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

[chiliwili69]

8 hours ago, R3animate said:

Avg: 122.567 - Min: 109 - Max: 166

 

Thank you for runing the test with our second 5800X. Now they are over the Intel line with just stock OC.

 

You get more fps than the another 5800X in the table. This could be due to your NB or Uncore Frequency (Infinity Fabric frequency in AMDs). 

 

Please, if you have VR, don´t wait to run the benchmark in VR with the 1080. We would like to know how the 5800X performs in VR even with a 1080.

 

Also, the problem of posting startup config files is that they also contain other personal parameters (like IPD) beyond the graphics settings. I think everyone can easily create their own startup config files only once and use later for runnning them multiple times.

[WallterScott]

11 часов назад, 13/JG5_Ehrler сказал:

What resolution has your monitor? Did you use DSR?

Monitor 2K (2560 1440). DSR - x2.25.  What is your result in 3dmark Port Royal?

9 часов назад, R3animate сказал:

This is essentially stock, I can test again at a later point with +200mhz offset (I hit 5040mhz while testing the rig out, holding off on OCing until I see more data online from other 5800x users).

 Great result!! And if you still overclock the processor, with good cooling, it should be even better!

[thermoregulator]

44 minutes ago, chiliwili69 said:

 

Thank you for runing the test with our second 5800X. Now they are over the Intel line with just stock OC.

 

You get more fps than the another 5800X in the table. This could be due to your NB or Uncore Frequency (Infinity Fabric frequency in AMDs). 

 

Please, if you have VR, don´t wait to run the benchmark in VR with the 1080. We would like to know how the 5800X performs in VR even with a 1080.

 

Also, the problem of posting startup config files is that they also contain other personal parameters (like IPD) beyond the graphics settings. I think everyone can easily create their own startup config files only once and use later for runnning them multiple times.

The difference between 5800x tested  won't be due to Infinity fabric - the "slower" 5800x has higher Infinity Fabric frequency (infinity fabric frequency should be linked to RAM frequency). The difference between the two could in RAM timings and subtimings (quite important) and cooling. Remember, Zen 2 and Zen 3 behave more like graphic cards, they boost higher, if there is thermal headroom. For this reason I suggest to provide info on cooling solution, a well as more info about RAM timings in ZEN benchmarks.

 

There is no point in conservative overclocking (fixed frequency) of Zen 2 and Zen 3 for gaming, at least with high end chips like 5800x and higher. Single core performance will decrease that way. Even PBO won't always improve performance in lightly threaded games, sometimes it leads to higher voltage and temperature, which means also lower clocks.

With zen 2 and zen 3 it's sometimes alchemy,  it's very different from the Intel systems. The results are counterintuitive sometimes - power saving features leads to better performance, features that should improve the performance just increase power consumption, voltage and generated heat, while the performance actually decreases.

[LuftManu]

Okay! first OC test CPU

 

I used +100 PBO augment plus 1.100 Vcore SOC and 0.950 VDDP

 

Corsair H110i PRO XT (Maximum performance enabled) Peak of 63 ºC

 

Temps report:  Current / min / max / average

 

Motherboard: Gigabyte Aorus B550 Pro
 CPU:                Ryzen 5 5600X
 CPU Freq:        4.8 Ghz
 L3 cache:        32 MB
 Cores:               6
 Threads:           12
 RAM type:        DDR4

NB freq:             1800 -> this is Infinity fabric, right?
 RAM size:        16 GB
 RAM Freq:       3600 MHz 
 RAM Latency: 16
 GPU:                 Gtx Asus Strix 1080

 

 

Test done twice. Looks like being chillier helped on the first one. The second one seemed more stable and in a avg temp.

2020-11-23 11:32:52 - Il-2
Frames: 7662 - Time: 60000ms - Avg: 127.700 - Min: 111 - Max: 175

 

2020-11-23 11:36:53 - Il-2
Frames: 7560 - Time: 60000ms - Avg: 126.000 - Min: 105 - Max: 170

Edited November 23, 2020 by LF_Gallahad

[SCG_Fenris_Wolf]

Spoiler

 

The tests must run in Fullscreen mode.

 

Using DSR to get a 4K resolution on a smaller monitor without Fullscreen mode casts the game way beyond the boundaries of the monitor. Also, this testing method only works on 

1080p or 4K monitors.

It does not work on widescreen monitors, as you won't get 3840x2160 or a comparable resolution.

 

 

Edit:

Regarding GPU test: Il-2 crashes when using DSR from 1080p towards 4K and full screen. For wide-screen it doesn't fit a resolution at all. When not selecting fullscreen it casts the game window way beyond the monitor so you cannot recover the game without editing startup.cfg , nor select any menu anymore.

 

The GPU testing method doesn't work unless you have a real 4K screen available. IL-2 even locks up Windows with using DSR to 4k+fullscreen. Best delete this asap.

Edited November 23, 2020 by SCG_Fenris_Wolf

[56RAF_Roke]

Motherboard: MSI B450 Gaming Pro Carbon AC

CPU: Ryzen 5 3600  (stock 4.2)

L3 Cache: 2x 16MB

Cores: 6

Threads: 12

Ram Type: DDR4 3200Mhz

NB Freq: 1600

RAM latency: 14

GPU: Palit 3080

 

CPU

 

2020-11-18 21:49:15 - Il-2

Frames: 4234 - Time: 60000ms - Avg: 70.567 - Min: 62 - Max: 95

 

GPU

 

2020-11-18 21:44:03 - Il-2

Frames: 6240 - Time: 60000ms - Avg: 104.000 - Min: 89 - Max: 133

 

Can you say bottleneck?   

 

Which is fine, I'll be moving to a 5600X soon I hope. Interesting to see that even in the GPU test my 3080 is still being held back noticably relative to some of the 1080s here.

 

VR test to follow.

 

Roke

 

 

[R3animate]

5 hours ago, chiliwili69 said:

 

Thank you for runing the test with our second 5800X. Now they are over the Intel line with just stock OC.

 

You get more fps than the another 5800X in the table. This could be due to your NB or Uncore Frequency (Infinity Fabric frequency in AMDs). 

 

Please, if you have VR, don´t wait to run the benchmark in VR with the 1080. We would like to know how the 5800X performs in VR even with a 1080.

 

Also, the problem of posting startup config files is that they also contain other personal parameters (like IPD) beyond the graphics settings. I think everyone can easily create their own startup config files only once and use later for runnning them multiple times.

 

Got it -- sounds good. BIOS update today fixed my Infinity Fabric -- instead of 1600 it'll be going up to 1800 (3600mhz mem). I'll re-run a regular CPU run and a VR test run and post results on both with the new 3600mhz mem.

 

My thoughts on differences between similar Ryzen chips: I have two* dual-ranked 16gb dimms in dual channel -- meaning I have quad rank memory. The other 5800x user may not. This will likely show a small difference, 5%-ish. 

Edited November 23, 2020 by R3animate

[wju]

CPU TEST:
2020-11-23 18:25:08 - Il-2
Frames: 5040 - Time: 60000ms - Avg: 84.000 - Min: 72 - Max: 113

2020-11-23 18:27:14 - Il-2
Frames: 5013 - Time: 60000ms - Avg: 83.550 - Min: 71 - Max: 112

 

GPU TEST:
2020-11-23 18:32:27 - Il-2
Frames: 5508 - Time: 60000ms - Avg: 91.800 - Min: 73 - Max: 108

2020-11-23 18:34:10 - Il-2
Frames: 5574 - Time: 60000ms - Avg: 92.900 - Min: 78 - Max: 110

 

VR TEST: REverb G1 SS100%, i.e. 2200x2160x2 = 9.504Mpix
2020-11-23 18:41:02 - Il-2
Frames: 3203 - Time: 60000ms - Avg: 53.383 - Min: 43 - Max: 69

2020-11-23 18:42:44 - Il-2
Frames: 3302 - Time: 60000ms - Avg: 55.033 - Min: 45 - Max: 72

MB:                 MSI X299 GAMING PRO CARBON AC
CPU:               i7 7820x  
CPU Freq:       4.7 Ghz
L3 cache:       11 MB
Cores:             8
Threads:        16
RAM type:      DDR4  Quad
RAM size:       32 GB
Uncore Freq:  2800 MHz
DRAM Freq:   1733 MHz , it is quad, so 4x1733 ??
RAM Latency: 16
GPU:                  1080Ti

 

Edited November 23, 2020 by wju

[Jaws2002]

A test with My current setup. 3950x/overclocked 1080TI. I ordered a 5950x with my local suppliers. It won't be here until about mid December but I'm not that desperate.

My local supplier scored a 5900x and said if I really want it he'll give it to me. I was very tempted, but I'll hold for the 5950x.

 

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                      16

NB frequency    1800MHz

GPU                      Gigabyte Aorus 1080Ti (factory overclocked)

 

CPU test:  

 

2020-11-22 18:10:14 - Il-2
Frames: 5205 - Time: 60000ms - Avg: 86.750 - Min: 75 - Max: 119

2020-11-22 18:12:48 - Il-2
Frames: 5190 - Time: 60000ms - Avg: 86.500 - Min: 75 - Max: 119

2020-11-22 18:25:16 - Il-2
Frames: 5204 - Time: 60000ms - Avg: 86.733 - Min: 77 - Max: 118

 

GPU 4k test (with DSR to get the 4k resolution.):

 

2020-11-23 09:57:32 - Il-2
Frames: 5777 - Time: 60000ms - Avg: 96.283 - Min: 80 - Max: 114

2020-11-23 13:19:31 - Il-2
Frames: 5794 - Time: 60000ms - Avg: 96.567 - Min: 80 - Max: 114

 

 

    

Edited November 23, 2020 by Jaws2002

[chiliwili69]

9 hours ago, thermoregulator said:

The difference between 5800x tested  won't be due to Infinity fabric - the "slower" 5800x has higher Infinity Fabric frequency (infinity fabric frequency should be linked to RAM frequency).

 

From the two 5800X tested of the table:

The slower CPU test (116 fps) has an Infinity Fabric freq (shown in CPU-Z as uncore freq) of 1800MHz. (RAM freq is 3600, and CL 16)

The faster CPU test (122.57 fps) has an Infinity Fabric freq (shown in CPU-Z as uncore freq) of 3200MHz (RAM freq is 3200, and CL16).

 

That´s why I was thinking that it could be an explanation.

 

[thermoregulator]

6 minutes ago, chiliwili69 said:

 

From the two 5800X tested of the table:

The slower CPU test (116 fps) has an Infinity Fabric freq (shown in CPU-Z as uncore freq) of 1800MHz. (RAM freq is 3600, and CL 16)

The faster CPU test (122.57 fps) has an Infinity Fabric freq (shown in CPU-Z as uncore freq) of 3200MHz (RAM freq is 3200, and CL16).

 

That´s why I was thinking that it could be an explanation.

 

Zen 3 chips max IF speed is 2000 Mhz, I believe the IF frequency of faster chip was 1600 Mhz, which corresponds to 3200 Mhz RAM speed.

[chiliwili69]

11 hours ago, thermoregulator said:

infinity fabric frequency should be linked to RAM frequency

 

From my very limited knowledge.

In Intel CPUs the NB freq (Norton Bridge, also called Uncore freq in Intel CPUs) is the frequency of the non-core parts of the CPU. This is L3 cache, memory controller, etc.

We were including that value in our previous benchmarks to see if there was any influence of that over the performance.

 

With Zen2 line AMD introduced the term "Infinity Fabric"(IF) for the bus comunication of the core with the non-core components including memory controller.

 

In the past remagen Test, we were using the NB freq colum for the IF freq, and here we continue to do the same.

 

The IF freq is shown by CPU-Z in the Uncore freq (or NB freq) and it is a key item for having better performance. 

 

So you have two frequencies to play with: The IF freq (FCLK) and the RAM freq (MCLK), you can run it at 1:1 ratio  (IF1800-RAM1800 but means RAM 3200 because Double Rated) 

 

 

 

Edited November 23, 2020 by chiliwili69

[thermoregulator]

5 minutes ago, chiliwili69 said:

 

From my very limited knowledge.

In Intel CPUs the NB freq (Norton Bridge, also called Uncore freq in Intel CPUs) is the frequency of the non-core parts of the CPU. This is L3 cache, memory controller, etc.

We were including that value in our previous benchmarks to see if there was any influence of that over the performance.

 

With Zen2 line AMD introduced the term "Infinity Fabric"(IF) for the bus comunication of the core with the non-core components including memory controller.

 

In the past remagen Test, we were using the NB freq colum for the IF freq, and here we continue to do the same.

 

The IF freq is shown by CPU-Z in the Uncore freq (or NB freq) and it is a key item for having better performance. 

 

So you have two frequencies to play with: The IF freq (FCLK) and the RAM freq (MCLK), you can run it at 1:1 ratio like the faster CPU test (IF3200-RAM3200) or with 1:2 like the slower CPU test (IF1800-RAM3600). 

 

 

 

I believe actual frequency of 3600 DDR RAM is 1800 Mhz (DDR means "double data rate", or something like that). So 3600 DDR RAM frequency is 1800 Mzh = 1800 IF frequency.

[chiliwili69]

9 hours ago, thermoregulator said:

The difference between the two could in RAM timings and subtimings (quite important) and cooling.

 

In the past we were just keeping record of the CL but no the subtimings. We saw that the important factor was the RAM frequency and less the CL.

I don´t know if we should keep this data. If we demostrate it is relevant for the performance (beside CL), we should then keep track of it.

 

Regarding the freq of AMD CPU you are right. The way OC is done in Intel line (all cores to same fixed freq) might not be optimal for the AMD CPUs.

The cooling solution (and the ambient room temperature and other fans inside the case) in AMD lines might influence the final result.

Keeping track of that could complicate the table. But I am open to request/include whatever we think it would be practical to keep.

[Jaws2002]

 

5 minutes ago, chiliwili69 said:

you can run it at 1:1 ratio like the faster CPU test (IF3200-RAM3200) or with 1:2 like the slower CPU test (IF1800-RAM3600). 

 

Where you see ram speeds of 3200, 3600, 4000 MHz, that's not actual ram speed. the Actual ram speed is half that, because we are talking about DDR Double Data Rate. the memory speed in that second test is actually running at 1800Mhz, in 1:1 with the infinity fabric speed. 

CPUz shows the actual frequency of the memory: The frequency is 1800MHz, but we call it PC 3600.

 

[chiliwili69]

40 minutes ago, thermoregulator said:

Zen 3 chips max IF speed is 2000 Mhz, I believe the IF frequency of faster chip was 1600 Mhz, which corresponds to 3200 Mhz RAM speed

OK, then the IF freq reported by @R3animate shold be wrong (can not be higher than 2000) or maybe CPU-Z is reporting that NB wrongly.

9 hours ago, LF_Gallahad said:

2020-11-23 11:32:52 - Il-2
Frames: 7662 - Time: 60000ms - Avg: 127.700 - Min: 111 - Max: 175

 

2020-11-23 11:36:53 - Il-2
Frames: 7560 - Time: 60000ms - Avg: 126.000 - Min: 105 - Max: 170

 

Hey!, that´s again a good result. And your temps are quite good! You are now at the top!

I will record the second test then.

8 hours ago, SCG_Fenris_Wolf said:

Regarding GPU test: Il-2 crashes when using DSR from 1080p towards 4K and full screen. For wide-screen it doesn't fit a resolution at all. When not selecting fullscreen it casts the game window way beyond the monitor so you cannot recover the game without editing startup.cfg , nor select any menu anymore.

 

The GPU testing method doesn't work unless you have a real 4K screen available. IL-2 even locks up Windows with using DSR to 4k+fullscreen. Best delete this asap.

 

I think you are quite right here but if we limit the GPU test only to 4K owners then the number of potential tests will be lower.

The game also crash when in a 4K monitor (mine) you try to run at 1080p in full screen. My test for 1080p were not done at full screen in my 4K monitor.

 

 

I also have a 1920x1080p monitor and also a 2560x1080p monitor for work. I can try to put them with the PC were I run IL-2 and run the GPU test with DSR in those and compare results.

[chiliwili69]

7 hours ago, 56RAF_Roke said:

Can you say bottleneck?

Thank you for your test.

I think the CPU test is expectable, acording to previous tests of Remagen with Zen2 CPUs.

Clearly the GPU test is being constrained by the CPU, so I think it is not a valid test (When designing the procedure I wanted to design a GPU test not constrained by the CPU). But in your case you have a very unbalanced system CPU vs GPU. So when you get the new 5600X we should see a nice increase in the GPU test.

6 hours ago, R3animate said:

instead of 1600 it'll be going up to 1800 (3600mhz mem)

OK, now you clarify your IF frequency. In your post you put 3600 instead of 1800.

20 hours ago, R3animate said:

VR Test Results - Valve Index / SS @ 100%: (All settings as instructed)

2020-11-23 12:19:32 - Il-2
Frames: 4697 - Time: 60000ms - Avg: 78.283 - Min: 48 - Max: 91

 

I believe I am now GPU bound here in VR more than anything.

 

Wow! thats a nice result in VR. Clearly GPU bounded now in this test.

[R3animate]

17 minutes ago, chiliwili69 said:

Thank you for your test.

I think the CPU test is expectable, acording to previous tests of Remagen with Zen2 CPUs.

Clearly the GPU test is being constrained by the CPU, so I think it is not a valid test (When designing the procedure I wanted to design a GPU test not constrained by the CPU). But in your case you have a very unbalanced system CPU vs GPU. So when you get the new 5600X we should see a nice increase in the GPU test.

OK, now you clarify your IF frequency. In your post you put 3600 instead of 1800.

 

Whoops! Fixed, I meant 1800 Infinity Fabric :). 

Edited November 23, 2020 by R3animate

[chiliwili69]

2 hours ago, wju said:

DRAM Freq:   1733 MHz , it is quad, so 4x1733 ??

Thanks for your test. It is a Quad channel Intel CPU. In the past Remagen test I created a different group for those chips, but here I will put together since they don´t provide a significant adavantage for IL-2.

2 hours ago, Jaws2002 said:

My local supplier scored a 5900x and said if I really want it he'll give it to me. I was very tempted, but I'll hold for the 5950x

 

thanks Jaws2002 for your test with that previous Zen2 which was beated by Intel in the Remagen Tests. Now it will be your revenge when your new monster will be put in your PC.  ?

 

1 hour ago, thermoregulator said:

I believe actual frequency of 3600 DDR RAM is 1800 Mhz (DDR means "double data rate", or something like that). So 3600 DDR RAM frequency is 1800 Mzh = 1800 IF frequency

 

Yes, I has been corrected in the table. It was IF at 1800MHz. 

[Jaws2002]

44 minutes ago, chiliwili69 said:

Thanks for your test. It is a Quad channel Intel CPU. In the past Remagen test I created a different group for those chips, but here I will put together since they don´t provide a significant adavantage for IL-2.

 

thanks Jaws2002 for your test with that previous Zen2 which was beated by Intel in the Remagen Tests. Now it will be your revenge when your new monster will be put in your PC.  ?

 

 

 No. Thank you for putting all this together!   A lot of people are able to make an informed buying decision, because of this benchmarks. 

 

   While in Il-2 and gaming in general, Intel CPUs were the stronger option, in heavy work loads, my 3950x is still a monster. 

 I just want a more complete monster.

 

 

  

 

 

?

   

 

  

Edited November 23, 2020 by Jaws2002

[Gambit21]

6 hours ago, Jaws2002 said:

I ordered a 5950x with my local suppliers. It won't be here until about mid December but I'm not that desperate.

My local supplier scored a 5900x and said if I really want it he'll give it to me. I was very tempted, but I'll hold for the 5950x.

 

 

That's hard core dude...hard core.

[Jaws2002]

51 minutes ago, Gambit21 said:

 

That's hard core dude...hard core.

 

 

I was very tempted to just grab the 5900x and throw it in the case right now. I'm actually surprised I didn't just fold to the temptation. ? 

  In the end it was the lack of Graphics card stock that helped with the decision. I still have to figure out what card to get and actually find one, before I can really enjoy the upgrade.

I'll be fine with what i have for another two three weeks.

 

[SCG_Fenris_Wolf]

 Motherboard: MSI X570 Tomahawk WIFI
 CPU:             AMD Ryzen 9 5900X
 CPU Freq:        4950 MHz
 L3 cache:        2x 32 MBytes
 Cores:           12
 Threads:         24
 RAM type:        DDR4
 RAM size:        2 x 16 GB (Dual Rank)
 Uncore Freq:     1900 MHz FCLK==MCLK
 RAM Freq:        3800 MHz
 RAM Latency:     16-16-16-16-32 
 GPU:             MSI RTX 3080 Trio X Gaming (TDP 102%, +100MHz core, + 250MHz memory)

 

 

CPU Test (PBO disabled, stock speed)

2020-11-23 18:47:27 - Il-2
Frames: 7668 - Time: 60000ms - Avg: 127.800 - Min: 111 - Max: 175

 

GPU Test

<place holder> //DSR stretches image beyond screen. Will test in my living room on 4k TV. I will test RTX 3080 Founders Edition OC vs MSI RTX 3080 Trio X Gaming.

 

VR Test (CPU: PBO disabled, no overclock)

2020-12-04 22:14:55 - Il-2
Frames: 5346 - Time: 60000ms - Avg: 89.100 - Min: 83 - Max: 92

Il-2 2020-12-04 22-14-55-88 fps.rar

Edited December 4, 2020 by SCG_Fenris_Wolf

[chiliwili69]

13 minutes ago, SCG_Fenris_Wolf said:

Frames: 7668 - Time: 60000ms - Avg: 127.800 - Min: 111 - Max: 175

 

Thank Fenris! this is a nice achievement just at stock speed!

Let´s see how the other tests go.

I have run again the test in my PC with the last update IL-2 (v4.503) since my initial test was done with the v4.502.

 

CPU test (from 15-Nov with V4.502)
Frames: 4933 - Time: 60000ms - Avg: 82.217 - Min: 72 - Max: 112

 

CPU test (today with V4.503)
Frames: 4560 - Time: 60000ms - Avg: 76.000 - Min: 66 - Max: 102
Frames: 4572 - Time: 60000ms - Avg: 76.200 - Min: 67 - Max: 103
Frames: 4654 - Time: 60000ms - Avg: 77.567 - Min: 68 - Max: 106

 

I thought the new version was not going to impact the performance but I was wrong. (Or maybe was other Windows updates or I don´t know).

@hsthhsth Could you run again the CPU test with the new IL-2 version 4.503?

 

 

[SCG_Fenris_Wolf]

Internet reports issues with latest Nvidia drivers / SteamVR version. It might be that, and not IL-2's update =)

[chiliwili69]

I have been thinking a bit more about how to configure a procedure to perform the GPU Test correctly for people who has not 4K native monitors (or TVs).

 

I am only thinking about the monitor with 1920x1080 native resolution which has the same proportions than a 4K monitors (just multiple every axis by two, so it is four times the pixels).

 

Other monitor with Ultrawide view (like 2560x1080) hs a different proportion and the FOV will change. In the 4K or in the 1920x1080 the FOV is 67 degrees. So if a different FOV is used there will be more objects in the scene a the performance can not be compared.

So Ultrawide monitors can be used for the CPU test (since they don´t stress GPU), but not for the GPU tests.

 

As Fenris said, if a 1920x1080 native monitor uses the DSR factor=4, then you have available the 3840x2160 resolution in the IL-2 game, but then the image is cropped and you can not access the menus. 

 

But another way to test that is to use a 1920x1080 native monitor with DSR=4 but chose the 1920x1080 resolution in IL-2, so the image is not cropped.

I have run this 3 times each:

 

4K GPU test: Using 4K native monitor
Frames: 5559 - Time: 60000ms - Avg: 92.650 - Min: 75 - Max: 112
Frames: 5617 - Time: 60000ms - Avg: 93.617 - Min: 78 - Max: 110
Frames: 5642 - Time: 60000ms - Avg: 94.033 - Min: 78 - Max: 108

 

4K GPU test: 1920x1080p monitor with DSR x4 and Full screen mode
Frames: 5394 - Time: 60000ms - Avg: 89.900 - Min: 70 - Max: 105
Frames: 5413 - Time: 60000ms - Avg: 90.217 - Min: 74 - Max: 105
Frames: 5432 - Time: 60000ms - Avg: 90.533 - Min: 75 - Max: 106

 

The averages of the three runs show this:

 

In theory it should be the same but it is not. Perhaps it takes extra effort to the GPU to average the 4 pixels in one pixel.

CONCLUSION: So 4K native monitor or TVs has to be used for the GPU test. I will update the instructions.

Also, I will remove all the previous test which used DSR to reach 4K resolution. So we will be sure we compare apples with apples.

Edited November 24, 2020 by chiliwili69

[hsthhsth]

Ok, tested with new version.

 

First test, just ran mission once.

Second test, ran mission three times. Always started from the menu.

Third test, tweaked GPU to OC settings, still CPU test, ran mission 4 times. Last time choose "Restart mission".

 

Seems that FPS is falling during multiple runs. Can't understand why.

20201124.zip

[chiliwili69]

9 hours ago, hsthhsth said:

Seems that FPS is falling during multiple runs. Can't understand why.

 

I think your variations has nothing to do with the way you run the mission (doing restart or started from menu or restarting IL-2 ).

I have plotted your runs:

 

Firstly I assume that you wait the 25 seconds (it is not a exact measure) to allow loading all buildings in the train station (not only the ones near to you but also in the far view).

Just in case you can wait an extra time to be sure.

 

If you continue to have this then I would plot the CPU freq using MSI afterburner during the test. Just to see if during some periods your CPU is throttled due to power limit or temp limit or amps limit.

 

 

[Voxman]

Thanks guys for pulling this together.  I do miss the "Black Death" track as we did back in the old days to fine tune video settings, then OC on the CPU.  Hopefully this will be the new standard.

 

Here's my performance on a system I just built at the beginning of this month.  I haven't OC'd yet.  I have liquid cooling on the CPU, so my intent is to OC sometime this week or month:

 

 
 Motherboard: Gigabyte Z490 Aorus Elite
 CPU:                i7-10700K
 CPU Freq:        4.8 Ghz
 L3 cache:       16 MB
 Cores:           8
 Threads:         16
 RAM type:        DDR4
 RAM size:        2x8GB
 NB Freq:          4300 MHz
 RAM Freq:        2134 MHz
 RAM Latency:      15
 GPU:              Gigabyte RX 5700 XT

2020-11-24 16:54:12 - Il-2
Frames: 4824 - Time: 60000ms - Avg: 80.400 - Min: 70 - Max: 112

 

Just for fun as I run the system at 1440, I ran the mission and here's my result at that resolution and with the 4K settings:

2020-11-24 17:04:22 - Il-2
Frames: 3863 - Time: 60000ms - Avg: 64.383 - Min: 57 - Max: 73

[Gambit21]

15 hours ago, chiliwili69 said:

I have been thinking a bit more about how to configure a procedure to perform the GPU Test correctly for people who has not 4K native monitors (or TVs).

 

 

I have a plan, but need to test it.

Here's the idea. Small "free" map, no player aircraft. Just a ton of complex tanks (which are very expensive on the CPU) and then measuring Time Dilation with subtitle's/timers.

So there would be no FPS so to speak because you would only have the default camera that activates when no player aircraft is used. 

 

You'd hit play, then set your stopwatch according to the subtitle instructions, and measure Time Dilation over 2 minutes or so.

Might not be what you're after as it eliminates measuring FPS, but takes the graphics card out of the equation basically.

 

If you do like that idea, or if you do @SYN_Vander I won't be offended if you build it if you have time.

I might not have time for some days.

Edited November 25, 2020 by Gambit21

[sevenless]

On 11/18/2020 at 11:24 AM, LF_Gallahad said:

Hello, first 5600X Ryzen benchmark. @chiliwili69. 

 

CPU TEST

2020-11-18 11:18:56 - Il-2
Frames: 7155 - Time: 60000ms - Avg: 119.250 - Min: 101 - Max: 166
 

 

Very interesting results for the new AMD CPU. Here are my Intel CPU results with my trusty old Nvidia 980. Still undecided whether a RTX 3070 or AMD 6800 will replace it. I will stay with 1080p for the next 1-2 years for sure.

 

CPU test

2020-11-25
Frames: 5277 - Time: 60000ms - Avg: 87.950 - Min: 73 - Max: 117

 Motherboard: ASUS Maximus XI Hero Wi-Fi
 CPU:                 9900K
 CPU Freq:        5.2 Ghz
 L3 cache:        16 MB
 Cores:              8 
 Threads:          16 
 RAM type:        DDR4
 RAM size:        32 GB
 NB Freq:          4300 MHz
 RAM Freq:       3800 MHz (2 x 1900 Mhz)
 RAM Latency: 19 
 GPU:                 GTX 980

 

Edited November 25, 2020 by sevenless

[SYN_Vander]

5 hours ago, Gambit21 said:

 

I have a plan, but need to test it.

Here's the idea. Small "free" map, no player aircraft. Just a ton of complex tanks (which are very expensive on the CPU) and then measuring Time Dilation with subtitle's/timers.

So there would be no FPS so to speak because you would only have the default camera that activates when no player aircraft is used. 

 

You'd hit play, then set your stopwatch according to the subtitle instructions, and measure Time Dilation over 2 minutes or so.

Might not be what you're after as it eliminates measuring FPS, but takes the graphics card out of the equation basically.

 

If you do like that idea, or if you do @SYN_Vander I won't be offended if you build it if you have time.

I might not have time for some days.

Yes that would be a good setup to stress test cpu. Since there is a range of cpu’s to test it’s perhaps a good idea to spawn more tanks every minute or so. That way you can see after one test the relationship between # of tanks and time dilation; you can see where the threshold is for your system.

However, I also don’t have time the next couple of days or this weekend. Really tied up with my regular job.

[chiliwili69]

7 hours ago, Voxman said:

RAM Freq:        2134 MHz

Thanks for your test. Let´s see how much the OC improve the performance.

Regarding your RAM speed, this is quite low for your CPU.

Did you enable any of the XMP profiles of your RAM? I think that witha memory on 3200 or more your system will improve greatly.

5 hours ago, sevenless said:

Avg: 87.950

Thank you for your test.

Your results is clearly constrained by your GPU. This is something I wanted to avoid for the CPU test. So I marked your result in color.

If you are going to play just at 1080p monitor for 60Hz you are quite fine with that 980 card.

In any case, if you upgrade your GPU you can run again the CPU test and you should achieve a much higher value aligned with other 9900K.

 

6 hours ago, Gambit21 said:

Just a ton of complex tanks (which are very expensive on the CPU)

 

Hey, but I was talking about the GPU test, not the CPU test.

 

The current CPU test I think it is doing OK his job.

 

The current GPU test is also OK, but the only problem is that it needs to be done in 4K monitors (or TVs are also valid). At not all people have 4K monitors. (although probably all have 4K TVs in their homes)

[wju]

On 11/24/2020 at 11:26 AM, chiliwili69 said:

I have been thinking a bit more about how to configure a procedure to perform the GPU Test correctly for people who has not 4K native monitors (or TVs).

 

my GPU test was done not on native 4K but 2560x1440 native with DSR;

windows screen resolution set to 3540*2160 and dtto in IL-2 setting, to be consistent, you should better delete my GPU record form the overwiev

Edited November 25, 2020 by wju

[sevenless]

4 hours ago, chiliwili69 said:

Thank you for your test.

Your results is clearly constrained by your GPU. This is something I wanted to avoid for the CPU test. So I marked your result in color.

If you are going to play just at 1080p monitor for 60Hz you are quite fine with that 980 card.

 

Thanks to you for preparing the spreadsheet. Gives a really good overview. Yes an upgrade is due for the GPU. Hope to get in the 2080TI ballpark in 1080p with either the 3070 or 6800, once I actually can buy them. Both still are out of stock here in Germany. Even the ASUS online shop doesn´t have them available. LOL!

[Voxman]

6 hours ago, chiliwili69 said:

Regarding your RAM speed, this is quite low for your CPU.

Did you enable any of the XMP profiles of your RAM?

I haven’t done this.  This is the first rig I’ve built in about 9 years, and a hiatus that long from building rigs. 
I haven’t delved into this rigs settings as I’m trying to find time.  Are XMP settings found in the BIOS?  The RAM I’m using is Patriot Viper Steel Series DDR4 16GB (2 x 8GB) 4400MHz    
 

 

[Voxman]

I went into my BIOS and found the XMP settings and set to a profile

Two screenshots...before and after.

*Edit:  Added a third screenshot with a 2nd XMP setting within BIOS.

 

Hopefully today, or within the next few days, I'll start OC'ing this rig to optimize.  One thing is, since I'm coming back from the caveman days of OCing and new technology, is there a recommended guide on OCing properly?  I do have some old bookmarks from old websites and prior to building my rig referred to them, but some have gone over 9 years.  Also to add, I only play IL2 forgotten battles.  No other games.

 

Edited November 25, 2020 by Voxman
added another screenshot

[Voxman]

Running the tests again and using XMP profiles I get the following:
 

xmp profile 1 RAM BIOS:

1080P CPU test
2020-11-25 07:23:29 - Il-2
Frames: 5305 - Time: 60000ms - Avg: 88.417 - Min: 76 - Max: 121
GPU test:
2020-11-25 07:31:00 - Il-2
Frames: 3858 - Time: 60000ms - Avg: 64.300 - Min: 56 - Max: 73

XMP profile 2 in RAM BIOS

 

1080CPU Test
2020-11-25 07:51:27 - Il-2
Frames: 5156 - Time: 60000ms - Avg: 85.933 - Min: 75 - Max: 116

GPU test
2020-11-25 07:41:13 - Il-2
Frames: 3840 - Time: 60000ms - Avg: 64.000 - Min: 56 - Max: 72

 

A nice little bump in the CPU test for sure.

Thank you for pointing out the XMP profile setting!

 

[Gambit21]

7 hours ago, chiliwili69 said:

Hey, but I was talking about the GPU test, not the CPU test.

The current CPU test I think it is doing OK his job.

 

Ahh...OK. 

I still might make it just out of curiosity as it will almost completely isolate the CPU.

[chiliwili69]

3 hours ago, Voxman said:

Added a third screenshot with a 2nd XMP setting within BIOS

This third has a better NB freq. You can try to run the CPU benchmark with this config to see if there is an impact before OC the CPU.

 

Sorry I have seen you did it already

2 hours ago, Voxman said:

 

A nice little bump in the CPU test for sure

 

The NB freq of your first profile is quite low (800). 

The XMP profiles are always an easy a free upgrade. Not neccesarily the higher freq delivers the more fps. Also latencies and timmings.

 

3 hours ago, Voxman said:

is there a recommended guide on OCing properly? 

 

I learnt from this forum some years ago. I wrote down the little (but enough) I learnt:

https://forum.il2sturmovik.com/topic/29881-overclocking-a-4790k-for-better-bos-performance/

 

but there are other guides like this one also valid for your 10700K:

https://www.overclockers.com/how-to-overclock-the-i9-10900k-a-guide-for-taming-the-beast/

 

First thing you need is a good CPU cooler. I prefer the AIO water cooling. Zero problems so far.

Edited November 25, 2020 by chiliwili69