Benchmark for CPU/RAM performance: Remagen 4.002 to 4.005

November 29, 2019

NOTE: This Remagen benchmark doesn´t run anymore. There is a new benchmark based in mission:

This is a benchmark test targeted to measure the performance of different CPUs/RAMs (and GPU to some extent) when running IL-2 in a 1080p monitor (FullHD: 1920x1080).

The benchmark is a 3 minutes flight in the new Rheinland-Summer map at 19:00 around the famous Remagen bridge. Just two SpitfireIX vs 4 Ju88 and 1 He-111 H16 at 600m altitude. Wheather average and clouds at 1200m

You can follow the instructions below to run it in your PC.

So you will be able to compare your performance with similar rigs and determine if something is wrong in your PC. (Hardware, software, settings)

We can also use it to measure the impact of different settings in the fps.

You will also know what are the best CPU/RAM and GPUs for our favourite game: IL-2 GB!!

I didn´t describe a VR test here since it is too complex to design a fair: Multiple devices, different resolution, different FOVs, different internal SS ratios, different panel frequencies, capped by 45-90 limits, activate ASW tech, etc, etc. But just leave a column at the right if someone wants to run it in VR.

This monitor info will be useful for VR, since everything which is good for monitor is good for VR. ?

======================================================================================================================================

Instructions for measuring and comparing IL-2 performance on 1080p monitor

Note: Before running this test you should first spend the required time to Overclock your CPU to the maximum (and stable) capacity of your PC. There is multiple info on the web for Overclock.

Be sure to run all your cores at the same manually fixed CPU frequency and set he AVX Offset=0 . So all the test performed (CPU-Z, Passmark Performance Test and IL-2 test) are run at the same frequency with no CPU throttling.

1. GET YOUR PC INFO and RUN CPU passmark

Install and execute the free tool CPU-Z from: https://www.cpuid.com/softwares/cpu-z.html
Collect the following info from CPU-Z:

Motherboard: MSI Z97M-Gaming
CPU: 4790K
CPU Freq: 4.8 Ghz
L3 cache: 8 MB

Cores: 4 (number of active physical Cores)

HT: Off (HyperThreading. If Cores=Threads then HT is Off)
RAM type: DDR3
RAM size: 16 GB

NB Freq: 4000 MHz
RAM Freq: 2933 MHz (this is 2xDRAM freq if Dual channel)
RAM Latency: 12 (CAS Latency CL)
GPU: 1080Ti
STMark: 2782

To get the STMark (means Passmark CPU Single-Threaded) you need to install the free tool "Passmark Performance Test v9.0" from here (30 days free trial but it works as well after that period) :
You only need to run the CPU Single Threaded test by pressing the RUN or RERUN:

2. TOOLS AND TOOL SETUP.

Install Fraps ( free fps measuring software) from: http://www.fraps.com/

Run Fraps.

Under the "general" tab, uncheck "fraps window always on top".

Under the "99 fps" tab, set benchmarking hotkey to "P".

Also check and set the "stop benchmark after __ seconds" to 180 seconds

Get the benchmark Flight Record track: Download it from this link

Unzip it ((a . trk file and a folder) and drop the contents into your "IL-2 Sturmovik Battle of Stalingrad\data\Tracks" folder as shown below:

3. SETTINGS.

Turn off any IL-2 Mod to run the test.

If your GPU uses Nvidia G-Sync or AMD Free-sync, you must switch it off to don´t limit fps during the test.

In IL-2 Settings:

Game: Map Scenery Distance to 20Km.

Graphics: set as below

Note: Be sure you run the clouds quality to Low. We want to be sure we don´t overload the GPU.

Also in your NVIDIA (or AMD) control panel put the following (or closer for AMD) settings:

I use all global (I think mostly default) settings for IL-2 except Power to "Prefer Maximum performance"

4. BENCHMARKING.

Start Fraps, if it's not already running.

Start IL-2, load and Start the Remagen_Benchmark_v1 flight record. You should see this:

Press "H" to hide the HUD.

Press "Backspace" to hide the IL-2 fps counter (in case you had it active, in any case it doesn´t affect performance at all)

Press "P" to both: Start the Fraps measurement and unpause the track. Left the 3 minutes track run without touching anything.

5. REPORT RESULTS.

Find the folder you installed Fraps. Inside that folder is a "Benchmarks" folder and inside that you will find a "FRAPSLOG.TXT" with the Frames, Avg, Min and Max fps values.

There will be also the recorded fps in *.csv files that you can open with notepad or excel if you like to plot them.

Report your Frames, Avg, Min, Max values in a post and write also your system specs from section 1 (IL-2 release, Motherboard, CPU, etc)

I will compile all collected info and will update it in the below spreadsheet, in Remagen4.0 tab:

https://docs.google.com/spreadsheets/d/1gJmnz_nVxI6_dG_UYNCCpZVK2-f8NBy-y1gia77Hu_k/edit?usp=sharing

Edited November 15, 2020 by chiliwili69

November 29, 2019

Thank you very much for the effort. Can you please rebrand this to a VR test.

Reasons:

1. We are in the VR section of the forums.

2. What is good for the monitor is not automatically good for VR. There is no "disagreement" about it, it's a simple fact we can't help. The bottleneck in VR is mostly single thread CPU speed. On monitor, GPU throttles earlier.

Different VR headsets create different results. We should keep this in mind too, and have the subsections.

Thank you very much for creating a new test we can use!! And the thread as well!

Edited November 29, 2019 by SCG_Fenris_Wolf

November 29, 2019

**************************************************************************************

Motherboard: MSI X299 Gamming Pro Carbon AC
CPU:             i7 7820x
CPU Freq:        4.7 Ghz AVX=0,AVX512=0
L3 cache:        11 MB
RAM type:        DDR4
RAM size:        32 GB
RAM Freq:        3466 MHz
RAM Latency:        16
GPU:                 1080Ti MSI, max. core 2015Mhz
STMark:           2703

FRAPS TXT:
2019-11-29 16:02:11 - Il-2
Frames: 22324 - Time: 180000ms - Avg: 124.022 - Min: 82 - Max: 215

BOX 4.002

**************************************************************************************

@Fenris : the google link contains also VR fields, so may be this is the first step only ?

@chilli thanks for your effort, there are never enoug standartized tests,

anyway, I hope for Jason´s statement from October:

" I need to design a standard test for everyone and check results. "

Edited November 29, 2019 by wju
fps graph added

November 29, 2019

This is the plot of the fps along the 3 minutes.

My goal was to create a track which is mostly limited by CPU, that´s why I run it at 1080p, but also see when it is GPU limited.

You can see that only during the loop is when the CPU is not the bottleneck:

2 hours ago, SCG_Fenris_Wolf said:

Reasons:

1. We are in the VR section of the forums.

2. What is good for the monitor is not automatically good for VR. There is no "disagreement" about it, it's a simple fact we can't help. The bottleneck in VR is mostly single thread CPU speed. On monitor, GPU throttles earlier.

Thanks for your feedback. My main interest is VR (that´s why it is in VR section), but since VR it is mostly limited by CPU (at least for devices like Rift, Rift-S, Index, VivePro and GPUs like 1080 or above) I wanted to know the best CPU/RAM (including new AMD line and future Intel 10th) and also how every parameter affects the CPU.

If you try to know that in VR it is mission impossible since you need to take into account many factors:

- Software of every VR device and the different betas software.

- Resolution of every VR device

- Frequency of the VR display (72Hz, 80 Hz, 90Hz)

- The internal Supersampling which is different in every device

- The average fps is misleading when the fps is capped at 90 o 45fps.

Even Though I have created a column at the right for people to report their own VR test in addition to the monitor test. That info would be also valuable but more difficult to extract knowledge.

November 29, 2019

1 hour ago, wju said:

STMark: 2703

Thanks wju for being the first to test it. Perfectly reported.

Here my test at test settings:

2019-11-29 17:09:46 - Il-2
Frames: 22617 - Time: 180000ms - Avg: 125.650 - Min: 81 - Max: 215

Just run four test with LOW, BALANCED, HIGH and ULTRA settings. Not a huge difference between them:

LOW
2019-11-29 17:14:55 - Il-2
Frames: 25593 - Time: 180000ms - Avg: 142.183 - Min: 92 - Max: 232

BALANCED
2019-11-29 17:23:41 - Il-2
Frames: 23434 - Time: 180000ms - Avg: 130.189 - Min: 88 - Max: 218

HIGH
2019-11-29 17:09:46 - Il-2
Frames: 22617 - Time: 180000ms - Avg: 125.650 - Min: 81 - Max: 215

ULTRA
2019-11-29 17:31:39 - Il-2
Frames: 21781 - Time: 180000ms - Avg: 121.006 - Min: 80 - Max: 213

Edited November 29, 2019 by chiliwili69

November 29, 2019

What is the "expected" and the "delta" in the reporting form?

Also, for cpu-z, although I have a i5-7600 which cpu-z states on the specification line is 3.50 GHz, the core speed indicator (in the Clocks (Core #0)) indicates core speed of 3.9 - 4.1 GHz (fluxuating)- which number shall I report?

Edited November 29, 2019 by Redwo1f

November 29, 2019

Nice, will try this and report later next week. I5-7600k ,if I remember right ST mark is somewhere between 2800-2900 when oc to 5Ghz.

Edited November 29, 2019 by Hartigan

November 29, 2019

Can't add to the online sheet yet as I don't have access (set to view only)...but will add to it when I do.

Motherboard: Gigabyte B250M-DS3H
CPU: i5-7600
CPU Freq: 4.1 Ghz
L3 cache: 6 MB
RAM type: DDR4
RAM size: 16 GB
RAM Freq: 2400 MHz
RAM Latency: 15
GPU: GTX 1060 (6 Gb)
STMark: 2319

HIGH settings specified in first post:

Frames: 13562 Time: 180000ms - Avg: 75.344 - Min: 51 - Max: 170

ULTRA

Frames: 13211 Time: 180000ms - Avg: 73.394 - Min: 49 - Max: 169

BALANCED

Frames: 16029 Time: 180000ms - Avg: 89.050 - Min: 57 - Max: 173

LOW

Frames: 17097 Time: 180000ms - Avg: 94.983 - Min: 60 - Max: 188

(note: this system rarely ever suffers from micro-stutters during il2 BoX play, and none were experienced during this testing either).

Edited November 30, 2019 by Redwo1f
Added Frames

November 29, 2019

@chiliwili69 have you compared the results of a track like this with live gameplay?

The last time I tried doing that, the results I got live were much lower than from the recorded track. I think that might be due to the AI and flight modeling. I don’t suspect that using recorded tracks as benchmarks will produce valid results.

November 30, 2019

Motherboard: Asrock z87 Extreme 6
CPU: 4670K
CPU Freq:       4.5Ghz
L3 cache: 6 MB
RAM type:       DDR3
RAM size:       16 GB
RAM Freq: 2400 MHz
RAM Latency: 10
GPU: 1080
STMark: 2702

Version 4.002 Box Settings:

High ( per Chiliwili69 settings)

Time: 180000ms - Avg: 111.711 - Min: 74 - Max: 193

With Low cloud setting:

Frames: 21258 - Time: 180000ms - Avg: 118.100 - Min: 74 - Max: 199

Edited December 10, 2019 by Stolle
needed to revise my graphic settings will edit with corrections re ran test with low cloud settings

November 30, 2019

Interesting other tidbits that I have found during standardized testing on my system as I experiment:

BALANCED

- SSAO on (-0.9% fps)

- Clouds Extreme (up from high) (-7.3% fps)

-Shadows medium (down from high) (+26% fps)

HIGH

-SSAO on (-4.0% fps)

-Clouds Extreme (up from high) (-13.2% fps)

-Shadows medium (down from high) (+28% fps)

--wanna gain fps in a big way - turn down shadows!

Edited November 30, 2019 by Redwo1f

November 30, 2019

5 hours ago, SharpeXB said:

@chiliwili69 have you compared the results of a track like this with live gameplay?

The last time I tried doing that, the results I got live were much lower than from the recorded track. I think that might be due to the AI and flight modeling. I don’t suspect that using recorded tracks as benchmarks will produce valid results.

This is a benchmark to compare systems, as such the only way to do this is with a track or the results are not comparable.

As far as I know any other method would not produce valid/consistent results. Read Chilli's posts again.

Even if the result is lower/higher than in actual live play it does give a comparison in IL-2 rather than a pure synthetic benchmark

Cheers, Dakpilot

November 30, 2019

Hate to be a noobie, but do I need to purchase Bodenplatte in order to run this benchmark?

November 30, 2019

22 minutes ago, whitav8 said:

Hate to be a noobie, but do I need to purchase Bodenplatte in order to run this benchmark?

I’m not sure. You don’t need to own Bodenplatte to join a multiplayer server that uses it, so it might be that the game will play a replay off that map also. Just won’t let you fly the planes yourself. Give it a go it only takes 2 minutes to try.

November 30, 2019

4 hours ago, whitav8 said:

Hate to be a noobie, but do I need to purchase Bodenplatte in order to run this benchmark?

No, because a time ago I did run a benchmark with a plane I did not own.

@chiliwili69 first of all a BIG THANK, for keeping this alive, but as what @SharpeXB did wrote down is interesting, are you sure, can an recorded Bopl action video be compared to an real game benchmark? I’m afraid we are only comparing the results of the BoX video play capability from our hardware.

November 30, 2019

1 hour ago, Dutch2 said:

@chiliwili69 first of all a BIG THANK, for keeping this alive, but as what @SharpeXB did wrote down is interesting, are you sure, can an recorded Bopl action video be compared to an real game benchmark? I’m afraid we are only comparing the results of the BoX video play capability from our hardware.

A very valid point...

November 30, 2019

1 hour ago, Dutch2 said:

I’m afraid we are only comparing the results of the BoX video play capability from our hardware.

8 minutes ago, dburne said:

A very valid point...

It's a limited benchmark, for sure, but it's the best we have at the moment. I made a specific Ai benchmark map a while ago, which would be closer to reality because it's calculating Ai decisions and flight-paths, but Chili found that it produced very different results from run to run. Probably because the Ai make different decisions each time, so even if you start the same mission over and over again, you get a different fight and therefore different benchmark numbers.

For a more stable benchmark we would need to start the whole map with the exact same "random seed" (technical term) so that all decisions, effects, everything basically, was the same each time. And maybe there are other things that would need to be controlled by the developers. It's probably not a small job to produce a good built-in benchmark. And you need to decide what to include -- are you trying to be representative of gameplay (RDR2 benchmark is actually about on-par with most gameplay) or are you trying to heavily load the system for the purposes of comparing hardware (synthetic benchmarks like Firestrike tend to do this).

Anyway I'm a giant insomniac tonight so I'm going to read Chili's instructions and benchmark my rig. Back soon with results.

November 30, 2019

19 hours ago, Redwo1f said:

What is the "expected" and the "delta" in the reporting form?

The expected value is the Avg fps expected for a given STMark value and based in a linear correlation (STMark vs Avg fps) of all the tests. There will be a graph below in the sheet showing the correlation.

The delta is the difference between the expected and the achieved.

We were using that in previous tests (Balapan, Samuel, etc) to check is the achieved value was close to the expected. If the delta is positive then the PC is performing above average, if it is negative is performing below average.

19 hours ago, Redwo1f said:

the core speed indicator (in the Clocks (Core #0)) indicates core speed of 3.9 - 4.1 GHz (fluxuating)- which number shall I report?

Your CPU is a non-K (or non-X) ending CPU. This means you can not fix the frequency at which you want to run it.

The base clock speed is the specification (3.5 GHz) but the Intel Turboboost technology increases the speed up to 4.1 GHz based on the CPU load and CPU temp.

If there is no load to CPU it reduces the speed to up 3.5, but when you run IL-2 (if there is no limitation on temp or power or amp) the freq should be maintained at 4.1.

You can check that by running the free tool MSI afterburner and trending the CPU clock during the test in the background.

November 30, 2019

Motherboard: MSI Z370 SLI Plus
CPU:         8086K
CPU freq:    5.0 Ghz        (hyper-threading disabled, 6-cores only)
L3 Cache:    12 MB
RAM Type:    DDR4
Ram size:    16 GB
Ram freq:    3466 Mhz (OC from 3200 stock)
Ram latency: 16
GPU:         RTX 2080  (EVGA 2080 XC Ultra, 'stock' pre-OC by manufacturer)
STMark:      3017, 3011, 3017  (ran 3 times)

Benchmarks:   Frames / Avg / Min / Max
Run 1:        24956 / 138.644 / 95 / 227
Run 2:        24866 / 138.144 / 96 / 224
Run 3:        24821 / 137.894 / 96 / 224

(Game NOT restarted between runs, I simply exited to main menu and then reloaded the track. This slight reduction in performance between runs is something I observed in VR benching too).

Now for fun, let's run the CPU at 4.5 and 4.0 Ghz instead.

CPU freq:    4.5 Ghz        (hyper-threading disabled, 6-cores only)
STMark:      2723
Benchmark:   23668 / 131.489 / 90 / 216

CPU freq:    4.0 Ghz        (hyper-threading disabled, 6-cores only)
STMark:      2421
Benchmark:   21464 / 119.244 / 82 / 197

Also, great benchmark track! Really fun to watch. You should play the new map "Crossing the Rhine" on Combat Box -- Remagen is there, with many players fighting over it!

Edit: Also, you might want to calculate memory latency. I don't think it's just speed / CAS. Here's a calculator that gives you nanoseconds: https://notkyon.moe/ram-latency.htm

Edited November 30, 2019 by Alonzo

November 30, 2019

15 hours ago, Redwo1f said:

Can't add to the online sheet yet as I don't have access (set to view only)...but will add to it when I do

Thanks for the test. Normally I upload the data to the sheet every 2-3 days.

13 hours ago, SharpeXB said:

chiliwili69 have you compared the results of a track like this with live gameplay?

The last time I tried doing that, the results I got live were much lower than from the recorded track. I think that might be due to the AI and flight modeling. I don’t suspect that using recorded tracks as benchmarks will produce valid results.

No I didn´t. We don´t exactly know what is reproduced in a recorded track. Probably the AI and perhaps some physics are already calculated or maybe the track record the input produced by the input devices. But independently of that all the graphics and building the scene is performed by the CPU again. So even if the fps are not exactly the same than in game play we can use it to see what CPUs/RAMs are best suited to IL-2 and hence to IL-2 VR.

13 hours ago, Redwo1f said:

--wanna gain fps in a big way - turn down shadows!

Exactly!

I wanted to use this test to calculate the impact of every single settings.

I think you can live with shadows medium and gain those fps.

7 hours ago, whitav8 said:

Hate to be a noobie, but do I need to purchase Bodenplatte in order to run this benchmark?

We had the same question in the past. I think you don´t need to purchase it. But will not put my hand on the fire.

I chose the new Rheinland map since it is the newest.

2 hours ago, Dutch2 said:

No, because a time ago I did run a benchmark with a plane I did not own

OK, you answer that. Great then.

2 hours ago, Dutch2 said:

can an recorded Bopl action video be compared to an real game benchmark? I’m afraid we are only comparing the results of the BoX video play capability from our hardware

I think it is not a pure video play since the is less fps when shooting and hitting planes (so Damage model should calculate that again) than when I was just flying along the forest.

I asked developer what the recorded track really do, but get no answer.

November 30, 2019

Thanks much for the benchmark! (You are correct about not needing to purchase Bodenplatte - although I should! )

Here are my test results:

=================================================

Motherboard: ASUS MAXIMUS XI HERO
CPU: 9700K
CPU Freq: 5.0 Ghz
L3 cache: 12 MB
RAM type: DDR4
RAM size: 32 GB
RAM Freq: 3200 MHz
RAM Latency: 17
GPU: 2070
STMark: 3018

Results with Graphic Settings as specified:

2019-11-30 06:07:30 - Il-2
Frames: 23616 - Time: 180000ms - Avg: 131.200 - Min: 94 - Max: 228

=====================================================

Results with Graphic Settings as set for VR (Shadows Quality = medium, Mirrors=off,Landscape = x2, Horizon=70km, grass=normal,clouds=medium):

2019-11-30 06:34:57 - Il-2
Frames: 29360 - Time: 180000ms - Avg: 163.111 - Min: 106 - Max: 251

Results with Graphic Settings as set for VR and actually in VR (Reverb at 2160x2116 SS and NO ASW/reprojection)

2019-11-30 08:46:17 - Il-2
Frames: 10859 - Time: 180000ms - Avg: 60.328 - Min: 48 - Max: 91

Edited November 30, 2019 by whitav8

November 30, 2019

3 hours ago, Alonzo said:

It's probably not a small job to produce a good built-in benchmark.

Back during BoS Early Access we were all asked to post test results in order for the Devs to work out the graphic presets. Obviously having a benchmark track for us to all use would have been beneficial. But I don’t think that’s possible in this game due to the AI, otherwise they would certainly have provided one. I suppose it’s a safe assumption that whichever hardware performs better on a track will perform better live, or it’s the only tool that there is. But it’s an easy test to do. Just run a live mission with Fraps benchmarking it and then run the track of that same gameplay again and benchmark that. Then compare.

November 30, 2019

Motherboard: Gigabyte X570 Aorus Master
CPU: AMD Ryzen 5 3600x
cpu freq: 4400.55MHz
L3 cache: 2x16 MB
Ram type: DDR4
Ram size 32 GB
Ram speed: 3600MHz
Ram latency: 16
GPU: 1080Ti
STMark: 3040

Low: 2019-11-30 11:44:17 - Il-2
Frames: 28129 - Time: 180000ms - Avg: 156.272 - Min: 103 - Max: 252

Balanced: 2019-11-30 11:35:27 - Il-2
Frames: 25899 - Time: 180000ms - Avg: 143.883 - Min: 95 - Max: 237

High: 2019-11-30 11:00:41 - Il-2
Frames: 24850 - Time: 180000ms - Avg: 138.056 - Min: 92 - Max: 237

Ultra: 2019-11-30 11:17:50 - Il-2
Frames: 23608 - Time: 180000ms - Avg: 131.156 - Min: 89 - Max: 236

Edited November 30, 2019 by Jaws2002

November 30, 2019

1 hour ago, SharpeXB said:

I suppose it’s a safe assumption that whichever hardware performs better on a track will perform better live, or it’s the only tool that there is. But it’s an easy test to do. Just run a live mission with Fraps benchmarking it and then run the track of that same gameplay again and benchmark that. Then compare.

Yes, that was my assumption regardless of the AI or physics being calculated.

But as you say, it an easy test to do and compare. I will try that.

November 30, 2019

@jaws2002,

Great to see how well AMD Ryzen is performing! One question, you show the results for all different graphic settings (low, balanced,...) but did you provide results for the settings exactly as chiliwili69 established at the very top of this thread?

@chiliwili69,

Great new benchmark! Thanks for the effort. I would like your opinion of the usefulness of adding a couple runs (as I did above) with our VR graphical settings but not running in VR mode and then the run with VR but no ASW/reprojection and SS set at our choice but reported in pixels horizontal x vertical instead of percentage. I realize that the VR results are subject to a lot of variables, but it still is fun to compare.

Thanks

Dave

Edited November 30, 2019 by whitav8

November 30, 2019

4 hours ago, Alonzo said:

Edit: Also, you might want to calculate memory latency. I don't think it's just speed / CAS. Here's a calculator that gives you nanosecond

Thanks, I have just edited the spreadsheet to indicate the True Latency in ns. It is just the formula 2000*CAS/Freq.

2 hours ago, whitav8 said:

Frames: 23616 - Time: 180000ms - Avg: 131.200 - Min: 94 - Max: 228

Thanks for your run. I saw your CPU acts exactly the same than Alonzo (same CPU speed, same cache, same STMark) but your result is slightly lower than Alonzo (7fps).

There could be two reason for that: the GPU and the RAM latency. With more data from mor users we will know more about it.

58 minutes ago, Jaws2002 said:

CPU: AMD Ryzen 5 3600x
cpu freq: 4400.55MHz

Thanks for running the first new ryzen series! As we suspected the Ryzen are now on pair with Intel regarding IL-2. With less clock speed but more instructions per clock they are able to deliver same results.

How difficult is to overclock the 3600X? it is 4.4 your max freq?

November 30, 2019

48 minutes ago, whitav8 said:

but did you provide results for the settings exactly as chiliwili69 established at the very top of this thread?

Yes, the HIGH run is the one with the settings as per the instructions.

49 minutes ago, whitav8 said:

I would like your opinion of the usefulness of adding a couple runs (as I did above) with our VR graphical settings but not running in VR mode and then the run with VR but no ASW/reprojection and SS set at our choice but reported in pixels horizontal x vertical instead of percentage. I realize that the VR results are subject to a lot of variables, but it still is fun to compare

Thanks for running the test and your additional tests with you VR settings in monitor and VR.

I think all info is always valuable to others users with the Reverb. The difficult thing would be to collect all this info in a systematic way, since every VR user would use their own VR settings. So it would be difficult to compare later.

I think the most practical thing would be for you to run the benchmark in VR using the specified settings of the benchmark, even if this would produce a lower value. You are free to use the SS you want. The column on the right was created for that purpose.

Another thing we could do is to define a different lower settings to run the test at VR. But here we would need to define a common settings. I am open to suggestions but we would need to agree on "the most common settings for VR". (those settings will need to be changed from the IL-2 game, not modifying settings files)

November 30, 2019

I have been playing with the clouds setting. Apparently there is very little influence in this test on monitor 1080p.

Clouds low
Frames: 22688 - Time: 180000ms - Avg: 126.044 - Min: 83 - Max: 216

Clouds medium
Frames: 22419 - Time: 180000ms - Avg: 124.550 - Min: 82 - Max: 218

Clouds High
Frames: 22617 - Time: 180000ms - Avg: 125.650 - Min: 81 - Max: 215

Clouds extreme
Frames: 22207 - Time: 180000ms - Avg: 123.372 - Min: 82 - Max: 217

My "theory" on that is that the clouds calculation is really executed by the GPU, and in this particular case there is room in the GPU load to accomodate any kind of cloud.

Here the graph:

5 hours ago, Alonzo said:

Now for fun, let's run the CPU at 4.5 and 4.0 Ghz instead.


CPU freq:    4.5 Ghz        (hyper-threading disabled, 6-cores only)
STMark:      2723
Benchmark:   23668 / 131.489 / 90 / 216

CPU freq:    4.0 Ghz        (hyper-threading disabled, 6-cores only)
STMark:      2421
Benchmark:   21464 / 119.244 / 82 / 197

Thanks, I have updated these additional runs. The STMark is along the lines of OCvsSTMark tab in the sheet. So it is coherent.

But the fps you achieve are about 12 to 18 fps higher than expected (based in the few runs we have right now). That´s strange. Could you be sure you run these at the specified settings?

November 30, 2019

I have been running Extreme Clouds since they became available.

November 30, 2019

2 hours ago, whitav8 said:

@jaws2002,

Great to see how well AMD Ryzen is performing! One question, you show the results for all different graphic settings (low, balanced,...) but did you provide results for the settings exactly as chiliwili69 established at the very top of this thread?

I setup the game settings exactly as in the settings screenshot Cili provided. the only difference is gamma correction. Mine is 1.0. I dopn't think that setting changes anything about performance, it's just something you setup based on the monitor brightness.

I adjusted all the settings like in the screenshot above, clicked on "high", restarted the game so the change takes effect, and then ran the benchmark.

after that one, i just changed settings in game to the other presets, (ultra, balanced and low) and repeated the benchmark, with restarting the game in between so the settings take effect.

I think we can do a few extra things to improve results compatibility. We should agree on identical settings in the Nvidia control panel, and all run the benchmark with the same settings. I usually run at 1440p, maybe i dropped some settings in there and maybe my results look better than they should.

2 hours ago, chiliwili69 said:

Thanks for running the first new ryzen series! As we suspected the Ryzen are now on pair with Intel regarding IL-2. With less clock speed but more instructions per clock they are able to deliver same results.

How difficult is to overclock the 3600X? it is 4.4 your max freq?

That's the maximum freq i see in the CPU boosting, in CPUZ. It fluctuates a lot and the maximum boost moves from core to core to keep the thermals balanced, so it's not something i see all the time. I don't run with "all core" overclock, because that limits the maximum single core boost. The base clock for this cpu is 3.8GHz.

I ran a slight overclock. I think i just clicked on the "game mode" in the Ryzen master and restarted the computer and it applied a light overclock.

The single core performance is pretty decent on this little bastard.

Edited November 30, 2019 by Jaws2002

November 30, 2019

Hello, maybe that's abit off topic, but rather than fps, that's more stutter that currently bother me...

It should be nice to have a protocol to measure them.

November 30, 2019

3 hours ago, chiliwili69 said:

My "theory" on that is that the clouds calculation is really executed by the GPU, and in this particular case there is room in the GPU load to accomodate any kind of cloud.

That could be completely true. For me (having the weakest overall system posted so far), the relative cost of increasing the cloud detail increases as the overall game detail setting increases - and that seems to support your hypothesis (with my video card being less able/capable of taking on the higher detail when it is already working harder at the higher game settings).

December 1, 2019

7 hours ago, chiliwili69 said:

[Alonzo tests at 5.0, 4.5 and 4.0 ghz]

But the fps you achieve are about 12 to 18 fps higher than expected (based in the few runs we have right now). That´s strange. Could you be sure you run these at the specified settings?

My test procedure was to run the main test as you specified, 3 times (first set of results). Then I rebooted into BIOS, set 4.5ghz, ran tests again. Then rebooted, set 4.0, ran tests again. Unless something changed underneath me somehow, the only thing changing between those runs is the CPU speed. I do have the RTX 2080 SC Ultra, which I think is a bit better than a 1080ti but not much. It has a good beefy cooler so with a short 3 minute test it's probably doing well on the GPU boost. It might account for the extra FPS.

December 1, 2019

I think some nvidia CP settings could change results? Quality and power settings maybe standardisation could be recommended here?

I normally set to high quality

Perhaps default nvidia settings with 'prefer maximum performance' only changed ?

Will do benchmark when have a chance

Cheers, Dakpilot

December 1, 2019

16 hours ago, SharpeXB said:

I suppose it’s a safe assumption that whichever hardware performs better on a track will perform better live, or it’s the only tool that there is. But it’s an easy test to do. Just run a live mission with Fraps benchmarking it and then run the track of that same gameplay again and benchmark that. Then compare.

Well, I have just followed the procedure you indicated:

1. Start Fraps with "P" key and 3 minutes

2. Launch IL-2, QMB same as test, then Press "Record" key , "P" to unpause (and run fraps) and "A" for autopilot (I am not use to play in monitor ? ). The AI shutted down two bombers.

3. Load later the recorded flight with fraps

4. Compare the fps files. I had to shift in time 3 seconds the recorded flight to synchronize the shape of the curve.

The results:

GAME LIVE (With AI of my plane and recording the flight): 157.3 avg fps

RECORDED FLIGHT: 155.5 avg fps

As you can see the shapes are very similar, but sometimes the Game is above, sometimes the Record is above. I can not find an explanation for this small differences.

The shooting to bombers occurs in 10-20 second and 70-80.

The AI made two loops, at 25-35 and at 85-90.

In any case, I think that a Flight record is a valid way to measure the IL-2 performance, even if they are not exactly exactly the same.

13 hours ago, Jaws2002 said:

We should agree on identical settings in the Nvidia control panel, and all run the benchmark with the same settings

1 hour ago, Dakpilot said:

I think some nvidia CP settings could change results? Quality and power settings maybe standardisation could be recommended here?

I normally set to high quality

Perhaps default nvidia settings with 'prefer maximum performance' only changed ?

thanks for the feeback!

In VR I think that only the power mode of the NVIDIA affects the performance. I always have that mode to "Prefer Maximum performance".

I think that in monitor there cloud be more parameters which affects the IL-2 game. I will put in the first post my NVidia settings (just to put a common base, perhaps not the optimum but just a common base).

In any case, since this is a track which is mostly limiting the CPU I think the the Nvidia control panel will not affect too much the results, but it would be good to set a common base for testing.

December 1, 2019

13 hours ago, c6_lefuneste said:

Hello, maybe that's abit off topic, but rather than fps, that's more stutter that currently bother me...

It should be nice to have a protocol to measure them

Yes, the stutter is a very different thing, and perhaps more difficult to reproduce.

A couple of days ago I was experimenting this for first time. It was in SP in a QMB in a solo flight with I-16 plane in spring Rheinland map. I was glad to see them since I was intrigued why I was not having them before. (that´s weird way of being happy with a problem). So in future I will dig on that problem too.

With this test I wanted first determine how every setting affects CPU and GPU on monitor. With that info we can have a base knowledge to be used later in monitor and VR for other problems.

December 1, 2019

**************************************************************************************

Motherboard: ASUS ROG MAXIMUS XI APEX
CPU: i9 9900k
CPU Freq:    5.3 Ghz AVX=0
L3 cache:        16 MB
RAM type:        DDR4
RAM size:    16 GB
RAM Freq:    4400 MHz
RAM Latency: 17
GPU: 2080Ti MSI, max. core 2130Mhz
STMark: 3216

FRAPS TXT:
2019-12-01 16:30:36 - Il-2
Frames: 31979 - Time: 180000ms - Avg: 177.661 - Min: 119 - Max: 287

BOX 4.002

**************************************************************************************

December 1, 2019

6 hours ago, chiliwili69 said:

The results:

GAME LIVE (With AI of my plane and recording the flight): 157.3 avg fps

RECORDED FLIGHT: 155.5 avg fps

Good to know, thanks

December 1, 2019

As I have quite a few rigs available right now, I will have to do some testing, when I will have some time for it.

Rigs I have available:

Rig 1:

Ryzen 3900x, MSI Creation X570, 32 gb 3600 cl 16 RAM, RTX 2080ti, x72 Kraken 360 AIO, Fractal design meshify S2, 5 140 mm intake fans, HP Reverb, 4K VA monitor

ABBA 1.0.0.4B, stock - I have tried lots of BIOSes an settings, performance of each vary. This BIOS is not the best in therms of performance - clock during playing IL2 is aprox 4300 Mhz, temps - spikes above 70 degrees Celsius.

Rig 2:

I7 9700K, ASUS ROG STRIX Z390-E, 32 GB 3200 CL 16 RAM, RTX 2080ti, Fractal design Celsius S36 360 AIO, Fractal design meshify S2, 4 140 mm intake fans, RIFT S, 4K TN monitor

5.2 OC, no AVX offset - have only tried ASUS auto OC yet - this is the result. Voltage is too high (up to 1.4 V), temps OK (bellow 70 degrees celsius)

Rig 3:

I7 7700K, Asrock Z270 extreme 4, 16 gb 3200 cl16 RAM, gtx 1080ti, silentium pc grandis cooler, NZXT h440, FHD monitor

4.8 OC, no AVX offset

Rig 4

Threadripper 1920x, asrock x399 taichi, 32 gb 3200 cl 16 4 channel RAM, 3x radeon RX 5700, FHD monitor

stock

__________________________________________________________________________________________________________________________________________________________________________

Based on feelings, rig 2 is still much faster than rig 1 in IL2 (sadly, I really wish my 1st rig was faster :-) ). Even though their single core performance in Cinebecnh si aprox. equal, I get better fps on Intel in IL2. On Intel, I am also able to compress time by 8x, on Ryzen i can do that only by 4x, and sometimes just to 2x time speed increase. Does this depend just on CPU speed?

On rig 2, I am locked on 80 fps most of the time. On rig 1, I am on 45 fps on reverb with reprojection on (perfectly playable, not much artifacts though)

Performance of I7 7700K and 3900x seems to be equal, based just on the feelings. OCed I7 7700K micht be faster in some occasions, 3900x seems to be more stable - less fps drops.

I will try to do some more testing.

December 1, 2019

4 hours ago, WallterScott said:

Motherboard: ASUS ROG MAXIMUS XI APEX
CPU: i9 9900k
CPU Freq: 5.3 Ghz AVX=0

Hey! That´s a beast in all departments, CPU, OC, RAM, GPU!

Your preformance is quite beyond expected. Probably your RAM is helping.

Benchmark for CPU/RAM performance: Remagen 4.002 to 4.005

Recommended Posts

chiliwili69

SCG_Fenris_Wolf

wju

chiliwili69

chiliwili69

Redwo1f

Hartigan

Redwo1f

SharpeXB

Stolle

Redwo1f

Dakpilot

whitav8

Alonzo

Dutch2

dburne

Alonzo

chiliwili69

Alonzo

chiliwili69

whitav8

SharpeXB

Jaws2002

chiliwili69

whitav8

chiliwili69

chiliwili69

chiliwili69

dburne

Jaws2002

C6_lefuneste

Redwo1f

Alonzo

Dakpilot

chiliwili69

chiliwili69

WallterScott

SharpeXB

thermoregulator

chiliwili69

Create an account or sign in to comment

Create an account

Sign in

Browse All

Activity