Video Settings -> "Custom Resolution Multiplier" set at 100%.

Link to post
Share on other sites
22 hours ago, Charlo-VR said:

VR test with Reverb G2 at 90Hz, 100% SteamVR (3164 x 3092 x 2 = 19,566,176 pixels) in SteamVR Settings -> General tab (and repeated on Video tab) "Resolution per eye".
2020-12-07 12:40:40 - Il-2
Frames: 4507 - Time: 60000ms - Avg: 75.117 - Min: 54 - Max: 91

Thank you for this test. This means that a 3090 is able to handle 19.5 Mpixels at at least 75fps, which is a good information. Let´s see if other people can also run the VR test with 19.5 million pixels and try other GPUs and CPUs.

Link to post
Share on other sites
2 minutes ago, chiliwili69 said:

Thank you for this test. This means that a 3090 is able to handle 19.5 Mpixels at at least 75fps, which is a good information. Let´s see if other people can also run the VR test with 19.5 million pixels and try other GPUs and CPUs.

 

Just fyi I think 100% resolution may differ from rig to rig, depending on the hardware.

For instance for me 100% Steam VR is 3172x3100.

Link to post
Share on other sites
12 hours ago, SCG_Fenris_Wolf said:

@Chiliwili69, we may have to modify the test. The VR test bottlenecks the CPUs first. Which means that any test in which the CPU is bottleneck, we cannot make any assumption about the GPU. We need a vr CPU and a VR GPU one, to differentiate.

 

Initially we were runing the VR test at Low PReset but we were achieving high numbers, so we decided to run it with High settings to load the CPU a little bit more.

I think the VR test done with 9.5 millionpixels is good to test the CPU performance in VR. You for example were very close to 90.

 

For the new VRtest2 with 19.5 millions we try to see if the GPU is capable to manage those pixels. In the case of Charlo (and 10700K at 5.1) we were bounded by the CPU at 74fps.

Perhaps, for the VRTest2, we could agree to run it at Low Preset, to unload the CPU a bit. What do you think?

 

In general it is a bit difficult to make VR tests for all VR devices, CPU and GPUs without hitting the boundaries (45-90). That´s why I think it is better to look at the monitor tests in 1080p and 4K. But the problem is that not all people have 4K available. Whatever CPU or GPU is good in monitor will be good in VR.

Link to post
Share on other sites

ok, i managed to upgrade, now i have 5600x. Re-run the tests, steam vr set to 222%:

 

Monitor:

2020-12-08 21:03:24 - Il-2
Frames: 6871 - Time: 60000ms - Avg: 114.517 - Min: 100 - Max: 155

 

Oculus Rift CV1

2020-12-08 21:16:28 - Il-2
Frames: 2670 - Time: 60000ms - Avg: 44.500 - Min: 38 - Max: 53

Link to post
Share on other sites
7 minutes ago, dburne said:

Just fyi I think 100% resolution may differ from rig to rig, depending on the hardware.

For instance for me 100% Steam VR is 3172x3100.

 

Yes, I was saying that here: https://forum.il2sturmovik.com/topic/66924-syn_vander-benchmark-v6-to-measure-il-2-performance-in-monitor-vr/?do=findComment&comment=1033201

 

There is very small variations in the total number of pixels.

 

For Charlo: 3164 x 3092 x 2 = 19,566,176 pixels

For Don: 3172x3100x2= 19,666,400 pixels

 

This small difference (it is 0.5%) will not affect significantly the VRTest2 results.

 

If you or Charlo have time, Could you run the VRTest2 with High and Low presets (all other settings like in VRTest2)?

Link to post
Share on other sites

 

Before I start @chiliwili69

In my 4770K run in the spreadsheet there is information on Cooler, GPU and RAM Model which is not from me.

 

Finally I can post some results. It took me quite some time to settle with the following and I tried a few things. I kept the CPU to PBO = Auto, but did some DRAM OC. The GPU is probably to the max of what it can give:

Motherboard:      MSI B550 Gaming Carbon Wifi
CPU:                      R5 5600X
CPU Freq:             max 4.65 Ghz - PBO Auto
L3 cache:             32 MB
Cores:                  6
Threads:             12
RAM type:          DDR4
RAM size:          32Gb (2x16GB)
NB Freq:            1900 MHz (or Uncore Frequency or Infinity Fabric freq),  1.050V
RAM Freq:         3800 MHz Dual Channel
RAM timings:    16-16-16-32 , Ryzen DRAM Calculator Fast values
GPU:                   1080Ti - MSI Afterburner OC Curve 2050Mhz - Memory 5961Mhz

 

Optional:

CPU Cooler: Stock Wraith Prism (Some QA didn't catch the missing screws with my Arctic Freezer).
RAM Model: G-SKILL Trident Z Neo F4-3600C16D-32GTZN, IF 1900,
GPU Model: Gigabyte AORUS GTX 1080 Ti Xtreme Edition 11G

 

CPU:

2020-12-08 21:27:20 - Il-2
Frames: 7028 - Time: 60000ms - Avg: 117.133 - Min: 101 - Max: 157

 

VR (Index):

2020-12-08 21:32:45 - Il-2
Frames: 5109 - Time: 60000ms - Avg: 85.150 - Min: 64 - Max: 91

 

General Remark: Don't forget to turn off the anti-virus. Malwarebyte eats 10 FPS off the VR Test average . I turned off all the OC, to finally remember what already happened to me in DCS. Well, in DCS it was intermediate micro-freezes, here it just slows the overall game down.

 

I would have liked a better score, when I see what @LF_Gallahad has reached, but the stock settings on the 5600X gives me good thermal and performance balance, for the moment (around 76C). I will play around with the CPU when I get the missing screws for my cooler, I had to use the stock wraith cooler for now ( which IMHO, is probably plenty enough for the 5600X, especially if you have no plans to go for an all core OC).

The IF and DRAM OC showed overall score improvement in cinebench, but the difference in Il2 wasn't significant enough. The system is stable an the DIMMs temps are good, so I keep it that way.

The GPU is pushed to its maximum, the Afterburner OC feature really has calculated a  good curve where the clock never dips under 2000Mhz, which I can't reach with only setting the offset. The GPU memory speed is the sweet spot, more is possible, but didn't give more fps in VRMark (Cyan  Room), even less.

 

What I think about the Ryzen 5600X: I didn't like  all core OC, because you have to push the voltage to reach 4.5 and it is always all cores. I came to > 90 degrees, which AMD has said, is perfectly ok, but I don't feel comfortable with it. What is splendid is the capability of the Precision Boost ( not the PBO , it is not the same thing) to only boost a few cores, and for a light threaded game like Il2, that is perfect. It keeps the thermals in a very decent range, you notice this when you do a single core run in cinebench.

I wasn't a friend of PBO enabled, as it also takes the thermals up, but I must admit it gives a few FPS more. Any PBO setting (+100 . +200) where the voltage control is on auto, goes up  way over 1.4V and produces more heat.

The new settings with the PBO scalar (is that already the announced PBO2 ?), which seems to basically enable a better distributed PBO of selected cores (like PBO = auto) , given you put enough voltage on it (or keep the VC on auto), seems the more promising configuration. The same behavior as plain PB , but with the freedom to push it to the thermal limit. Definitely better for light threaded usage than all core OC. I will probably test this when I mount the other cooler.

 

In VR results I had runs with this result:

2020-12-06 23:27:31 - Il-2
Frames: 5177 - Time: 60000ms - Avg: 86.283 - Min: 72 - Max: 91

but I don't recall what the CPU and GPU settings were, I just found it in my FRAPS log file , so there is definitely more to get.

 

 

For now I want to fly and not benchmark anymore :). I have spend 90% of my time with the new system installing and tweaking, I need a break.

Edited by hellfirelx
Somehow posted before finishing to write the whole report
  • Like 3
Link to post
Share on other sites
11 hours ago, JimmySolarium said:

Avg: 114.517

 

Good info. We see that only by upgrading from 3600X to 5600X CPU (6 cores) you got +35fps in the CPU test. Being over other 9900K overclocked processor.

 

I suposse to left oll other settings the same. If this is the case your Infinity Fabric freq is 1600. For the 5600X you should aim to reach 1800 and RAM 3600 or 1900 with RAM 3800.

 

Regarding VR, you current 2070S is more than engouh for the RfitCV1. But for the G2 you will be in the reprojection zone even for 50%SS.(9.5 Mpixels).

 

I suposse that with the new CPU you have seen a good performance increase (or play at higher settings) in your regular IL-2 VR with more moderate SS (PD=1.2 for example)

11 hours ago, hellfirelx said:

CPU:

2020-12-08 21:27:20 - Il-2
Frames: 7028 - Time: 60000ms - Avg: 117.133 - Min: 101 - Max: 157

 

Many thanks. That´s good result given that you only used the stock cooler and all cores OC with PBO.

I am sure that when you put the better cooler your numbers will reach other 5600X.

 

On the VR test I think the the 85 or 86 fps values are CPU limited mostly, but 85 is a very good result!. My 1080Ti is not fully loaded even at 130%SS with the Index.

So you VR result should also increse a bit with the better cooling.

 

 

11 hours ago, hellfirelx said:

What is splendid is the capability of the Precision Boost ( not the PBO , it is not the same thing) to only boost a few cores, and for a light threaded game like Il2, that is perfect. It keeps the thermals in a very decent range, you notice this when you do a single core run in cinebench.

 

You could also try to disable the Simultaneous Multi Threading (SMT) item and have only 6 physical cores running. Just to try.

 

And with the PRecision Boost, how do you force IL-2 to run only on the OC cores?

 

In any case, with the better CPU cooler you will be quite OK running the 6 cores at +200.

 

Another option is to disable 2 cores and run only with 4 cores physical. But I don´t know if this will slow down the overall performance since many other processes are running as well.  Just try. But not hurry, enjoy your PC first. We tweak for the purpose of playing IL2-VR, not for pleasure. 😉 

  • Thanks 1
Link to post
Share on other sites

BTW, I made yesterday some test varying the Index SS values to know to waht extend the CPU was limiting the performance.

This is the results using the same setting of the VRTest but just changing the SS values:

 

Index at 90Hz
20% (900x1000) 1.8 Mpixels
Frames: 2819 - Time: 60000ms - Avg: 46.983 - Min: 43 - Max: 61
Frames: 2844 - Time: 60000ms - Avg: 47.400 - Min: 43 - Max: 58

 

52% (1452x1612) 4.7 Mpixels
Frames: 2680 - Time: 60000ms - Avg: 44.667 - Min: 42 - Max: 51
Frames: 2703 - Time: 60000ms - Avg: 45.050 - Min: 41 - Max: 55
Frames: 2685 - Time: 60000ms - Avg: 44.750 - Min: 38 - Max: 49

 

106% (2076x2304) 9.6 Mpixels
Frames: 2654 - Time: 60000ms - Avg: 44.233 - Min: 40 - Max: 47
Frames: 2652 - Time: 60000ms - Avg: 44.200 - Min: 41 - Max: 46
Frames: 2655 - Time: 60000ms - Avg: 44.250 - Min: 41 - Max: 47

 

216% (2960x3292) 19.5 Mpixels
Frames: 2638 - Time: 60000ms - Avg: 43.967 - Min: 38 - Max: 46
Frames: 2661 - Time: 60000ms - Avg: 44.350 - Min: 41 - Max: 46
Frames: 2651 - Time: 60000ms - Avg: 44.183 - Min: 39 - Max: 46

 

The view with 52% really bad (note that this is the same than physical display), many jaggies and pixelated image. The view at 20% was like going to the 80´s arcades.

 

Link to post
Share on other sites

IL2 4.503

 

Motherboard: SABERTOOTH Z97 MARK 2

CPU: I5 4690K 4.5Ghz

L3 Cache: 6MB

Cores: 4

Threads: 4

Ram Type: DDR3 1600Mhz

NB Freq: 800

Memory Frequency        799.9 MHz (1:6)
CAS# latency (CL)        10.0
RAS# to CAS# delay (tRCD)    10
RAS# Precharge (tRP)        10
Cycle Time (tRAS)        30

 

GPU: Geforce 2060 6 Gb


2020-12-09 15:12:53 - Il-2
Frames: 4298 - Time: 60000ms - Avg: 71.633 - Min: 58 - Max: 97

 

Edited by Norz
Link to post
Share on other sites
9 hours ago, chiliwili69 said:

On the VR test I think the the 85 or 86 fps values are CPU limited mostly, but 85 is a very good result!. My 1080Ti is not fully loaded even at 130%SS with the Index.

So you VR result should also increse a bit with the better cooling.

 

 

Yes indeed, during the test the workload was around 45% . For playing, I enabled 4K Textures and low shadows an went down to 80Hz. Then I started with 120% no AA. I still had an average of 76fps, still had all asw or smoothing measures off.  The dropped frames were not noticeable to me and of quite low count. I thought, I still miss some AA, lets push to 200%%, which is near G2 territory. the gpu finally came to a 80% usage and the fps or frametimes were still intact. Of course this is in a simple duel quick mission. The most taxing to my understanding are the scripted campaigns, and for the moment  I didn't have time to go back into online flying. I will do the vr test2 just to see where I end up. I am one of those guys who cancelled his G2 order because I was on the amazon.es disaster with a pre-order from late september. l'll reconsider the 2160p headsets earliest in a year I guess..

 

 

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...