Jump to content

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


chiliwili69
 Share

Recommended Posts

RAAF492SQNOz_Steve
10 hours ago, DBCOOPER011 said:

 

Thanks for the information, your setup looks like its tuned very nicely! I tried running my memory above 3800 Mhz, but kept getting whea errors. I don't think the memory clock on my 5800X can handle those higher frequencies..


I actually sold my 3090 vision a few days back to an individual for a good price, and just received an 3090 FTW3 ultra from EVGA yesterday. Been tinkering with it for a little bit and it seems to be a good card. I just set it up to my 4K monitor and ran passmark and got a little better score then I had with my 3090 vision. I think I'm doing something wrong with passmark, as I'm getting much better performance in 3dmark timespy/port royal but not so much in passmark...

 

 

 

Made a discovery that may help explain your Passmark results.

 

I accidently ran Passmark, with HP Reverb G2 and Steam software still loaded in memory (i.e. sleep mode) and noted a big drop in Passmark scores.

In the screen captures below, the one on the left is with HP Reverb software loaded and scores are down across the board. The one on the right was run just after I shutdown the Reverb software and Steam VR.  Do you have some background programs running that could affect your score?

 

image.thumb.png.c00a9950b7af7eaa59766a3bad83b5c1.png

 

Added in Edit, settings for both runs

Settings

5800X Settings:

  • PBO Enabled with mostly Auto selected (except below)
  • +100 MHz
  • Curve Optimizer: -18

 

3090 Settings:

  • Memory Clock: +600 MHz
  • GPU Clock: +100 MHz
  • Power Target: 102% (this is the max increase that the MSI Gaming Trio X 3090 will allow).
  • GPU Temp: 72C
Edited by RAAF492SQNOz_Steve
Added CPU and GPU settings
Link to comment
Share on other sites

RAAF492SQNOz_Steve
4 hours ago, Voyager said:

Looking through the results people have posted, and is there any difference between the 3080 Ti and the 3080? I do notice that the 3090 is up to 20% faster than the 3080, but the Ti numbers seem about the same as the 3080 plain? 

 Really depends on what you will be using the GPU for and games you want to play.  If only using a 4K monitor, a 3080 should be fine.

Also depends on if you are a "fiddler" and like to tune your PC.

 

For Hi Res VR I would be going for a 3080 Ti if you can get it for a good discount over a 3090.

 

Get one with the option to increase power rating if you are one of the aforesaid fiddlers. In most circumstances, a 3080 ti should be on a level footing with a 3090 for all practical purposes. With the Ti having an extra 2 Gb of RAM over the 3080, it give some extra RAM headroom for games that actually use it.

 

In a discussion with Fenris some time ago, the question of how much RAM is needed for Hi Res VR headset users, playing IL2, arose.

I have just completed a IL2 Bodenplatte, single player mission, with 6 or so figthers and another six bombers in heavy cloud.

Ran the HP Reverb G2 at 100% resolution (in 90 Hz mode) and discovered that my GPU used less that 8 Gb of the 24 Gb that my card has.

In this mission, framerates never dropped below 75 Hz and averaged 80 Hz with a mildly tuned RTX 3090. If I fly a mission, without cloud, I get a rock solid 89 - 90 Hz.

 

image.png.2dd57b442d0bda3c407d6a17fe7e32cd.png

Link to comment
Share on other sites

Voyager

@RAAF492SQNOz_Steve Interesting. Honestly, for Il-2, I think the 3080 plain should do me. Its more DCS, FS2020, and some other things that I've been doing that need more ram and vram. That's part of why I'm doing trades on 3080 vs 3090 vs waiting for next gen and doing a complete rebuild. I'm kind of at a spot where anything single thing I change on my computer, aside from dropping in a 3080 or better, is going to decrease my Il-2 performance.

Link to comment
Share on other sites

RAAF492SQNOz_Steve

Ok, Have been fiddling with my CPU, GPU and done what turned out to be a minor memory upgrade and got some useful improvements in performance especially on Frame Lows.  Should also note that I had not set IL2 Sturmovik  Nvidia setting to "Prefer Max Performance", when doing original tests, so my original posted results are a bit on the low side to begin with.

 

Latest results.............................

Motherboard: MSI Tomahawk X570
CPU:               5800x ( -18 undervolt, PBO +100Hz )
CPU Freq:       4.95 Ghz ( Max frquency achievable, average freq is lower )
L3 cache:       32 MB
Cores:             8
Threads:         16
RAM type:      DDR4
RAM size:       32Gb (2x16GB)

Uncore Freq: 2000 Mhz
RAM Freq:      4000 MHz
RAM timings:  16-16-16-16-36-300-1T
Ram type: G.Skill F4-4000C16-16GTZNA
GPU:      MSI RTX3090 Gaming X Trio @102% power 
CPU Cooling: 240mm liquid cooler

GPU driver version: 466.77

v6 Benchmark used.

 

EDIT Comment added 20/07/21

Did a BIOS upgrade on my motherboard and had to back off RAM secondary timings to try and address a intermittent stability issue with one of my CPU's cores that is causing crashes.  Have retested results since the changes and can no longer obtain the results below.  Will do some fine tuning on the weekend that should mostly re-instate my VR2 result but expect that I will not be able to get close to the 1080p results reported below.  Based on my current setting test results, will only end up with about 130 fps average.  Suspect/ guess that I did not change the graphics setting from High to Ultra when commencing the 1080p tests. Apologies for un-intentionally providing some mis-leading results.  Will post corrected results after next weekend.

 

CPU Test 1080p:

Frames: 8174 - Time: 60000ms - Avg: 136.233 - Min: 123 - Max: 185

 
VRTest2 HP Reverb G2, 100% Steam resolution

Frames: 5116 - Time: 60000ms - Avg: 85.267 - Min: 67 - Max: 91

 

Original results from 4 - 5 months ago with different RAM (but I do not think the new RAM provided all the gains, CPU and GPU tuning also helped)

image.thumb.png.20c9527e55f9c9f8359abb108a4e67cf.png

Edited by RAAF492SQNOz_Steve
Reported error with test results
Link to comment
Share on other sites

chiliwili69
On 7/11/2021 at 6:15 AM, RAAF492SQNOz_Steve said:

G.Skill F4-4000C16-16GTZNA

 

Thank you for this new test after your RAM upgrade, you got +12fps in the CPU test!.

 

I don´t think the  "Prefer Max Performance" of the Nvidia settings affects tot he CPU test, since it is just stressing the CPU not the GPU. So you original mark was quite valid for the CPU test.

 

So with that new memory you went from 2x8Gb (3800Mhz) to 2x16Gb (4000Mhz), so from single rank to dual rank. So, some part of your gain could be due to the increase of frequency and the other part due to going from single rank to dual rank memory.

 

I understand that the G-skill F4-4000C16-16GTZNA is just using the default xmp profile which is guarantied by the QVL of Gskill. Right?

 

I would upgrade to that memory, but unfortunately my B550 Gigabyte Mobo is usually in the QVL of top memories. Next time I will build a PC I will go to the top Motherboards which are in the QVL of G-skill. Like Asus, or MSI or EVGA

Link to comment
Share on other sites

RAAF492SQNOz_Steve
2 hours ago, chiliwili69 said:

 

Thank you for this new test after your RAM upgrade, you got +12fps in the CPU test!.

 

I don´t think the  "Prefer Max Performance" of the Nvidia settings affects tot he CPU test, since it is just stressing the CPU not the GPU. So you original mark was quite valid for the CPU test.

 

So with that new memory you went from 2x8Gb (3800Mhz) to 2x16Gb (4000Mhz), so from single rank to dual rank. So, some part of your gain could be due to the increase of frequency and the other part due to going from single rank to dual rank memory.

 

I understand that the G-skill F4-4000C16-16GTZNA is just using the default xmp profile which is guarantied by the QVL of Gskill. Right?

 

I would upgrade to that memory, but unfortunately my B550 Gigabyte Mobo is usually in the QVL of top memories. Next time I will build a PC I will go to the top Motherboards which are in the QVL of G-skill. Like Asus, or MSI or EVGA

 

I actually think that the change of RAM just by itself did not add much to the fps performance improvement as the Patriot Vipers were really punching above their price range with regard to performance (I just did not know it!).

The 2 x 8Gb Patriot Viper had lower primary timings at 15-14-14-14 (and slightly tighter secondary timings) that pretty much compensated for the extra 200MHz increase gain of the  G-skill F4-4000C16-16GTZNA RAM operating frequency.  The G-Skill DIMMs did, eventually after much tuning, have a 2 Gb better score  using Maxxmem but most of this would be due to my CPU Undervolt allowing the CPU to run about 200MHz faster.  Did not try a CPU undervolt when the Patriot Viper RAM was fitted :(

 

IL2 does not need more than 16Gb so for IL2 the RAM upgrade (from a $130 AUD 2 x 8Gb set to the $500 AUD G.Skill 2 x 16Gb set) is quite wasted.

The upgrade will however be useful for when I get around to using MS Flight Sim.  My original purchase decision for the G.Skill RAM was based on  another reviewers claim of 14-14-14-14 timings at 4000 MHz which sounded like DDR4 nirvana!!!!!!

 

I would guestimate that the fps improvements are due to changes in the following proportions:

    CPU undervolt and PBO +100 MHz over clock, so can now do a all core 4950Hz = 30%

    GPU Power increase (102%  power limit and +10% Core volts) = 25%

    IL2 Nvidia setting "Prefer Max Performance" = 30%

    RAM upgrade to G-skill F4-4000C16-16GTZNA RAM = 15%

It really was the combination of many small improvements, made to the above four area's, that delivered the useful gains in performance especially for Min fps scores.

 

Posted a review of the New 32Gb RAM on Newegg. Mine is the one with a Four star rating by Steve

https://www.newegg.com/global/au-en/g-skill-32gb-288-pin-ddr4-sdram/p/N82E16820374145?Item=N82E16820374145&Description=ddr4 5000&cm_re=ddr4_5000-_-20-374-145-_-Product

 

Re: "I understand that the G-skill F4-4000C16-16GTZNA is just using the default xmp profile which is guarantied by the QVL of Gskill. Right?"

The G.Skill XMP profile when applied yielded worse results than what I had already achieved with the Patriot Viper's. It was only when I really went to town on tightening up the G.Skill's secondary timings (and cranked up the CPU frequencies) that I was able to better the Patriot Viper's results. In my opinion the XMP profiles supplied with the G.Skill RAM are abysmal. I am referring to Maxxmem test results here, when assessing RAM performance, as my Aida64 trial has long expired. 

 

On another note:

I am a bit of a newbie when it comes to tuning PC's and am very impressed with how easy it was to do the CPU undervolt and target max frequency adjustment. 5 minutes on Youtube looking at a "How to do it Video" and I was ready to go.

Well done AMD! Would recommend that all Ryzen 5000 series CPU owners give it a shot!

 

Overall,

I will admit that I do enjoy the PC tuning game using mild overclocks etc that do not result in turning my PC into a a running on the absolute limit, power hungry heat generating monster.   :) Getting those fps min improvements has made it all worthwhile. Flying around the Kuban Peninsula with Ultra settings at a pretty consistent 90 Hz & MSAA x 2, with the VR HP G2, ( no clouds and medium shadow, so Yes compromises still required) is stunning.

 

Edited by RAAF492SQNOz_Steve
The usual typo
Link to comment
Share on other sites

PA_Willy

Could you share with us your CPU undervolt for 5800X? Mine do 4850GHz (without OC or PBO changes)

Edited by PA_Willy
Link to comment
Share on other sites

RAAF492SQNOz_Steve
6 hours ago, PA_Willy said:

Could you share with us your CPU undervolt for 5800X? Mine do 4850GHz (without OC or PBO changes)

 

CPU:               5800x ( -18 allcore undervolt, PBO +100Hz )
CPU Freq:       4.95 Ghz ( Max frequency achievable, average freq is lower )

Was able to post with a PBO+150 i.e. 5000 MHz but PC became unstable and would crash after 20 minutes or so.

 

Passmark CPU scores now over 32000

image.png.154f550d61137c005727dac430fa9d2b.png

 

Userbenchmark score now averages 112% (make sure you do more than one run as background processes on your computer can impact results)

image.png.0290655d4c1d10ea3256b40ab2f58859.png

Link to comment
Share on other sites

chiliwili69
20 hours ago, RAAF492SQNOz_Steve said:

    GPU Power increase (102%  power limit and +10% Core volts) = 25%

    IL2 Nvidia setting "Prefer Max Performance" = 30%

 

These two items are just related to the GPU, and the GPU is not really loaded during the CPU test since it is at 1080p and with clouds low. You can test this by yourself by deactivating it and running the CPU test again.

 

I believe runnin the RAM at 4000MHz is the main factor, you can try to reduce it to 3800 or 3600 to see the effect.

On 7/11/2021 at 6:15 AM, RAAF492SQNOz_Steve said:

Ram type: G.Skill F4-4000C16-16GTZNA

 

Which is your RAM model: F4-4000C16-16GTZNA (which is 2x8Gb) or F4-4000C16-32GTZNA (which is 2x16Gb)?

 

you said the F4-4000C16-16GTZNA, but you indicated 2x16Gb

20 hours ago, RAAF492SQNOz_Steve said:

Well done AMD! Would recommend that all Ryzen 5000 series CPU owners give it a shot!

 

I am alss not an expert in PC tunning, I would like to see how much can I gain with that undervolt thing in my 5600X. Is it as easy as you said?

Link to comment
Share on other sites

RAAF492SQNOz_Steve

Re: RAM identifier code query

The RAM is DDR4-4000C16D-32GTZNA according to the box and is indeed  2 x 16 Gb.

However Zen timings reports it as as F4-4000C16-16GTZNA so take your pick on the name.  :)  

 

image.png.6a91e59629ae95560456890305fbf5f3.png

 

Re: performance percentages

I am referring to VR2 test results as that is the score that is important to me and that is what I tuned the various components for.

 

Only added the test for 1080p and scores as an afterthought.

Edited by RAAF492SQNOz_Steve
Comment on performance percentages.
Link to comment
Share on other sites

RAAF492SQNOz_Steve
On 7/11/2021 at 2:15 PM, RAAF492SQNOz_Steve said:

Ok, Have been fiddling with my CPU, GPU and done what turned out to be a minor memory upgrade and got some useful improvements in performance especially on Frame Lows.  Should also note that I had not set IL2 Sturmovik  Nvidia setting to "Prefer Max Performance", when doing original tests, so my original posted results are a bit on the low side to begin with.

 

Latest results.............................

Motherboard: MSI Tomahawk X570
CPU:               5800x ( -18 undervolt, PBO +100Hz )
CPU Freq:       4.95 Ghz ( Max frquency achievable, average freq is lower )
L3 cache:       32 MB
Cores:             8
Threads:         16
RAM type:      DDR4
RAM size:       32Gb (2x16GB)

Uncore Freq: 2000 Mhz
RAM Freq:      4000 MHz
RAM timings:  16-16-16-16-36-300-1T
Ram type: G.Skill F4-4000C16-16GTZNA
GPU:      MSI RTX3090 Gaming X Trio @102% power 
CPU Cooling: 240mm liquid cooler

GPU driver version: 466.77

v6 Benchmark used.

 

EDIT Comment added 20/07/21

Did a BIOS upgrade on my motherboard and had to back off RAM secondary timings to try and address a intermittent stability issue with one of my CPU's cores that is causing crashes.  Have retested results since the changes and can no longer obtain the results below.  Will do some fine tuning on the weekend that should mostly re-instate my VR2 result but expect that I will not be able to get close to the 1080p results reported below.  Based on my current setting test results, will only end up with about 130 fps average.  Suspect/ guess that I did not change the graphics setting from High to Ultra when commencing the 1080p tests. Apologies for un-intentionally providing some mis-leading results.  Will post corrected results after next weekend.

 

CPU Test 1080p:

Frames: 8174 - Time: 60000ms - Avg: 136.233 - Min: 123 - Max: 185

 
VRTest2 HP Reverb G2, 100% Steam resolution

Frames: 5116 - Time: 60000ms - Avg: 85.267 - Min: 67 - Max: 91

 

Original results from 4 - 5 months ago with different RAM (but I do not think the new RAM provided all the gains, CPU and GPU tuning also helped)

image.thumb.png.20c9527e55f9c9f8359abb108a4e67cf.png

 

Have had to report an error with SYN_Vander Benchmark results I obtained in a recent post.  Have put explanation in the original post.

@chiliwili69 Please remove the results (contained in the above post) from your spreadsheet.

 

Will have another go on the weekend to provide reliable SYN_Vander Benchmark results

 

Edited by RAAF492SQNOz_Steve
fixed @ name to correct person
Link to comment
Share on other sites

chiliwili69
4 hours ago, RAAF492SQNOz_Steve said:

Have had to report an error with SYN_Vander Benchmark results I obtained in a recent post.  Have put explanation in the original post.

@chiliwili69 Please remove the results (contained in the above post) from your spreadsheet.

 

Done. Thanks for correcting it. Let´s see how the weekend test goes.

Link to comment
Share on other sites

  • 1 month later...

Hope I'm not too late to join the party!

 

SYN: v6
IL2 Version: 4.602
Motherboard: Asus Maximus Hero X (BIOS 2701)
CPU: i7-8086K  (Passmark CPU Single Thread score: 3131)
Actual Cores: 6
Threads: 12
CPU GHz: 5.1 (AVX offset 0)
RAM: 16GB (Samsung M378A1K43CB2-CTD, Passmark Memory overall score: 3910)
RAM sticks: 2x8 (single-rank, dual-channel)
RAM MHz: 3466
Uncore: 4800
Timings: 19/19/19/43
 
GPU: 6900XT (1125 mV undervolt, +15% power limit,  ReBAR enabled, default clocks, AMD reference card, water cooled)
Public GPU Passmark: 27088
 
VR Headset: Pimax 5k+ (early backer version, 90Hz, Normal FOV, Render Scale 1)
 
CPU 1080p: (full screen mode with FreeSync disabled.. Window'ed mode it was frame limited for some reason)
2021-09-14 14:12:49 - Il-2
Frames: 5439 - Time: 60000ms - Avg: 90.650 - Min: 79 - Max: 132
 
VR Test1:  
2400 x 1980 = 4,752,000 pixels (88%)
2021-09-14 14:30:33 - Il-2
Frames: 4158 - Time: 60000ms - Avg: 69.300 - Min: 58 - Max: 85
 
VR Test2:
3436 x 2832 = 9,730,752 pixels (180%)
2021-09-14 14:37:49 - Il-2
Frames: 4183 - Time: 60000ms - Avg: 69.717 - Min: 59 - Max: 83
 
VR Test2: FSR enabled (Defaults: 0.76 ultra quality, 0.9 sharpness, 0.5 radius)
2021-09-14 14:42:12 - Il-2 VR test 2, FSR enabled
Frames: 4199 - Time: 60000ms - Avg: 69.983 - Min: 59 - Max: 85
 
VR Test2: FSR enabled, Pimax at 110Hz
2021-09-14 14:55:12 - Il-2
Frames: 4041 - Time: 60000ms - Avg: 67.350 - Min: 56 - Max: 82
 
GPU usage does not go beyond 50%, average GPU power draw of 120W during the benchmarking. This wont keep me warm through the winter haha! Either IL2 needs to be more accommodating/efficient with how it manages its workload on an AMD GPU, or AMD needs to emulate whatever methods NVIDIA are doing with their drivers to keep the CPU and GPU in-sync better. Or maybe Windows 11 will surprise us with its WDDM revamping and we can soon forget all about this?

  • Like 1
Link to comment
Share on other sites

chiliwili69
On 9/14/2021 at 8:14 PM, halz said:

GPU usage does not go beyond 50%, average GPU power draw of 120W during the benchmarking. This wont keep me warm through the winter haha! Either IL2 needs to be more accommodating/efficient with how it manages its workload on an AMD GPU, or AMD needs to emulate whatever methods NVIDIA are doing with their drivers to keep the CPU and GPU in-sync better. Or maybe Windows 11 will surprise us with its WDDM revamping and we can soon forget all about this?

 

It is never late!, thank very much for these series of tests.

 

They confirm even more the problem that latest AMD card have with IL-2, in monitor and in VR. That´s why I am marking all AMD cards in orange in the table.

You have the the top card of AMD and watercooled, but it seems there some issues limiting the performance in VR (even maybe in monitor 1080p).

 

Bear on mind, that this low performance of AMD cards is only in IL-2, other games doesn´t suffer this issue. This was reported here:

 

Link to comment
Share on other sites

BH_Adabadoo_VR

i have an asus rog maximus xiii hero motherboard

 

it's got an i9-11900KF and 32GB gskill cl14 in 2 sticks inside.

 

Its clocking pretty good but it's not locked on all cores.

To do you test you want locked on all cores.  I can't figure out how with this motherboard.

 

With my last motherboard it was easy.  With this one I can sync all cores to 50 but that only limits it to 5100 it doesn't lock it there.

 

Anyone know what to set to lock it on all cores to a core multiplier of 50 or 51 or 49 to test stability and do the run??

 

Thanks all!!!

Link to comment
Share on other sites

chiliwili69

I have no experience with that BIOS Mobo or CPU.

With my previous intel CPU (4790K) I was using "MSI Command Center" for simple overclocking all cores, instead of using BIOS.

 

You can try with the Intel XTU tool as well:  https://www.intel.es/content/www/es/es/gaming/resources/overclocking-xtu-guide.html

 

You can also try to run the benchmark with the standard Turboboost, just to check if you achieve the same results than keeping all cores to 5.0 GHz.

 

FYI, In most of the Ryzen Zen 3 tests performed, we don´t use manual OC to keep all cores to one frequency. We just trust in the AMD PBO or AutoOC which do a very good job just out of box.

Link to comment
Share on other sites

16 hours ago, BH_Adabadoo_VR said:

Anyone know what to set to lock it on all cores to a core multiplier of 50 or 51 or 49 to test stability and do the run??

 

Setting the Windows Power Profile to "HIgh Performance" should push all of the cores to the maximum clock. Otherwise, the BIOS should have an option to disable CPU C-States. That will force it to always run at max speed, with or without Windows. Will consume a fair amount more power while the system is idle.

Edited by halz
Link to comment
Share on other sites

BH_Adabadoo_VR

It has all of the ability it's just not really something you do with an i9-11900k

 

The older cpus lock because you have a core ratio you set and it locks them all at that.

 

These new cpus have limits instead as you all state above.  As stated the mother board and cpu are doing a great job of keeping it at various frequencies asked for but not locked and as heat or load changes so does frequency.  Generally without affinity the OS is spreading the thread across  all CPUs at frequency for thread load.

 

So ALL cpus are running 5400 sharing the load a bit at a time not just the best cores.  But trying to lock at a frequency that the i9-11900k can hit on all cores under all  loads is not good for benchmarking or gaming or even general use.  It's much better used if you set core by load and start at your highest frequency and work down.  I did not do well on the silicon lottery.  My CPU is sp50 but it's getting great scores and is stable with and without  AVX at reasonable temps getting pretty high frequency with and without AVX for prime.  About 4850 with AVX and 5050 without  AVX prime.  It gets to 5400 on all cores spreading the load of 1 core loaded thread. 2 cores drops to 5300 from heat.  All cores without avx not prime testing gets 5200 solid all cores.  Not the best results but not bad and much better than trying to lock at 5200.

 

I've set power is ultimate performance with cpu fully at 100 all the time and active cooling.  If your ok with it bouncing about abit but doing the best it can under all  loads I think that's as good as it gets for this rig.  I'll tweak it some more and run a bunch of benchmarks but I don't think all core sync is doing nearly as good as it adjusting cores.

 

It's good I've been tweaking for your benchmark, I did  gain 2-5%cpu over what I was previously happy with in a bunch of benchmarks though the fan is noisier now as it does get hotter.

 

Once I'm happy it's stable and solid I'll  run the benchmark you guys made.  When I get some time I'll also run my six year old computer that is locked at 4400 and has a 1080ti.  I'm curious.  

 

Thanks all

9 hours ago, halz said:

 

Setting the Windows Power Profile to "HIgh Performance" should push all of the cores to the maximum clock. Otherwise, the BIOS should have an option to disable CPU C-States. That will force it to always run at max speed, with or without Windows. Will consume a fair amount more power while the system is idle.

I don't see any benefit when I've tried  this in any benchmark so far.  It just wastes a bit of power and heat reserve.  It goes as far as it can when it gets load pretty nicely without this. Perhaps it could help with latency a tiny bit but your already ready and in load while you load your game and setup VR.  IL2 is never idle really when your ready to play so things are already loaded up.  Only advantage of this is that the fans might be closer  to load spool and so not need to catch up under load start.  But that's not and advantage for a long test.  It will throttle fans speed up it finds stable temp creeps freq up as it can till fan is max and temp is at threshold and stays there till test over.

 

Am I missing something??

Any benefit to setting Game mode, process affinity to best cores, process priority??

 

I am curious about setting CPU affinity for the game threads.  I haven't tried this.  When in game are there a few threads that affinity can be set for to specific cores you want?  Has anyone seen any advantage from that?  Any benefit from boosting priority?

 

Game mode has improved overtime and has become a big benefit to OBS recording with Nvidia cards.  Setting game mode has helped there while I fly.  So I now set game mode for IL2 and OBS.  But I have not tried affinity or priority for the processes.

Link to comment
Share on other sites

chiliwili69
On 9/18/2021 at 1:49 AM, BH_Adabadoo_VR said:

though the fan is noisier now as it does get hotter.

 

This 11900K has 8 cores, 2 less than the previous 10900K. So it generates less heat so it can overclock higher.

 

You can also try to disable 2 cores in the BIOS and run with just 6 cores. It will generate less heat, so perhaps you can achieve higher clocks. For IL-2, 4 or 6 active cores is more than enough.

On 9/18/2021 at 1:49 AM, BH_Adabadoo_VR said:

process affinity to best cores, process priority??

 

As far as I remember, there are no measurable benefits from using process affinity and process priority with IL-2. Otherwise I will be using it.

 

I have been using process affinity and propcess priority in another single-threaded application at work and there I could achieve a measurable extra 5%, but not more.

Link to comment
Share on other sites

Upgraded  a lot of my computer in the last couple of weeks, and because we had a 4K TV, and I was tearing apart my machine, I was able to run the GPU test on the 1080 TI. Not the new card though.

 

So results

Configuration 1: Ryzen 5800X, 1080 TI, DDR4-3600

 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 471.96


VR Headset: Vive Pro 2

Game Version 4.603


    CPU Test        
    Min        Max    Avg
1    103    158    117.067
2    102    157    116.783
3    102    156    116.567
   102.3    157    116.806
            
     GPU Test    
    Min    Max    Avg
1    67    101    86.267
2    70    99    86.167
3    69    100    85.833
    68.7    100    86.089
            
    VR Test 1 (2148x2148)        
    Min    Max    Avg
1    42    80    46.367
2    42    75    45.683
3    42    73    45.483
    42    76    45.84433333
            
    VR Test 2 (3128x3128)        
    Min    Max    Avg
1    42    46    44.817
2    42    46    44.883
3    42    46    44.8
    42    46    44.833
 

Configuration 2: Ryzen 5800X, 1080 TI, DDR4-2133

    CPU Test        
    Min    Max    Avg
1    91    143    105.1
2    91    144    106.583
3    91    146    106.367
    91    144.3    106.016
            
    GPU Test        
    Min    Max    Avg
1    65    98    84.083
2    68    96    84.033
3    66    95    83.067
    66.3    96.3    83.727
            
            
    VR Test 1 (2148x2148)        
    Min    Max    Avg
1    43    58    43.117
2    42    57    45.086
3    42    60    45.183
    42.3    58.3    44.462
            
    VR Test 2: 3100x3100        
    Min    Max    Avg
1    42    46    44.783
2    42    46    44.817
3    30    46    44.533
    38    46    44.711
 

 

Configuration 3: Ryzen 5800X with 3080 Ti, and DDR4-3600

 Motherboard:    Gigabyte X570 Aorus Xtreme v 1.2

 Bios Version F34
 CPU:        AMD Ryzen 7 5800X
 CPU Freq:    4450 MHz
 L3 cache:    2x16 MB
 Cores:        8
 Threads:    16
 RAM type:    DDR4
 RAM size:    128 GB (4x32)
 NB Freq:    1799.6 MHz
 RAM Freq:    1799.6
 RAM timings:    16-22-22-42
 GPU:        EVGA 3080 TI FTW3, Core 1800Mhz, Memory 9501Mhz (Factory overclock)

 Driver Version 4712.12


VR Headset: Vive Pro 2

Game Version 4.604b

    CPU Test        
    Min    Max    Avg
1    99    154    114.75
2    104    164    118.433
3    102    154    115.617
    101.7    157.3    116.267
            
    GPU Test   (Not performed)     
    Min    Max    Avg
1            
2            
3            
    #DIV/0!    #DIV/0!    #DIV/0!
            
            
    VR Test 1 (2188x2188)        
    Min    Max    Avg
1    44    91    69.15
2    42    90    59.333
3    42    91    65.717
    42.7    90.7    64.733
            
    VR Test 2: 3128x3128        
    Min    Max    Avg
1    42    77    45.967
2    42    89    47.133
3    42    70    45.317
    42    78.7    46.139
 

-------------------------------------------------------------

 

Was not able to perform the GPU test, since it required having the computer set up and running before moving it back to the office, but the main findings are, going from DDR4 2133 to DDR4 3600 is about a 10% performance difference. But going from 16-19-19-39 timing down to 16-22-22-42 timings was not a big impact.

 

The VP2's locking in of 45Hz does weird things to the VR benchmarks, however, the big thing there seems to be that the 3080 TI is around 40% faster than the 1080 TI at 9.5Mp, but it looks like both run out of juice at 19Mp. That said, my personal settings are at ~16Mp and there, I'm still seeing a~40% boost over the 1080 TI. (Test not listed, but I'm getting the same framerates at 1.0 pixel density as I was at 0.7 on the 1080 TI card.)

Link to comment
Share on other sites

14 hours ago, Voyager said:

Was not able to perform the GPU test, since it required having the computer set up and running before moving it back to the office, but the main findings are, going from DDR4 2133 to DDR4 3600 is about a 10% performance difference. But going from 16-19-19-39 timing down to 16-22-22-42 timings was not a big impact.

 

The VP2's locking in of 45Hz does weird things to the VR benchmarks, however, the big thing there seems to be that the 3080 TI is around 40% faster than the 1080 TI at 9.5Mp, but it looks like both run out of juice at 19Mp. That said, my personal settings are at ~16Mp and there, I'm still seeing a~40% boost over the 1080 TI. (Test not listed, but I'm getting the same framerates at 1.0 pixel density as I was at 0.7 on the 1080 TI card.)

 

Many thanks for posting all those tests. There is a lot info in there. It is the first VP2 tested here!  And with a 3080Ti!

 

Comments:

 

1.- As you say, RAM frequency plays an important role. 10 fps in the CPU test is a lot.

 

2.- Your previous 1080Ti works well in the 4K tet and also in VR with your previous G1. Results are aligned with other tests.

 

3.- Your VP2 with the 1080Ti is quite below than your VR test1 with 9.5Mp with the G1. With the G1 you had Avg 72 but with the VP2 you have Avg 46. Something weird is happening with VP2.

 

4.- With the new 3080Ti, you seems to have the same low values in both VR tests due to the VP2. Do you still have the G1? You can try to run the G1 again with the 3080Ti.

 

We could have more data from people having VP2 and high end cards like 3090, 3080Ti and 3080. Just to reproduce that low values and investigate why.

Link to comment
Share on other sites

@chiliwili69 I've still got the G1 boxed up. I'll see if I can hook it up later this week and see how it does. 

 

I believe the VP2 locks the frame rate at at 45 or 90 fps and does not have frame rates between those, which artificially deflated the performance. But testing with the Reverb G1 should show what's up. 

Link to comment
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
 Share

×
×
  • Create New...