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

[chiliwili69]

11 hours ago, FoxbatRU said:

I dig deeper into the PC settings a little. And added a 4K test

Thanks, you squeezed it even more, you have avery good performance in both test!

[Daisy_Blossom]

About to install my new 3080 so I figured I'd do a benchmark right before and right after. Here's the before:

Motherboard: Gigabyte Z370 Aorus gaming 7
CPU:               8700K (delidded) 
CPU Freq:       5.1 Ghz
L3 cache:       12 MB
Cores:             6
Threads:         12
RAM type:      DDR4
RAM size:       32Gb (4x8GB)
NB Freq:         4000 MHz
RAM Freq:      3200 MHz
RAM timings:  14-14-14-34-560
Ram type: G.Skill F4-3200C14D-16GTZR (x2)
GPU:      EVGA 1080ti SC gaming (11G-P4-6393-KR)
HMD: Valve Index
Cooling: Full custom water loop CPU + Mobo + GPU

 

GPU drivers: 497.09

IL2 Version 4.701, Benchmark V6

 

CPU Test 1080p:

Frames: 5713 - Time: 60000ms - Avg: 95.217 - Min: 84 - Max: 129

 

VR Test 1: (90hz @ 106 super-sampling)

Frames: 4350 - Time: 60000ms - Avg: 72.500 - Min: 51 - Max: 91

 

GPU Passmark (3d graphics mark): 17247

[chiliwili69]

14 hours ago, Daisy_Blossom said:

VR Test 1: (90hz @ 106 super-sampling)

Frames: 4350 - Time: 60000ms - Avg: 72.500 - Min: 51 - Max: 91

 

Thanks. Noted. The biggest difference will be inthe VRTest2, if you can measure it.

[Daisy_Blossom]

On 12/23/2021 at 1:55 AM, chiliwili69 said:

 

Thanks. Noted. The biggest difference will be inthe VRTest2, if you can measure it.

 

Unfortunately I already swapped the cards. I'll post the new data as soon as I'm home from Xmas travels. Do you think I'll have much improvement in real-world VR flying with the valve index and the new 3080 without upgrading my CPU?

[chiliwili69]

12 hours ago, Daisy_Blossom said:

Do you think I'll have much improvement in real-world VR flying with the valve index and the new 3080 without upgrading my CPU?

 

It depends on the settings you are using and the Hz mode.

 

The Index 80Hz mode is a great advantage in CPU and GPU terms.

 

With the 3080 and the new clouds you will not be bottlenecked by the GPU in any scenario. You would be able to raise the clouds to Extreme and the SS% SteamVR up to 140% or 150% and also use MSAAx2 or x4.

 

On the other hand, you will need to carefully select the settings that affects the CPU (shadows, mirror, canopy reflections, )

 

Using fpsVR to test your settings while you play is very instructive since you see the frametimes of both CPU and GPU.

[TheRealBullit]

On 12/14/2021 at 9:51 AM, TheRealBullit said:

I will be replacing my 9700K with a 12700k soon.
So I tested my system to see the differences.
Here are the tests in advance with the 9700K clocked at 4800 Mhz

 

Motherboard:      ASRock Z390 Taichi

CPU:                     9700k
CPU Freq:            4800 MHz
Cores:                  8
Threads:              8
RAM size:            32Gb     (2x16GB)
RAM Freq:           3200 MHz (Dual channel)
NB Freq:              4300 MHz (Uncore Frequency)
RAM timings:       16-18-18-38

GPU:                    3090
Headset:             Reverb G2

 

CPU Test:
Frames: 5562 - Time: 60000ms - Avg: 92.700 - Min: 81 - Max: 127

 

VR Test1:
Frames: 3738 - Time: 60000ms - Avg: 62.300 - Min: 50 - Max: 92 

 

VR Test2:
Frames: 3654 - Time: 60000ms - Avg: 60.900 - Min: 50 - Max: 87

 

 

The new processor is here.

So far, neither CPU nor RAM is overclocked.

With these results, the question arises whether an overclocking would be necessary at all?

Obviously, the GPU seems to be the bottleneck.

 

Motherboard:      ASUS TUF Gaming Z690-Plus D4

CPU:                     12700K
CPU Freq:            5000 MHz
Cores:                  8+4
Threads:              20
RAM size:            32Gb     (2x16GB)
RAM Freq:           2133 MHz (Dual channel)
NB Freq:              3600 MHz (Uncore Frequency)
RAM timings:       15-15-15-36

GPU:                    3090
Headset:             Reverb G2

 

CPU Test:

Frames: 6576 - Time: 60000ms - Avg: 109.600 - Min: 97 - Max: 155

 

VR Test1:
Frames: 3962 - Time: 60000ms - Avg: 66.033 - Min: 54 - Max: 91 

 

VR Test2:

Frames: 3717 - Time: 60000ms - Avg: 61.950 - Min: 49 - Max: 87

Edited December 25, 2021 by TheRealBullit

[TheRealBullit]

It looks like the RAM speed has a big impact on performance. Here is the result with a RAM speed of 3800 MHz:

 

CPU Test:

Frames: 7163 - Time: 60000ms - Avg: 119.383 - Min: 108 - Max: 166

 

VR Test2:

Frames: 4457 - Time: 60000ms - Avg: 74.283 - Min: 57 - Max: 91

 

Edited December 25, 2021 by TheRealBullit

[DBCOOPER011]

Below is what I got on my daily driver settings...

 

Motherboard:   ASUS Prime Z690-A

CPU:                 12900K
CPU Freq:        Adaptive (54X2,53X3,52X5,51X8 TVB+1,E-Cores=Auto)
Cores:              8+8
Threads:          24
RAM size:        32Gb     (2x16GB)
RAM Freq:       5400 MHz (Quad channel)
NB Freq:          4700 MHz
RAM timings:  34-39-36-48

GPU:               3090 Hybrid
Headset:        Reverb G2, Index (Both 90Mhz)

 

CPU Test:
Frames: 8063 - Time: 60000ms - Avg: 134.383 - Min: 118 - Max: 180

 

VR Test1:
(G2)Frames: 5225 - Time: 60000ms - Avg: 87.083 - Min: 70 - Max: 91
(Index) Frames: 5388 - Time: 60000ms - Avg: 89.800 - Min: 87 - Max: 92

 

VR Test2:
(G2)Frames: 5121 - Time: 60000ms - Avg: 85.350 - Min: 66 - Max: 91
(Index) Frames: 5393 - Time: 60000ms - Avg: 89.883 - Min: 87 - Max: 92

 

[dburne]

5.4 GHz ram is very impressive.

[chiliwili69]

7 hours ago, TheRealBullit said:

12700K

Thank you. It is the first 12700K tested here. I believe it should provide results close to the 12900K.

 

As you check it out, the RAM frequency plays a very important role. You should always active the faster XMP profile of the RAM in your BIOS. I believe it is 3800Mhz of DDR4, right?

 

Perhaps you could do the CPU test at 3800Mhz, so we can evaluate how well the 12700K does with DDR4 memory.

 

Also, you got 108 fps in the VRTest1. But VR is capped to 90fps, maybe is a typo error.

 

I am not familiar with all the Turbo and OC possibilities of the new Alder Lake, but having a good CPUTest results will help in your VR in-game performance.

 

 

[chiliwili69]

1 hour ago, DBCOOPER011 said:

RAM Freq:       5400 MHz (Quad channel)

 

Thank you for your detailed tests with G2 and Index. You got a very good performance in VR for both devices. Your minimun value is 87 in both tests with the Index!!

Perhaps I will need to think to raise the bar for the VRTest1 &Test2, for example increasing the graphics settings.

 

Regarding the DDR5 memory, each stick has 2 data channel of 32-bits, so you use 4x32bit=128bit data, with the advantage to access 4 simultaneosly.

With DDR4 memory, each stick has 1 data channel of 64bits, so 2x64bit=128bit data, which is the same total, but only 2 access.

 

With the previous Remagen test, we tested some processors (i9-7900X) with "true" Quad-Channel with DDR4, 4x64bit=256bit total. Although they were not providing an advantage in IL-2 over the Dual-Channel DDR4.

 

Your adaptive CPU settings is also interesting. How do you know that IL-2 is always running in the two cores with the highest speed? ie 5.4GHz

 

 

 

[TheRealBullit]

1 hour ago, chiliwili69 said:

Thank you. It is the first 12700K tested here. I believe it should provide results close to the 12900K.

 

As you check it out, the RAM frequency plays a very important role. You should always active the faster XMP profile of the RAM in your BIOS. I believe it is 3800Mhz of DDR4, right?

 

Perhaps you could do the CPU test at 3800Mhz, so we can evaluate how well the 12700K does with DDR4 memory.

 

Also, you got 108 fps in the VRTest1. But VR is capped to 90fps, maybe is a typo error.

 

I am not familiar with all the Turbo and OC possibilities of the new Alder Lake, but having a good CPUTest results will help in your VR in-game performance.

 

 

I corrected the values from the VR1 test it was a copy & paste error.

 

Also the CPU test with 3800 MHz Ram Timing is added.

 

The Ram is a 4600 CL19-26-26-46, so I can't just load the XMP profiles, but have to test manually. Currently it runs with 3800 CL16-20-20-36.

[RAAF492SQNOz_Steve]

7 hours ago, TheRealBullit said:

It looks like the RAM speed has a big impact on performance. Here is the result with a RAM speed of 3800 MHz:

 

CPU Test:

Frames: 7163 - Time: 60000ms - Avg: 119.383 - Min: 108 - Max: 166

 

VR Test2:

Frames: 4457 - Time: 60000ms - Avg: 74.283 - Min: 57 - Max: 91

 

 Re: VR2 test result.  I suspect that something else needs to be tuned on your PC or one of the VR2 IL2 config settings has been missed (set too high and negative impact to your average fps?).

The VR2 avg of 74.3 fps, while a big improvement over your old PC, is lower than I would have expected for the 12700k. Would have expected a result much closer to that posted by DBCOOPER011 i.e. 80+ fps in VR2 result. 

 

As you have just discovered, overclocking your RAM from stock frequencies, can give you a big boost. Now that your are running at 3800 MHz there may not be much more performance gain to be obtained in that area without compromising PC stability.

 

Hope you come across the "missing ingredient" that gives you some extra fps as it really contributes to the smoothness of the in VR flying experience with the G2.

[102nd-YU-cmirko]

Merry Christmas all ^^

wishing you all the least amount of stutters and even better virtual clouds in the year ahead

 

 

Since I haven't seen much AMD cards, I'll post my 6900xt results - it's really super nice to see how many super fast PC's are on this forum and in the virtual skies

 

 

Motherboard:     Gigabyte X570 Aorus Pro

CPU:             Ryzen 5800X    
CPU Freq:    ~4950Mhz (PBO curve by Hydra https://www.igorslab.de/en/project-hydra-oc-sandbox-for-zen3-cpus-freeware-download/3/)
Cores:          8
Threads:       16      
RAM size:     32Gb (4x8GB)
RAM Freq:    3800 MHz
NB Freq:       1900 MHz (UCLK)
RAM timings:16-16-16-36
GPU:             XFX RX6900XT Speedster ZERO (EKWB and good cooling loop lets me set 2200Mhz min frequency GPU core, 2400Mhx max, 2200Mhz GPU ram speed)
HMD:             HP G2 (rev1, new cable) 
OS:                Win10
Cooling loop:-> AlphaCool Extreme 280 -> CPU -> 360mm -> EKWB -> NB wb -> 

VR2 original test
2021-12-26 11:19:35 - Il-2 - 90hz
Frames: 3964 - Time: 60000ms - Avg: 66.067 - Min: 52 - Max: 91

2021-12-26 11:26:35 - Il-2 - 60Hz 
Frames: 3591 - Time: 60000ms - Avg: 59.850 - Min: 56 - Max: 61
for posterity all testing was done with original il2 openvr_api.dll ^^

 

 

CPU TEST 

2021-12-27 11:56:38 - Il-2 
Frames: 5937 - Time: 60000ms - Avg: 98.950 - Min: 87 - Max: 137

 

VR1 original test 60Hz

2021-12-27 12:08:36 - Il-2 - 60Hz
Frames: 3590 - Time: 60000ms - Avg: 59.833 - Min: 56 - Max: 61

 

VR1 original test 90Hz 

2021-12-27 12:18:13 - Il-2 - 90Hz
Frames: 3821 - Time: 60000ms - Avg: 63.683 - Min: 50 - Max: 91

 

-----
VR2 "my eyecandy" test

I play on 60Hz + ASW +120% SS + latest FSR/NIS with lots of eyecandy, online mostly

 

test done with "my eyecandy":
high preset, shadows high, mirrors medium, distant x4, canopy off, 130km, blurred filter, medium roughness, grass off, extreme clouds, msaa off, 4k on, distant buildings on
60Hz + eyecandy + 120% SS + fholger openvr_api.dll ("fsr": "enabled": true,"useNIS": true, "renderScale": 0.77, "sharpness": 0.99, "radius": 0.4, "applyMIPBias": true, "debugMode": false)

 

2021-12-26 12:23:41 - Il-2 - 60Hz + upper settings
Frames: 3421 - Time: 60000ms - Avg: 57.017 - Min: 44 - Max: 61

 

Edited December 27, 2021 by 102nd-YU-cmirko

[Voyager]

@102nd-YU-cmirko That looks pretty good for a VR2 result, and very encouraging given some of the other RDNA2 results we've seen. 

 

Did you also happen to run the CPU, GPU and VR1 tests? The folks we saw run those previously had some significant issues. It will be very interesting to find out if you're seeing similar performance across the board, or if Il-2 may have finally gotten whatever it was fixed. 

 

Thank you! 

[chiliwili69]

21 hours ago, 102nd-YU-cmirko said:

VR2 original test
2021-12-26 11:19:35 - Il-2 - 90hz
Frames: 3964 - Time: 60000ms - Avg: 66.067 - Min: 52 - Max: 91

2021-12-26 11:26:35 - Il-2 - 60Hz 
Frames: 3591 - Time: 60000ms - Avg: 59.850 - Min: 56 - Max: 61
for posterity all testing was done with original il2 openvr_api.dll ^^

 

Thank you for your tests. It is very interesting to see how the 6900XT cards performs.

 

As sadi above, it is also very important to run the CPUTest, (GPU test is you have a 4K monitor) and also the VRTest1.

 

Both, (CPU test and VRtest1) are useful since they just measure the CPU performance (assuming it is not bottlenecked by GPU, since they are not graphically demanding, not many pixels). In the past tests with AMD cards we saw that the performance in CPUtest and VRtests were somehow limited by the AMD cards. it seems they have an issue with IL-2 which has been reported to Dev Team.

 

In your VRtest2 with original dll, you have a large difference in your results, specially in the max fps (61 vs 91). That´s strange. During the test you should not move the head since depending on were you look at your fps varies (sky vs ground).

 

So, it seems that the issue is still there according to the VRtest2 with original test (compared to other tests with 3080-3090 and Ryzen CPUs).

But using the 60Hz and your settings you get a good average, very close to the 60Hz.

 

[102nd-YU-cmirko]

I have a WQHD monitor,  so can't do a native 4k test, updated the first post with full set of results with this current driver setup

for posterity: gpu is on 21.10.2, chipset drivers 3.10.22.706, and hydra recommended PBO curve per core , also, HMD rests on the table when running the test mission and I center it just before pressing start

Edited December 27, 2021 by 102nd-YU-cmirko

[chiliwili69]

On 12/26/2021 at 12:35 PM, 102nd-YU-cmirko said:

CPU TEST 

2021-12-27 11:56:38 - Il-2 
Frames: 5937 - Time: 60000ms - Avg: 98.950 - Min: 87 - Max: 137

 

VR1 original test 60Hz

2021-12-27 12:08:36 - Il-2 - 60Hz
Frames: 3590 - Time: 60000ms - Avg: 59.833 - Min: 56 - Max: 61

 

VR1 original test 90Hz 

2021-12-27 12:18:13 - Il-2 - 90Hz
Frames: 3821 - Time: 60000ms - Avg: 63.683 - Min: 50 - Max: 91

 

Thank you for runing those extra test.

 

In the CPU test, a 5800X should be around 120fps. But I think the 6900XT is limiting the CPU test and it should not be like that (it is just a 1080p test with low clouds and no MSAA). So, it seems that the IL-2 problem with the AMD GPUs is still there.

 

The VRtest1 at 90Hz also confirm more the issue. The VRtest1 is just 9.5million pixels (something really easy to handle for a 6900XT in any other game), but you 63fps is below the 85fps that a 5800X should deliver in the VRtest1.

 

The AMD issue has been reported here.

[Daisy_Blossom]

Update with the 3080 installed...

Motherboard: Gigabyte Z370 Aorus gaming 7
CPU:               8700K (delidded) 
CPU Freq:       5.1 Ghz
L3 cache:       12 MB
Cores:             6
Threads:         12
RAM type:      DDR4
RAM size:       32Gb (4x8GB)
NB Freq:         4000 MHz
RAM Freq:      3200 MHz
RAM timings:  14-14-14-34-560
Ram type: G.Skill F4-3200C14D-16GTZR (x2)
GPU:      EVGA 3080 FTW3 Ultra Gaming (10G-P5-3897-KL)
HMD: Valve Index
Cooling: Full custom water loop CPU + Mobo + GPU

 

GPU drivers: 497.29

IL2 Version 4.702, Benchmark V6

 

CPU Test: 

Frames: 5912 - Time: 60000ms - Avg: 98.533 - Min: 87 - Max: 132

 

VR Test 1: (90hz @ 106 super-sampling)

Frames: 4695 - Time: 60000ms - Avg: 78.250 - Min: 64 - Max: 91

 

GPU Passmark (3d graphics mark): 21638

[chiliwili69]

1 hour ago, Daisy_Blossom said:

VR Test 1: (90hz @ 106 super-sampling)

Frames: 4695 - Time: 60000ms - Avg: 78.250 - Min: 64 - Max: 91

 

In those both (1080Ti and 3080) VRTest1 you are limited by the CPU, not by the GPU.

 

But if you use the 80Hz mode of the Index, you would be fine with the 8700K in most of scenarios. And now with the 3080 you can apply as much %SS as you want and also go to Extreme clouds and MSAAx2 or perhaps MSAAx4.

[Daisy_Blossom]

1 hour ago, chiliwili69 said:

 

In those both (1080Ti and 3080) VRTest1 you are limited by the CPU, not by the GPU.

 

But if you use the 80Hz mode of the Index, you would be fine with the 8700K in most of scenarios. And now with the 3080 you can apply as much %SS as you want and also go to Extreme clouds and MSAAx2 or perhaps MSAAx4.

Thanks for the tips! What graphics preset would you recommend? I was using balanced with my 1080ti but fpsvr seems to show that high is now acceptable. I wish it was more clear what those presets actually changed...

[Voyager]

13 hours ago, chiliwili69 said:

 

Thank you for runing those extra test.

 

In the CPU test, a 5800X should be around 120fps. But I think the 6900XT is limiting the CPU test and it should not be like that (it is just a 1080p test with low clouds and no MSAA). So, it seems that the IL-2 problem with the AMD GPUs is still there.

 

The VRtest1 at 90Hz also confirm more the issue. The VRtest1 is just 9.5million pixels (something really easy to handle for a 6900XT in any other game), but you 63fps is below the 85fps that a 5800X should deliver in the VRtest1.

 

The AMD issue has been reported here.

 

Correct me if I'm wrong, but isn't the VR2 test fairly in line with the 3090/3080 Ti results, however? 

 

I still have to go dig out and reinstall my Reverb G1 and run tests, but if that is the case, that would imply the AMD issue is more of a frame rate cap than directly bad performance at high settings, yes? 

[chiliwili69]

13 hours ago, Daisy_Blossom said:

What graphics preset would you recommend? I was using balanced with my 1080ti but fpsvr seems to show that high is now acceptable. I wish it was more clear what those presets actually changed...

 

From previous posts from the past, my understanding of the presets determine the following parameters (correct me if I am wrong):

 

1. Number of polygons to define the mesh of every object. For example, a wheel in low could have 16 edges, in balanced 24 edges, and so on. I just invented the numbers.

2. Number of particles in special effects like smoke, fire, etc

3. Size of the bubble with detailed view

 

So, the higher the presets the more load for the CPU to compute more polygons, particles and objects. This might also affect partially the GPU load since the render is more complex.

 

In your case, I would first use the 80Hz mode of the Index and would try with the High Preset (but avoid complex mirror). Then see what are your frametimes for CPU and GPU for your usual plane (planes) and scenarios. Some maps like Kuban or Rhineland are more demanding.

[chiliwili69]

7 hours ago, Voyager said:

Correct me if I'm wrong, but isn't the VR2 test fairly in line with the 3090/3080 Ti results, however?

 

There are other tests using 3090/3080 shown here:

 

Rows 7&8 in green are using Ryzens (5800X and 5600X), like the tests of the 6900XT owner, and they are around 84 fps.

 

There are another test in orange made with a 12700K but using a low speed RAM profile (61.95fps) and high speed RAM profile (74.28). Here we see that the 12700K is not performing as well as the 12900K.

 

Then we have also another test in purple with a 9700K. Both tests VRTest1 and VRTest2 are limited by CPU. A 3090 can easily reach above 80 if not CPU limited.

 

[dburne]

17 hours ago, chiliwili69 said:

 

In those both (1080Ti and 3080) VRTest1 you are limited by the CPU, not by the GPU.

 

But if you use the 80Hz mode of the Index, you would be fine with the 8700K in most of scenarios. And now with the 3080 you can apply as much %SS as you want and also go to Extreme clouds and MSAAx2 or perhaps MSAAx4.

 

I would say maybe not so much on that one, even with my 9900k and 3090 FTW3 GPU I would not say I could throw as much SS as I want and run Extreme clouds with MSAA as well. Not and get a fluid experience.

[chiliwili69]

15 minutes ago, dburne said:

I would say maybe not so much on that one, even with my 9900k and 3090 FTW3 GPU I would not say I could throw as much SS as I want and run Extreme clouds with MSAA as well. Not and get a fluid experience.

 

I was telling that for an Index at 80Hz mode. You get an extra margin with that.

 

Remember also that the Index move less pixels than the G2 or VP2, so you can increase %SS to 130-150%.

 

I myself with the Index at 80Hz use 140%SS with UltraPreset and Extreme clouds and MSAAx2 and get 80fps during 99% of the total mission time.

 

But with the 8700K I will use High-Preset not Ultra. And I will avoid mirrors and canopy reflections.

[TheRealBullit]

On 12/25/2021 at 4:20 PM, TheRealBullit said:

It looks like the RAM speed has a big impact on performance. Here is the result with a RAM speed of 3800 MHz:

 

CPU Test:

Frames: 7163 - Time: 60000ms - Avg: 119.383 - Min: 108 - Max: 166

 

VR Test2:

Frames: 4457 - Time: 60000ms - Avg: 74.283 - Min: 57 - Max: 91

 

I have overclocked the CPU now.

The e-cores are deactivated, the performance cores run on 2x54, 3x53, 6x52, 8x51.  

The memory controller runs at 1900 MHz, the uncore frequency is 4200 MHz.

 

CPU Test:

Frames: 7716 - Time: 60000ms - Avg: 128.600 - Min: 115 - Max: 173

 

VR2 Test:

Frames: 5089 - Time: 60000ms - Avg: 84.817 - Min: 64 - Max: 91

 

[paul_leonard]

Results for my Alienware R10 5950x RTX3090.  I have run the benchmark using game openvr and FSR openvr.  I'm not savvy enough to understand why, but the FSR claim has been much improved performance but I wasn't seeing it in game, nor is there an appreciable improvment under the benchmark conditions.  Ran a couple several times as you will see the FSR degrades performance at 1920x1080 non-VR.  No overclocking.  Going back to game ready openvr.

 

The VR benchmarks are not especially impressive, but in-game on Ultra and settings like DBCOOPER01 recommends, I am a rock solid 90fps.

 

FSR 2.x using the following settings:

useNIS - false

renderscale - 0.77

sharpness - 0.99

radius - 0.7

applyMIPBias - true

debugMode - false

hotkeys enabled - none

 

iL-2 version 4.702b

Reverb G2

 

Motherboard: Alienware 0TYR0X-A00
CPU:               AMD Ryzen 9 5950X
CPU Freq:      3.6 Ghz
L3 cache:       2x32 MB
Cores:             16
Threads:         32
RAM type:       DDR4
RAM size:       64 GB
NB Freq:          3460 MHz
RAM Freq:       3460 MHz 
RAM Latency: 22-23-23-55
GPU:                 3090 24Gb

 

2021-12-31 17:25:38 - Il-2 - 2K
Frames: 6545 - Time: 60000ms - Avg: 109.083 - Min: 90 - Max: 155

2022-01-01 14:50:37 - Il-2 - 2K FSR
Frames: 6081 - Time: 60000ms - Avg: 101.350 - Min: 91 - Max: 141

2021-12-31 17:33:38 - Il-2 - 4K
Frames: 8251 - Time: 60000ms - Avg: 137.517 - Min: 111 - Max: 172

2021-12-31 13:46:23 - Il-2 4K FSR
Frames: 8299 - Time: 60000ms - Avg: 138.317 - Min: 112 - Max: 170

2021-12-31 16:53:18 - Il-2 - VR1
Frames: 3554 - Time: 60000ms - Avg: 59.233 - Min: 49 - Max: 87

2021-12-31 16:30:49 - Il-2 - VR1 FSR
Frames: 3619 - Time: 60000ms - Avg: 60.317 - Min: 52 - Max: 87

2021-12-31 16:48:57 - Il-2 - VR2
Frames: 3543 - Time: 60000ms - Avg: 59.050 - Min: 49 - Max: 83

2021-12-31 16:36:51 - Il-2 - VR2 FSR
Frames: 3687 - Time: 60000ms - Avg: 61.450 - Min: 52 - Max: 86

Per above, Windows 11

Paul

I think NB frequency sb be 1730 Mhz

[RAAF492SQNOz_Steve]

4 hours ago, TheRealBullit said:

I have overclocked the CPU now.

The e-cores are deactivated, the performance cores run on 2x54, 3x53, 6x52, 8x51.  

The memory controller runs at 1900 MHz, the uncore frequency is 4200 MHz.

 

CPU Test:

Frames: 7716 - Time: 60000ms - Avg: 128.600 - Min: 115 - Max: 173

 

VR2 Test:

Frames: 5089 - Time: 60000ms - Avg: 84.817 - Min: 64 - Max: 91

 

Now that is more like it!

 

That is a pretty good result. Wonder how much disabling the e-cores helped?

 

Have you monitored CPU (and GPU) temps when running the CPU overclock? This is just a general comment, do not know if there should be any concerns in that area.

[DBCOOPER011]

On 12/25/2021 at 1:28 PM, chiliwili69 said:

 

Thank you for your detailed tests with G2 and Index. You got a very good performance in VR for both devices. Your minimun value is 87 in both tests with the Index!!

Perhaps I will need to think to raise the bar for the VRTest1 &Test2, for example increasing the graphics settings.

 

Regarding the DDR5 memory, each stick has 2 data channel of 32-bits, so you use 4x32bit=128bit data, with the advantage to access 4 simultaneosly.

With DDR4 memory, each stick has 1 data channel of 64bits, so 2x64bit=128bit data, which is the same total, but only 2 access.

 

With the previous Remagen test, we tested some processors (i9-7900X) with "true" Quad-Channel with DDR4, 4x64bit=256bit total. Although they were not providing an advantage in IL-2 over the Dual-Channel DDR4.

 

Your adaptive CPU settings is also interesting. How do you know that IL-2 is always running in the two cores with the highest speed? ie 5.4GHz

 

 

 

 

For some reason I never was able to get the IL2 CPU test to boost to the target frequencies with Thermal Velocity boost(TVB) under adaptive mode. It would hit the target frequencies with cores synced though no problem. I don't know why this is, but other games/software utilize TVB well. I may be doing something wrong as well.  I verify this with task manager and HWinfo64

[TheRealBullit]

1 hour ago, RAAF492SQNOz_Steve said:

Now that is more like it!

 

That is a pretty good result. Wonder how much disabling the e-cores helped?

 

Have you monitored CPU (and GPU) temps when running the CPU overclock? This is just a general comment, do not know if there should be any concerns in that area.

The system runs more stable without the E-Cores, also I was able to raise the uncore frequency more.

At prime 95 small FFTs the core temps are around 90 C.

While running Il2 they are around 60 C.

Have not monitored the GPU yet.

[chiliwili69]

18 hours ago, TheRealBullit said:

The e-cores are deactivated, the performance cores run on 2x54, 3x53, 6x52, 8x51.  

The memory controller runs at 1900 MHz, the uncore frequency is 4200 MHz.

 

CPU Test:

Frames: 7716 - Time: 60000ms - Avg: 128.600 - Min: 115 - Max: 173

 

VR2 Test:

Frames: 5089 - Time: 60000ms - Avg: 84.817 - Min: 64 - Max: 91

 

Thank you. This gives a valuable info. We know now that the 12700K is as good as the 12900K for IL-2. It is just a question on using OC and also RAM speed.

You have now a very good performance in VR with the G2!

[chiliwili69]

16 hours ago, paul_leonard said:

Motherboard: Alienware 0TYR0X-A00
CPU:               AMD Ryzen 9 5950X
CPU Freq:      3.6 Ghz
L3 cache:       2x32 MB
Cores:             16
Threads:         32
RAM type:       DDR4
RAM size:       64 GB
NB Freq:          3460 MHz
RAM Freq:       3460 MHz 
RAM Latency: 22-23-23-55
GPU:                 3090 24Gb

 

Hey!  Thank you so much for giving us the opportunity to see how those top fully assembled PCs from Dell performs. In your case is an Alienware Aurora with 5950X processor and 3090 card. And for RAM you got 4x16 GB.

 

Thank you also for giving this extra info about how FSR performs.

 

At first look, I would say that you can not see the advantage of the FSR technique basically because two causes:

- your CPU (or perhaps GPU) is bottlenecked by temperature. I tend to think that the CPU turbo doesn´t speed up.

- your RAM latency bottleneck 

 

As you know, those top alienware PCs use the same AMD CPU and the same Nvidia processor than other custom build PCs, so looking at the specs they might be seem as good as other custom build PCs, but the evil is always at the details. 

For example, the RAM is 64Gb, but for IL-2 the size doesn´t matter at all (it is not needed more than 16Gb. Look here). The important thing is to look at the speed and latency.

Here Dell didn´t spend too much in the quality (latencies) of the RAM, just put 64Gb. I don´t know how easy is to replace in this R Series Aurora case your 64Gb RAM by just two sticks (2x8Gb) of faster memory, for example 3600Mhz CL16. But the problem could be also the Mobo, I don´t know what memories are supported by this Dell Mobo since they don´t appear in the Qualified Vendor List of any of the RAM providers.

 

This is the problem of large mass market providers (Dell Alienware, HP Omen, etc). They might pick the right CPU or GPU, but they tend to save money in everything else (RAM performance, Mobo RAM support, internal cooling for VRM and GPU, quality of PSU, CPU cooling, etc) except a nice looking case not optimal for air circulation.

But I also understand that they are easy to buy and grab.

 

Frequently those PCs are limited by temperature, either in CPU or GPU.

You could check everything with MSI afterburner. Just plot in a trend the CPU Temp, CPU freq, RAM freq, GPU temp, GPU freq while you run the benchmark.

Your CPU freq should raise to 4.85GHz or more if the alienware CPU watercooling block is doing the work.

 

There are some videos about upgrading some weaks point of those PCs, like this one: 

 

 

16 hours ago, paul_leonard said:

2021-12-31 16:53:18 - Il-2 - VR1
Frames: 3554 - Time: 60000ms - Avg: 59.233 - Min: 49 - Max: 87

 

2021-12-31 16:48:57 - Il-2 - VR2
Frames: 3543 - Time: 60000ms - Avg: 59.050 - Min: 49 - Max: 83

 

The fact that you are having exactly the same performance in the VR1 and VR1 test is an evidence that the problem is at the CPU or/and RAM level.

 

For replacing RAM (assuming problem is not in the CPU) this could be a good video:

 

Edited January 2, 2022 by chiliwili69

[dburne]

No more than 16 GB of ram may be not needed for IL-2 today but I imagine many are like me and play more games than IL-2. I certainly have some that use more than 16 GB.

I would not build a computer today without at least 32 GB of ram. And who knows how much ram IL-2 will use a couple of years from now...

Edited January 2, 2022 by dburne

[paul_leonard]

You're very welcome!  I did think it would be helpful to be able to post results for a pre-built machine.

 

I purchased the PC with liquid cooling and 32Gb RAM.  I'm also a minimum 32Gb guy.  Other games I have seem to really benefit from at least 32Gb.  Installed the next 32Gb later after a Dell rebate brought the cost to about $60 for 2 sticks (seemed like the thing to do at the time).  I've built/upgraded many PC's over the years but with age and a pandemic came supply (errrr demand) chain disruptions, less patience and age related risk aversion to THE COVID.  Having it delivered to the front door, understanding the monetary trade-offs, made sense at the time (and I suspect does so for many).

 

Seems to me like the first approach will be some gentle overclocking and perhaps more aggressive as we go.  It's been a really long time since I played the overclocking game and it often seemed to end with a trip to the computer store.  I agree, the Alienware case is small and airflow seems like an issue.  I've even toyed with jury-rigging external fans to draw/force more air through the case.  The problem better not be the CPU as I've got nowhere to upgrade to although I will explore the CPU turbo line of thinking some more.  If it is RAM, I must admit I get nervous about compatibility with a Dell proprietary mobo.  

 

I have put the XPOWER blower featured in the second video on my Amazon wishlist though.  Too funny.

[chiliwili69]

On 1/2/2022 at 6:57 PM, paul_leonard said:

I've built/upgraded many PC's over the years but with age and a pandemic came supply (errrr demand) chain disruptions, less patience and age related risk aversion to THE COVID.  Having it delivered to the front door, understanding the monetary trade-offs, made sense at the time (and I suspect does so for many).

 

Good to know, you took that informed decission and you already knew the potential pitfalls of the Alien Aurora PC.

 

Probably you should already know that you can order a full built and running PC but fully configurable. In my country there is a well know online shop where you can order that and I used them several times (for high-end PCs for work) with no problems. I suposse that kind of shops are also everywhere.

 

Let´s see what you find is limiting your performance.  With the RAM (if it is RAM problem) I would just try with 2 stick of 16Gb of 3600 or 3800MHz and CL16. If they don´t work you can return them and try with another. 

[paul_leonard]

I'm going to look at what might be going on with CPU first as it is reporting an abnormally low clock speed.  Dell motherboards are apparently finicky about RAM and 3600/3800MHz RAM in my parts is trés cher (Canada).  I do see now that the RAM timings are uber slow in comparison to others though.  There are discounters locally but prices are not cheap these days.  I priced out local discount supppliers for  WallterScott's rig that averaged 133fps [MB-$400,RTX3090-$2700{out of stock},5950X-$939,RAM(CL14)-$640=$4,679+case/SSD/cooler/1000W/cables(say$1,000)=$5,679] which is $320 more than the Alienware.  $1.00CAD=0.70Euro.  The big difference is the cost of that CL14 RAM.  Actually to find an in stock RTX3090 in Canada will cost $4,200.  So pre-built rig not a financial disaster.  Now back to benchmarking...

[chiliwili69]

Yes, I think it could be better to look first at CPU as well.

Regarding RAM, I think the performance difference between 3600-CL16 and 3800-CL14 is going to be small, but price difference it is not. So perhaps you can try first with less demanding modules (probably most chances to get accepted by the Mobo) like 3600-CL16 and see if that give you a good boost.

 

In Canada you can find 2x16Gb 3600-CL16 from 180$ CAD

https://ca.pcpartpicker.com/products/memory/#U=4&S=3600,3800&Z=32768002&L=140,160&sort=price&page=1

[paul_leonard]

What am I doing wrong in the configuration for the benchmark program?  I just went up to my cottage.  The monitor here is a Dell 4K S2127QS with a 60Hz refresh rate (I actually take up my Alienware desktop in an over the shoulder bag to the cottage).  The 4K test results here are locked at 60fps.  I've tried changing the startup.cfg vsync setting from 0 to 1 and back again.  60fps.  The original result posted for 4K was 137fps.  I'm having a not dissimilar problem at 1080p.  I can't reperform my 109fps result anymore, either at home of at the cottage.  The best I can get is around 90 fps which is suspiciously like the framerate of my G2.  The home monitor is a Dell Q3221QS also 60Hz. What is different is that I ran STeditor and resaved all campaigns because my Flight School missions weren't loading properly.  Any ideas?  I'm spending hours trying to figure out what kind of RAM to upgrade to based on chiliwili69's great advice in the above post [thank you] but can't pull the trigger until I can figure this new problem out.  Thanks Paul

[chiliwili69]

Uhmm,  both monitors have the chip with AMD Freesync technology, and that technology is supported by G-SYNC compatible from NVIDIA 10XX series in certain Freesync monitors.

 

So perhaps it is activated in your monitor S2127QS.

 

According to this page, NVIDIA has not officially included the Dell S2127QS as a "G-SYNC compatible" monitor, but appearently it works.

So, if this is the cause, just go to the NVIDIA control panel and deactivate G-SYNC (or deactivate FreeSync in the monitor)