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

January 8, 2021

21 hours ago, hellfirelx said:

The VRTest1 is the same, the 1080ti was really a beast.

The VRTest2 proves the 3080 is an improvement. That is basically the real win (besides RTX), you can push everything a bit higher and finish off by putting SS or AA as high as you can approaching 100% GPU usage.

Congrats or your new 3080 and thanks for re-testing. This is good info.

As you say, CPU test doesn´t change and VR TEst1 neither. But again, and luckely for you, you really achieve a high number for your 5600X and you Index.

I have exactly the same CPU and VR device, and I only achieve 77fps ( like 56RAF_Roke or R3animate). So your 87fps result test is really to be studied. How you can achieve that?

There could be two explanations:

- You configure something wherever that give you a nice boost over other 3 of us. If this is the case I really want to know it.

- There were some settings not correctly configured in your VR tests (could you make snapshots of SteamVR settings Video/General) or other hidden IL-2 game settings

I run the VR tests with Mottion Smoothing OFF.

January 8, 2021

I will see into making screenshots, I can also provide the content of the startup.cfg.

As already stated , check the reprojection in fpsVR, I have like the feeling that the 'enable legacy reprojection' setting in SteamVR , under the individual setting for a given application, is acting weird. I could not really find out, if it does anything at all, but if you put it to on, the reprojection is off in fpsVR. And as expected frames are dropped when frametime exceed the limit. You have to restart Il2 when you change this.

When your system can hold 70-90 fps it feels fine and smooth, even with sporadically dropped frames, but if your framerate drops too much , you might and probably will get vr sickness, which is the consequence of having frames missing.

Edit: I tried to gather info about reprojection, this applies to index and steamVR, but for other systems the same concepts are there with different names.

Motion smoothing , which is disabled for the benchmark is the ASW equivalent: Asynchronous Spacewarp.

Asychronous reprojection, which is the equivalent of asychronous timewarp, is always enabled. That's why fpsVR shows a reprojection ratio, even with motion smoothing disabled, it is not the same thing.

If I am right, legacy reprojection mode, was the old 'motion smoothing' called interleaved reprojection.

Now there is the option to override and turn off all reprojection of the compositor in fpsVR. I tried that and reprojection ratio is then disabled, and you have more dropped frames.

Furthermore I noticed that when I do that in fpsVR it basically changes the legacy reprojection toggle in SteamVR. The effect is the same when this toggle is enabled, fpsVR greys out reprojection ratio. So I concluded that this should be proof any reprojection is disabled.

Questions:

- May be there is interleaved reprojection going on, and fpsVR does not pick this up and it gives me a higher minimum framerate.

- May be it does effectively turn off the actual always on reprojection algorithm whatsoever, and the interleave mode is not kicking in or at last not for il2 ?

I am pretty sure this explains the fact that my minimum fps doesn't drop as low as yours @chiliwili69 or others, thus resulting in a higher average.

Also, Fenris and the other top liners have similar results as mine and I am pretty sure it doesn't come from their superior CPUs as this is exactly what the benchmark results show: Better framerates on higher resolutions rely almost only on GPU capacity.

Edited January 10, 2021 by hellfirelx
Additional info to reprojection

January 9, 2021

On 12/17/2020 at 5:16 PM, Voyager said:

Retest of system with 4.504

CPU Test
Terrain Roughness Off

2020-12-17 16:43:40 - Il-2
Frames: 5111 - Time: 60000ms - Avg: 85.183 - Min: 74 - Max: 117

Terrain Roughness High

2020-12-17 16:38:55 - Il-2
Frames: 5072 - Time: 60000ms - Avg: 84.533 - Min: 73 - Max: 117

Pseudo GPU Test (DQHD, 5120x1440p: 7.37m pixels)

Terrain Roughness Off (Only option)

2020-12-17 16:49:51 - Il-2
Frames: 5335 - Time: 60000ms - Avg: 88.917 - Min: 74 - Max: 102

VR Test 1

HP Reverb G1 SS: 100%

Terrain Roughness Off

2020-12-17 16:55:33 - Il-2
Frames: 3358 - Time: 60000ms - Avg: 55.967 - Min: 46 - Max: 80

Terrain Roughness High

2020-12-17 16:59:10 - Il-2
Frames: 3325 - Time: 60000ms - Avg: 55.417 - Min: 46 - Max: 77

VR Test 2

HP Reverb G1 SS: 206%

Terrain Roughness Off

2020-12-17 17:04:56 - Il-2
Frames: 2857 - Time: 60000ms - Avg: 47.617 - Min: 42 - Max: 57

Terrain Roughness High

2020-12-17 17:11:03 - Il-2
Frames: 2868 - Time: 60000ms - Avg: 47.800 - Min: 41 - Max: 54

Overall the test results for 4.504 look, to me, to be comparable with the 4.503 results, and setting the Terrain Roughness to High has minimum impacts on framerates for this test.

Pre-upgrade rerun of benchmark test:

Hardware

Motherboard: Gigabyte X570 Aorus Pro Wifi 1.0

Bios Version F31
CPU:       AMD Ryzen 7 3800X
CPU Freq:   4450 MHz
L3 cache:   2x16 MB
Cores:       8
Threads:   16
RAM type:   DDR4
RAM size:   64 GB (4x16)
NB Freq:   1796.4 MHz
RAM Freq:   1796.4
RAM timings:   16-19-19-39-58
GPU:        GTX 1080 Ti

Driver Version 461.09

VR Headset: HP Reverb G1

Game Version 4.505

CPU Test

2021-01-09 16:00:32 - Il-2
Frames: 4762 - Time: 60000ms - Avg: 79.367 - Min: 68 - Max: 106

Pseudo GPU Test (DQHD, 5120x1440p: 7.37m pixels)
2021-01-09 16:07:20 - Il-2
Frames: 5319 - Time: 60000ms - Avg: 88.650 - Min: 72 - Max: 101

VR Test 1

2021-01-09 16:13:15 - Il-2
Frames: 3168 - Time: 60000ms - Avg: 52.800 - Min: 43 - Max: 65

VR Test 2

2021-01-09 16:17:48 - Il-2
Frames: 2742 - Time: 60000ms - Avg: 45.700 - Min: 39 - Max: 52

January 10, 2021

I tinkered a bit with the memory today and I dropped the latency quite a bit. My 3600Mhz Gskill Neo came with 16-16-16-36 timings and i used to get just over 72ns latency in Aida 64.

I kept the 3600Mhz but I managed to bring the timings down to 14-14-15-14-30 after boosting the power from 1.35v to 1.4v, and that alone dropped the latency to 68.2 ns. Considering how crappy the latency of Zen2 is, how crap I am at memory overclocking, I think I did ok. ?

I ran the benchmark twice to see if it made any difference and sure enough, it gave me five frames average.

Everything else is the same as before.

Motherboard: Gigabyte X570 Aorus Master

CPU: Ryzen 9 3950X

CPU frequency: 4.7GHz

L3 Cache: 4x16MB

Cores 16

Threads 32

RAM type DDR4

Ram size 32GB (2x16GB)

Ram frequency 3600MHz

CAS 14

NB frequency 1800MHz

GPU Gigabyte Aorus 1080Ti (factory overclocked)

CPU test:

2021-01-09 20:46:32 - Il-2
Frames: 5573 - Time: 60000ms - Avg: 92.883 - Min: 84 - Max: 128

2021-01-09 20:29:18 - Il-2
Frames: 5520 - Time: 60000ms - Avg: 92.000 - Min: 82 - Max: 131

Hope i can get my hands on the replacement CPU soon. I'm getting tired of searching for it.

Edited January 10, 2021 by Jaws2002

January 10, 2021

Hi all...ran the benchmarking tests today in game version 4.505c

All PC details should be in my sign below...

For some reason when i do the Memory Latency test with AIDA64...i dont get a reading...all i get is the below (0.0ns)

Results below with description

Test 1 monitor using speaker
2021-01-10 09:30:35 - Il-2
Frames: 5255 - Time: 60000ms - Avg: 87.583 - Min: 76 - Max: 120
Test 1 monitor using USB headphones
2021-01-10 09:34:40 - Il-2
Frames: 5096 - Time: 60000ms - Avg: 84.933 - Min: 74 - Max: 113

VR1 Test using USB headphones
2021-01-10 09:50:45 - Il-2
Frames: 3206 - Time: 60000ms - Avg: 53.433 - Min: 44 - Max: 68
VR2 Test using USB headphones
2021-01-10 09:55:14 - Il-2
Frames: 3037 - Time: 60000ms - Avg: 50.617 - Min: 44 - Max: 59

Im happy with these results....i then switched back to the way I had been running my VR with Game - Map scenery distance = Unlimited

and the below graphics settings and resolution at 70pc in Steam VR (1848 x 1808)

vr normal 70pc settings + using USB headphones
2021-01-10 10:04:37 - Il-2
Frames: 2683 - Time: 60000ms - Avg: 44.717 - Min: 40 - Max: 47

Then i played with the resolution slider with my go to graphic settings above and settled on 110pc (2316 x 2218) - all NVIDIA control panel settings

at application controlled except for 'Prefer Maximum Performance' setting and i got the following

VR norm + 110pc + NVID def + Max Perf + using USB headphones
2021-01-10 10:31:55 - Il-2
Frames: 2405 - Time: 60000ms - Avg: 40.083 - Min: 33 - Max: 44

I am quite happy with this as i get no stutters and it looks amazing...curious to see how it performs on the MP servers.

What i dont understand out of all this is the DRam frequency figure of 1066.9 MHz...is that very low?

Can any computer savy people here read my other specs and identify where i may be able to squeeze more performance out of my rig...

As always, all help very much appreciated...cheers lads

Edited January 10, 2021 by Tonester

January 10, 2021

1 hour ago, Tonester said:

For some reason when i do the Memory Latency test with AIDA64...i dont get a reading...all i get is the below (0.0ns)

I don't know why that is, but from the screenshot above this line, i can tell that your memory XMP profile is disabled in BIOS.

In your sig your memory is a 3600Mhz kit, and in CPUz it shows 1066Mhz. It should be 1800Mhz, if your XMP was enabled.

January 10, 2021

6 hours ago, Jaws2002 said:

I tinkered a bit with the memory today and I dropped the latency quite a bit. My 3600Mhz Gskill Neo came with 16-16-16-36 timings and i used to get just over 72ns latency in Aida 64.

I kept the 3600Mhz but I managed to bring the timings down to 14-14-15-14-30 after boosting the power from 1.35v to 1.4v, and that alone dropped the latency to 68.2 ns. Considering how crappy the latency of Zen2 is, how crap I am at memory overclocking, I think I did ok. ?

I ran the benchmark twice to see if it made any difference and sure enough, it gave me five frames average.

Everything else is the same as before.

For reduced latency in Ryzen CPUs make sure Gear Down Mode is set to disabled in BIOS. Also make sure you are set to 1T and not auto or 2T. It should be possible to get those latencies down 10%.

January 10, 2021

8 hours ago, Jaws2002 said:

I ran the benchmark twice to see if it made any difference and sure enough, it gave me five frames average.

Many thanks for this test! and providing AIDA latency. I think this is the key factor we should follow, fpr Zen systems it should correlate well with fps.

This is the kind of things I want to try with my 5600X. How did you change timings? with RAM calculator?

8 hours ago, Tonester said:

Test 1 monitor using speaker
2021-01-10 09:30:35 - Il-2
Frames: 5255 - Time: 60000ms - Avg: 87.583 - Min: 76 - Max: 120
Test 1 monitor using USB headphones
2021-01-10 09:34:40 - Il-2
Frames: 5096 - Time: 60000ms - Avg: 84.933 - Min: 74 - Max: 113

Thanks you for your test. I do notice your RAM speed is 2133MHz, but your RAM model is 3600. Did you try to activate the XMP profile of your RAM in the BIOS?

This should give you a significant boost.

It is also very strange that you get different results with just speaker or USB headphones.

6 hours ago, Jaws2002 said:

I don't know why that is, but from the screenshot above this line, i can tell that your memory XMP profile is disabled in BIOS.

In your sig your memory is a 3600Mhz kit, and in CPUz it shows 1066Mhz. It should be 1800Mhz, if your XMP was enabled.

yeah, you look it first. ?

January 10, 2021

8 hours ago, ICDP said:

For reduced latency in Ryzen CPUs make sure Gear Down Mode is set to disabled in BIOS. Also make sure you are set to 1T and not auto or 2T. It should be possible to get those latencies down 10%

6 hours ago, chiliwili69 said:

How did you change timings? with RAM calculator?

I did this with the help of the newer version of DRAM calculator. In the versuon 1.63 i had, the best settings for my kit, at 3600mhz, were 16-16-16-32.

I installed version 1.7.3 and i got this new settings.

I used the settings in the safe profile for now and the safe profile has "gear down mode" to enabled, so i used that.

The comand rate is at 1T, like in the calculator.

In the end it worked but with the DRAM voltage at 1.4v instead of 1.37.

Maybe I'll try the fast profile as well, if i have time. But i'd rather do that with the new CPU.....whenever i can ger one.

EDIT. I tried this profile with "gear down mode" disabled and made no difference.

Edited January 10, 2021 by Jaws2002

January 10, 2021

10 hours ago, chiliwili69 said:

Thanks you for your test. I do notice your RAM speed is 2133MHz, but your RAM model is 3600. Did you try to activate the XMP profile of your RAM in the BIOS?

This should give you a significant boost.

.....

I went into BIOS and had a look at XMP profile..it was set to ‘Auto’...whatever that means...the only other option was ‘Profile 1’..whatever that means...i changed it to Profile 1...saved and exited...and now my PC wont start up....FML...it tries to start but just shuts itself down, then starts again in this endless loop...doesnt even get to the point where the prompt comes up on the screen to go to BIOS, STARTUP etc...infact i get nothing in the screen at all...any ideas?

10 hours ago, chiliwili69 said:

January 10, 2021

15 minutes ago, Tonester said:

I went into BIOS and had a look at XMP profile..it was set to ‘Auto’...whatever that means...the only other option was ‘Profile 1’..whatever that means...i changed it to Profile 1...saved and exited...and now my PC wont start up....FML...it tries to start but just shuts itself down, then starts again in this endless loop...doesnt even get to the point where the prompt comes up on the screen to go to BIOS, STARTUP etc...infact i get nothing in the screen at all...any ideas?

This is why folks need to be very cautious when working with their bios or recommending changes to a bios and have understanding of much of those settings by reading the MB manual and perusing the MB mfg forums.

There should be a way to reset the board to default . Check your MB manual for this or their forums.

Edited January 10, 2021 by dburne

January 10, 2021

51 minutes ago, Tonester said:

infact i get nothing in the screen at all...any ideas?

This kind of crashes are pretty normal when playing with memory settings.

In most cases "clearing CMOS fixes that.

Refer to your motherboard manual about how to clear CMOS.

I found your motherboard manual and here it is:

the jumper i circled in red is the clear CMOS jumper and the image below explains how to do it on your motherboard.

My board has a very convenient button at the top of the rear IO panel for that.

unplug the cable in the power supply, let the power supply button to ON, press the main case power button to drain any residual power in the board, then do the clear CMOS.

In most cases that should fix it. IF is not working, remove and reinsert the memory sticks. and try again.

Edited January 10, 2021 by Jaws2002

January 10, 2021

23 hours ago, Voyager said:

Pre-upgrade rerun of benchmark test:

Hardware

Motherboard:   Gigabyte X570 Aorus Pro Wifi 1.0

Bios Version F31
CPU:       AMD Ryzen 7 3800X
CPU Freq:   4450 MHz
L3 cache:   2x16 MB
Cores:       8
Threads:   16
RAM type:   DDR4
RAM size:   64 GB (4x16)
NB Freq:   1796.4 MHz
RAM Freq:   1796.4
RAM timings:   16-19-19-39-58
GPU:        GTX 1080 Ti

Driver Version 461.09

VR Headset: HP Reverb G1

Game Version 4.505

CPU Test

2021-01-09 16:00:32 - Il-2
Frames: 4762 - Time: 60000ms - Avg: 79.367 - Min: 68 - Max: 106

Pseudo GPU Test (DQHD, 5120x1440p: 7.37m pixels)
2021-01-09 16:07:20 - Il-2
Frames: 5319 - Time: 60000ms - Avg: 88.650 - Min: 72 - Max: 101

VR Test 1

2021-01-09 16:13:15 - Il-2
Frames: 3168 - Time: 60000ms - Avg: 52.800 - Min: 43 - Max: 65

VR Test 2

2021-01-09 16:17:48 - Il-2
Frames: 2742 - Time: 60000ms - Avg: 45.700 - Min: 39 - Max: 52

Test with Upgrade:

Hardware

Motherboard: Gigabyte X570 Aorus Pro Wifi 1.0

Bios Version F31
CPU:       AMD Ryzen 7 5800X
CPU Freq:   4450 MHz
L3 cache:   2x16 MB
Cores:       8
Threads:   16
RAM type:   DDR4
RAM size:   32 GB (2x16)
NB Freq:   1796.4 MHz
RAM Freq:   1796.4
RAM timings:   16-19-19-39-58
GPU:        GTX 1080 Ti

Driver Version 461.09

VR Headset: HP Reverb G1

Game Version 4.505

CPU Test

2021-01-10 16:18:54 - Il-2
Frames: 6854 - Time: 60000ms - Avg: 114.233 - Min: 99 - Max: 155

Pseudo GPU Test (DQHD, 5120x1440p: 7.37m pixels)
2021-01-10 16:23:39 - Il-2
Frames: 5552 - Time: 60000ms - Avg: 92.533 - Min: 74 - Max: 108

VR Test 1

2021-01-10 16:07:54 - Il-2
Frames: 4373 - Time: 60000ms - Avg: 72.883 - Min: 55 - Max: 91

VR Test 2

2021-01-10 16:12:30 - Il-2
Frames: 2706 - Time: 60000ms - Avg: 45.100 - Min: 41 - Max: 49

Note that I had to remove two of my ram sticks to get it to run at DDR4-3600 speeds. All four sticks run at the default DDR4 speeds. I haven't decided entirely what I want to do about that yet.

January 10, 2021

On 1/9/2021 at 12:53 AM, hellfirelx said:

I am pretty sure this explains the fact that my minimum fps doesn't drop as low as yours @chiliwili69 or others, thus resulting in a higher average.

Hi,

just chiming in on the VRTest topic:

I tested with 'legacy reprojection enabled and disabled. As stated before, the former setting disables the reprojection ratio counter and enables the dropped frames counter in fpsVR, the latter turns it on again and switches the dropped frames counter off. Not sure what to do with it, as we can debate what 'switching any ASW or reprojection off' really means, knowing that usually there is a compensation algorithm running in the background which in most cases is fiddly to disable on any platform. (I had a CV1, that used to be only possible with oculus tray and some sort of keyboard shortcut.)

Here the results for VRTest1 (with 3080):

With legacy reprojection enabled , motion smoothing off.

2021-01-10 22:50:36 - Il-2
Frames: 5310 - Time: 60000ms - Avg: 88.500 - Min: 80 - Max: 92

With legacy reprojection disabled, motion smoothing off.

2021-01-10 22:53:30 - Il-2
Frames: 5228 - Time: 60000ms - Avg: 87.133 - Min: 62 - Max: 91

Then I went back to older benchmarks and found this, which I ran with the 1080ti, where the reprojection must have been on :

2020-12-07 22:43:29 - Il-2
Frames: 4696 - Time: 60000ms - Avg: 78.267 - Min: 50 - Max: 91

2020-12-07 22:49:14 - Il-2
Frames: 4496 - Time: 60000ms - Avg: 74.933 - Min: 44 - Max: 91

You see, that explains it all. Now you can debate which setting would be truly respecting 'no reprojection on', but from a pure gameplay point of view, I keep it so that the reprojection ratio stays off in fpsVR. You probably have to try this for yourself.

As for the benchmark, if you prefer to keep the results consistent, choose which of my results you want to use for the sheet.

January 10, 2021

30 minutes ago, Voyager said:

Test with Upgrade:

Hardware

Motherboard:   Gigabyte X570 Aorus Pro Wifi 1.0

Bios Version F31
CPU:       AMD Ryzen 7 5800X
CPU Freq:   4450 MHz
L3 cache:   2x16 MB
Cores:       8
Threads:   16
RAM type:   DDR4
RAM size:   32 GB (2x16)
NB Freq:   1796.4 MHz
RAM Freq:   1796.4
RAM timings:   16-19-19-39-58
GPU:        GTX 1080 Ti

Driver Version 461.09

VR Headset: HP Reverb G1

Game Version 4.505

CPU Test

2021-01-10 16:18:54 - Il-2
Frames: 6854 - Time: 60000ms - Avg: 114.233 - Min: 99 - Max: 155

Pseudo GPU Test (DQHD, 5120x1440p: 7.37m pixels)
2021-01-10 16:23:39 - Il-2
Frames: 5552 - Time: 60000ms - Avg: 92.533 - Min: 74 - Max: 108

VR Test 1

2021-01-10 16:07:54 - Il-2
Frames: 4373 - Time: 60000ms - Avg: 72.883 - Min: 55 - Max: 91

VR Test 2

2021-01-10 16:12:30 - Il-2
Frames: 2706 - Time: 60000ms - Avg: 45.100 - Min: 41 - Max: 49

Note that I had to remove two of my ram sticks to get it to run at DDR4-3600 speeds. All four sticks run at the default DDR4 speeds. I haven't decided entirely what I want to do about that yet.

Big jump in performance.

Congrats.

January 10, 2021

3 minutes ago, Jaws2002 said:

Big jump in performance.

Congrats.

Thank you, though I did have to give up half my ram to get it. ?

I do think I need to get a new 3D Card to really push things in VR. Not sure if I want to keep stalking Best Buy for a 3080, or hold out to stalk a 3080 TI. Catch is, I'm looking to do DCS as well. I should probably run a bunch of missions and see how it plays in practice. We will see.

January 11, 2021

5 hours ago, Jaws2002 said:

This kind of crashes are pretty normal when playing with memory settings.

In most cases "clearing CMOS fixes that.

Refer to your motherboard manual about how to clear CMOS.

I found your motherboard manual and here it is:

the jumper i circled in red is the clear CMOS jumper and the image below explains how to do it on your motherboard.

My board has a very convenient button at the top of the rear IO panel for that.

unplug the cable in the power supply, let the power supply button to ON, press the main case power button to drain any residual power in the board, then do the clear CMOS.

In most cases that should fix it. IF is not working, remove and reinsert the memory sticks. and try again.

Thanks mate

so i fixed it by clearing cmos...once up and running i then flashed the Bios to 4.40 (from 1.30)...it will now accept XMP profile 1.0 which should be DDR4-3600...but i still see 1066.7 in CPU-Z no matter what i try...have tried Phantom Gaming Tuning utility but cant seem to make a change in there...have tried in the Bios in both EZ and Advanced mode to no avail..i tried intel XTU but that didnt help...i even rang the guy that built my PC but he is no help as he thinks OC is some kind of evil...he said Ripjaw 3600 16Gb is what i installed and if its only showing 1066.7 then thats what it must run at...which i think is a complete cop out and nonsense...there must be a way to make this computer/bios/motherboard/cpu (whichever applies here) make the DRam run at its advertised 3600...surely? Any ideas anyone?

January 11, 2021

Not all kits run at the advertized XMP settings. Some can't do it on all motherboards. But 3600mhz should not be a problem.

And you can tell the guy who built the PC that XMP is factory overclocking and should be covered.

If you pay for a 3600mhz kit you should be able to achieve that.

If you had this computer built recently, and paid for it, the guy who built it should have made sure XMP works before you picked it up.

January 11, 2021

On 1/11/2021 at 4:52 AM, Tonester said:

...there must be a way to make this computer/bios/motherboard/cpu (whichever applies here) make the DRam run at its advertised 3600...surely? Any ideas anyone?

+1, that the builder has to take responsibility for XMP check and compatibility , as without it no recent system runs the 3000+ kits at intented speed.

But on a constructive note :

try profile #2 , the difference is not always apparent in the timings, but it's a simple check, so do it.

try to set the dram parameters manually, eventually try reducing the speed to 3200 or less to see where the limit is.

check compatibility of RAM on motherboard or ram vendor site, worst case, is switch to another pair of ram, as sometimes, it just won't work.

Edited January 12, 2021 by hellfirelx
Typo

January 13, 2021

I finally managed to get a 5950x. I had to pay a bit more than I wanted, but It's here and it rocks. ?

I did a quick test of the infinity fabric and i got it to 1933 Mhz and then it crapped out. I'm happy with this for now, because i can run it reliably at 1900Mhz and the memory at 3800.

I inserted the timings from the DRAM calculator for 3800Mhz and sure enough it runs at a clean CL16. At least don't need to buy new memory. The kit I bought last year it's still pretty awesome. I got a 3600 Mhz, CL16 kit (16-16-16-36) and it runs perfectly fine at 3800Mhz CL16 (16-16-16-32)

The CPU is hitting 5.05 GHz on a few cores, so I'm happy. In the game it's a solids jump from my 3950X. My memory latency, running at 3800Mhz cl16, with the new chip, dropped to 58.2 ns. The same kit, running at 3600Mhz, CL16, on the 3950x had 72ns latency. This is quite a jump.

This was done with the CPU stock. Something is buggy with PBO and I'll try to figure it out later.

Motherboard: Gigabyte X570 Aorus Master

CPU: Ryzen 9 5950X

CPU frequency: 5.05GHz

L3 Cache: 64 MB

Cores 16

Threads 32

RAM type DDR4

Ram size 32GB (2x16GB)

Ram frequency 3800MHz

CAS 16 timings 16-16-16-32

NB frequency 1900MHz

GPU Gigabyte Aorus 1080Ti (factory overclocked)

CPU test

2021-01-12 20:32:37 - Il-2
Frames: 7642 - Time: 60000ms - Avg: 127.367 - Min: 112 - Max: 176

2021-01-12 20:34:43 - Il-2
Frames: 7674 - Time: 60000ms - Avg: 127.900 - Min: 110 - Max: 175

2021-01-12 20:36:39 - Il-2
Frames: 7768 - Time: 60000ms - Avg: 129.467 - Min: 112 - Max: 182

2021-01-13 18:28:34 - Il-2
Frames: 7807 - Time: 60000ms - Avg: 130.117 - Min: 115 - Max: 173

2021-01-13 18:32:55 - Il-2
Frames: 7667 - Time: 60000ms - Avg: 127.783 - Min: 110 - Max: 173

Temps during stress test stabilized at 64 degrees.

Now all i need is a 3080Ti and I'll be good.

Edited January 13, 2021 by Jaws2002

January 13, 2021

On 1/10/2021 at 4:29 PM, Voyager said:

Test with Upgrade:

Hardware

Motherboard:   Gigabyte X570 Aorus Pro Wifi 1.0

Bios Version F31
CPU:       AMD Ryzen 7 5800X
CPU Freq:   4450 MHz
L3 cache:   2x16 MB
Cores:       8
Threads:   16
RAM type:   DDR4
RAM size:   32 GB (2x16)
NB Freq:   1796.4 MHz
RAM Freq:   1796.4
RAM timings:   16-19-19-39-58
GPU:        GTX 1080 Ti

Driver Version 461.09

VR Headset: HP Reverb G1

Game Version 4.505

CPU Test

2021-01-10 16:18:54 - Il-2
Frames: 6854 - Time: 60000ms - Avg: 114.233 - Min: 99 - Max: 155

Pseudo GPU Test (DQHD, 5120x1440p: 7.37m pixels)
2021-01-10 16:23:39 - Il-2
Frames: 5552 - Time: 60000ms - Avg: 92.533 - Min: 74 - Max: 108

VR Test 1

2021-01-10 16:07:54 - Il-2
Frames: 4373 - Time: 60000ms - Avg: 72.883 - Min: 55 - Max: 91

VR Test 2

2021-01-10 16:12:30 - Il-2
Frames: 2706 - Time: 60000ms - Avg: 45.100 - Min: 41 - Max: 49

Note that I had to remove two of my ram sticks to get it to run at DDR4-3600 speeds. All four sticks run at the default DDR4 speeds. I haven't decided entirely what I want to do about that yet.

Hardware Test with default speed Ram

Motherboard: Gigabyte X570 Aorus Pro Wifi 1.0

Bios Version F31
CPU:       AMD Ryzen 7 5800X
CPU Freq:   4800 MHz
L3 cache:   2x16 MB
Cores:       8
Threads:   16
RAM type:   DDR4
RAM size:   32 GB (2x16), dual rank
NB Freq: 1064.5
RAM Freq:   1064.5
RAM timings:   15-15-15-36
GPU:        GTX 1080 Ti

Driver Version 461.09

VR Headset: HP Reverb G1

Game Version 4.505

CPU Test

2021-01-13 00:07:19 - Il-2
Frames: 6381 - Time: 60000ms - Avg: 106.350 - Min: 92 - Max: 149

VR Test 1

2021-01-13 00:18:57 - Il-2
Frames: 3608 - Time: 60000ms - Avg: 60.133 - Min: 50 - Max: 83

January 13, 2021

On 1/10/2021 at 11:29 PM, Voyager said:

CPU Test

2021-01-10 16:18:54 - Il-2
Frames: 6854 - Time: 60000ms - Avg: 114.233 - Min: 99 - Max: 155

Nice!. Just by only changing the 3800X by a 5800X you went from 79 fps to 114, +34fps in monitor and +20fps in VR.

On 1/10/2021 at 11:33 PM, hellfirelx said:

2020-12-07 22:43:29 - Il-2
Frames: 4696 - Time: 60000ms - Avg: 78.267 - Min: 50 - Max: 91

2020-12-07 22:49:14 - Il-2
Frames: 4496 - Time: 60000ms - Avg: 74.933 - Min: 44 - Max: 91

Yeah, this results were with your 5600X and the 1080Ti, putting Motionsmotthing OFF and Legacy reprojection disabled. This is fully aligned with my results.

Then I have to try to run the game with Legacy Reprojection Enabled to see if it improves the experience and also run the benchmark with this.

January 13, 2021

On 1/11/2021 at 4:52 AM, Tonester said:

Thanks mate

so i fixed it by clearing cmos...once up and running i then flashed the Bios to 4.40 (from 1.30)...it will now accept XMP profile 1.0 which should be DDR4-3600...but i still see 1066.7 in CPU-Z no matter what i try...have tried Phantom Gaming Tuning utility but cant seem to make a change in there...have tried in the Bios in both EZ and Advanced mode to no avail..i tried intel XTU but that didnt help...i even rang the guy that built my PC but he is no help as he thinks OC is some kind of evil...he said Ripjaw 3600 16Gb is what i installed and if its only showing 1066.7 then thats what it must run at...which i think is a complete cop out and nonsense...there must be a way to make this computer/bios/motherboard/cpu (whichever applies here) make the DRam run at its advertised 3600...surely? Any ideas anyone?

I am glad to see you recover the BIOS by clearing the CMOS, it is a quite rutinary thing for people to play with timming.

But XMP profiles is a very different thing. They should be safe since they has been tested previously. But your Mobo and RAM model should be in the QVL.

https://www.gskill.com/qvl/165/169/1536043466/F4-3600C19D-32GSXKB-Qvl

Which is exactly your model of RAM? Productc code?

Do you have 32Gb, are they 2x16Gb or 4x8Gb?

You can download all XMP certified by Intel in this page: https://www.intel.es/content/www/es/es/gaming/extreme-memory-profile-xmp.html

Edited January 13, 2021 by chiliwili69

January 13, 2021

There was a time when XMP didn't work more than it did, but this days, at least on AMD CPU's and kits below 4000Mhz, it's mostly point and click.

January 13, 2021

9 minutes ago, Jaws2002 said:

There was a time when XMP didn't work more than it did, but this days, at least on AMD CPU's and kits below 4000Mhz, it's mostly point and click.

Yeah same here for my last several Intel boards.

I always try to by ram on the QVL list for the board.

January 14, 2021

21 hours ago, Jaws2002 said:

I finally managed to get a 5950x. I had to pay a bit more than I wanted, but It's here and it rocks. ?

I did a quick test of the infinity fabric and i got it to 1933 Mhz and then it crapped out. I'm happy with this for now, because i can run it reliably at 1900Mhz and the memory at 3800.

I inserted the timings from the DRAM calculator for 3800Mhz and sure enough it runs at a clean CL16. At least don't need to buy new memory. The kit I bought last year it's still pretty awesome. I got a 3600 Mhz, CL16 kit (16-16-16-36) and it runs perfectly fine at 3800Mhz CL16 (16-16-16-32)

Congratulations for your new CPU. Just changing your CPU from Zen2 to Zen3 gave you +41fps in the monitor!

You also are lucky with the RAM! Your CPU is not going to be a limiting factor for your monitor (unless it runs at 144Hz)

January 14, 2021

Thank you! I think the memory and infinity fabric boost also made a big difference. I tried it once with the memory as i had it with the other chip, at XMP and i got 114FPS average. Once I inserted the settings from DRAM calculator, for 3800Mhz and boosted the infinity fabric to 1900Mhz, I got another fifteen FPS.

Now all I need is a 3080Ti, whenever that one is going to show up and I'll be golden.

January 14, 2021

8 hours ago, Jaws2002 said:

Once I inserted the settings from DRAM calculator, for 3800Mhz and boosted the infinity fabric to 1900Mhz, I got another fifteen FPS

Yeah, This is exactly what I would like to achieve with my 5600X and B550 Mobo. Let´s see if one day I can be in the 54ns latency. Other people with 5600X made it.

January 14, 2021

22 hours ago, chiliwili69 said:

On 1/10/2021 at 11:33 PM, hellfirelx said:

2020-12-07 22:43:29 - Il-2
Frames: 4696 - Time: 60000ms - Avg: 78.267 - Min: 50 - Max: 91

2020-12-07 22:49:14 - Il-2
Frames: 4496 - Time: 60000ms - Avg: 74.933 - Min: 44 - Max: 91

Yeah, this results were with your 5600X and the 1080Ti, putting Motionsmotthing OFF and Legacy reprojection disabled. This is fully aligned with my results.

Then I have to try to run the game with Legacy Reprojection Enabled to see if it improves the experience and also run the benchmark with this.

I have just made a the VRTest1 with Legacy Reprojection Enabled in per application settings (And motion smoothing off) and I got 88fpsAvg. Same than you. So this is 10fps more in VR.

I need to understand why? and how this affect the experience.

January 14, 2021

1 hour ago, chiliwili69 said:

I have just made a the VRTest1 with Legacy Reprojection Enabled in per application settings (And motion smoothing off) and I got 88fpsAvg. Same than you. So this is 10fps more in VR.

I need to understand why? and how this affect the experience.

Legacy reprojection enabled seems to be (for me) the only way to properly benchmark. If I don't do that, I get half-FPS mode instead of a properly fluctuating FPS. Maybe I did something wrong.

Visually at least, legacy reprojection looks awful on my Index. Every head movement causes a stutter.

January 14, 2021

46 minutes ago, Alonzo said:

Legacy reprojection enabled seems to be (for me) the only way to properly benchmark. If I don't do that, I get half-FPS mode instead of a properly fluctuating FPS. Maybe I did something wrong.

Visually at least, legacy reprojection looks awful on my Index. Every head movement causes a stutter.

I would think for benching one would just have reprojection off?

January 14, 2021

After a bit of tinkering with PBO.

Motherboard: Gigabyte X570 Aorus Master

CPU: Ryzen 9 5950X

CPU frequency: 5.05GHz

L3 Cache: 64 MB

Cores 16

Threads 32

RAM type DDR4

Ram size 32GB (2x16GB)

Ram frequency 3800MHz

CAS 16 timings 16-16-16-32

NB frequency 1900MHz

GPU Gigabyte Aorus 1080Ti (factory overclocked)

CPU test

2021-01-14 17:11:10 - Il-2
Frames: 7894 - Time: 60000ms - Avg: 131.567 - Min: 119 - Max: 183

2021-01-14 17:13:04 - Il-2
Frames: 7908 - Time: 60000ms - Avg: 131.800 - Min: 117 - Max: 184

2021-01-15 12:50:29 - Il-2
Frames: 7910 - Time: 60000ms - Avg: 131.833 - Min: 116 - Max: 185

2021-01-15 13:05:05 - Il-2
Frames: 7908 - Time: 60000ms - Avg: 131.800 - Min: 114 - Max: 185

Edited January 15, 2021 by Jaws2002

January 14, 2021

So for ram OC, do we think that it is ram throughout or memory latency that is the controlling factor? Was tinkering around with OC my original ram, and while I got it up to 3800, it didn't seem to impact the game performance.

I'll be running some more thorough tests tonight, including bandwidth and latency tests to make sure the OC's haven't actually made it perform worse, but wondering if we have a good idea why some Z3 with 3600/3800 get 10-15% more performance than others?

January 14, 2021

I think, in this game memory makes a bigger impact than newer games. it's all about high single core boost and fast, low latency memory.

When you overclock the memory, start with the safe settings from Ryzen DRAM calculator. That's a great tool to get the memory where it can get. This will also depend on your motherboard, because some higher end motherboards have very strong power delivery with more accurate power control.

I go lucky with the safe settings for 3800Mhz for my kit, but it didn't take all the timings for some reason, but sure enough it runs good and gave me no errors after a few hours of memory stress testing.

I'm sure I still have room for improvement with the memory, I just have to read up on it more....I'm just too clueless. ?

Here are the settings for my kit right now. You can see the ones that don't match with the suggested values. This are the "safe" settings for my kit. I tried once the "fast' settings and it told me to pi$$ off. ? There's plenty room for improvement here.

January 15, 2021

8 hours ago, Alonzo said:

Legacy reprojection enabled seems to be (for me) the only way to properly benchmark. If I don't do that, I get half-FPS mode instead of a properly fluctuating FPS. Maybe I did something wrong.

Visually at least, legacy reprojection looks awful on my Index. Every head movement causes a stutter.

7 hours ago, dburne said:

I would think for benching one would just have reprojection off?

The thing is there at least to types of reprojection, temporal and spatial. One is represented by motion smoothing, which can be turned off, the other theoretically is always on.

This article is older, but explains it:

https://uploadvr.com/reprojection-explained/

Especially this table is interesting:

So the one at the top is supposed to be always on, on the index. As I already stated in another post, if you check fpsVR ingame the reprojection ratio is active and fluctuates, if the reprojection is needed. So it is not off, by just turning off motion smoothing.

The interleaved reprojection in the table is the legacy one. When I set that to enable , in fpsVR the reprojection counter is inactive.

I could not verify anywhere if that is a proof ,of reprojection being simply off (may be legacy does not work with il2 ?).

The stutter @Alonzo experiences could come from this, the repro is off, your system starts to drop frames.Although I don't know, do you have an index?

But for some, the frame drop is acceptable or not noticaeble, and we gain some performance by having the non legacy, always on async reprojection off.

I don't know where to find a correct answer to my guesses.

Edited January 15, 2021 by Brandy_

January 15, 2021

On 1/14/2021 at 10:51 PM, Alonzo said:

Legacy reprojection enabled seems to be (for me) the only way to properly benchmark

That´s strange. My fps varies from 45 to 90 when running the benchmark with Legacy reprojection disabled.

I have been testing some QMB 8vs8 with not clouds and with Legacy reprojection enabled and it work quite OK. But I still really don´t know what legacy reprojection really is.

On 1/14/2021 at 11:54 PM, Jaws2002 said:

After a bit of tinkering with PBO

You get now a more consistent results. What did you do with PBO?

January 15, 2021

5 minutes ago, chiliwili69 said:

That´s strange. My fps varies from 45 to 90 when running the benchmark with Legacy reprojection disabled.

I have been testing some QMB 8vs8 with not clouds and with Legacy reprojection enabled and it work quite OK. But I still really don´t know what legacy reprojection really is.

Likely similar to Oculus ATW which preceded ASW.

Older tech.

January 15, 2021

On 1/15/2021 at 12:01 AM, Voyager said:

So for ram OC, do we think that it is ram throughout or memory latency that is the controlling factor?

we only have 5 memory latencies data, but it correlate well with fps:

5 minutes ago, dburne said:

ATW

But ATW is not old tech. It is something which is always enabled in the Oculus devices. The equivalent in SteamVR was Asynchronous Reprojection, and it is still used always in SteamVR.

January 15, 2021

On 1/14/2021 at 4:01 PM, Voyager said:

So for ram OC, do we think that it is ram throughout or memory latency that is the controlling factor?

On the Intel chips it looked like memory frequency was more important than latency. Which is odd, but there you go. It wasn't like latency did nothing, but frequency was more important (e.g. 3600-CL16 was better than 3200-CL14 even though both have same latency).

My theory is that something in the IL2 code is doing some cross-thread communication/locking, and whatever we can do to make that fast helps increase FPS. The switch from the old 3000-series to new 5000-series AMD architecture -- infinity fabric? -- seems to have done really well. I'd guess that whatever runs infinity fabric at its fastest and allows the CPU cores to synchronize across it will be most impactful.

January 16, 2021

31 minutes ago, chiliwili69 said:

But ATW is not old tech. It is something which is always enabled in the Oculus devices. The equivalent in SteamVR was Asynchronous Reprojection, and it is still used always in SteamVR.

I don't know if this is good advice, but I've been following this advice I found at https://forums.x-plane.org/index.php?/forums/topic/231901-hp-reverb-g2-issues-with-stuttering-what-settings-for-steamvr/&tab=comments#comment-2081364 :

"Open the settings file at :
<path-to-steam>\SteamLibrary\steamapps\common\MixedRealityVRDriver\resources\settings\default.vrsettings

Comment out these lines (add the // in front as here):
// "motionReprojectionMode": "none",
// "motionReprojectionIndicatorEnabled": false,"

I *think* that ensures I stay out of Asynchronous Reprojection mode, but all I know is that fpsVR tells me I remain in 80-90 fps in MP with rare stuttering and reasonably easy aircraft IDing with no ghosting. In MP I'd rather have occasionally minor stuttering close to the ground on full MP servers than the 45fps that doesn't allow me to spot and ID as easily.

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

Recommended Posts

chiliwili69

Brandy_Lx

Voyager

Jaws2002

Tonester

Jaws2002

ICDP

chiliwili69

Jaws2002

Tonester

dburne

Jaws2002

Voyager

Brandy_Lx

Jaws2002

Voyager

Tonester

Jaws2002

Brandy_Lx

Jaws2002

Voyager

chiliwili69

chiliwili69

Jaws2002

dburne

chiliwili69

Jaws2002

chiliwili69

chiliwili69

Alonzo

dburne

Jaws2002

Voyager

Jaws2002

Brandy_Lx

chiliwili69

dburne

chiliwili69

Alonzo

Charlo-VRde

Create an account or sign in to comment

Create an account

Sign in

Browse All

Activity