For people who will upgrade their PC

[chiliwili69]

If you plan to upgrade your whole PC, or just the CPU or GPU, please run the IL-2 benchmark before and after your upgrade. So we can quantify how good is the new hardware.

This will be specially helpful for new 9series GPUs and the new 9th gen CPUs.

This is a link where you can follow the instructions to run the benchmark, in VR, or in Monitor or in both:

https://forum.il2sturmovik.com/topic/29322-measuring-rig-performance-common-baseline-for-il-2-v3005/

 

Ideally the Dev team could create a common benchmark within the game (or a common flight record which run with new versions), so it will be easier to measure/compare/tweak.

[SharpeXB]

I’ve tried doing something similar with certain missions. One thing I noticed though was that live gameplay is more taxing on your system than running a replay track. 

So just running a replay unfortunately doesn’t give an accurate performance benchmark. 

[chiliwili69]

4 minutes ago, SharpeXB said:

One thing I noticed though was that live gameplay is more taxing on your system than running a replay track

 

Interesting.

How did you measure the difference (live vs replay)? How large is it?

 

Even if there is a difference the purpose of a replay benchmark is also quite valid since the hardware/settings are measured against the same test (which is a replay).

[dburne]

47 minutes ago, SharpeXB said:

I’ve tried doing something similar with certain missions. One thing I noticed though was that live gameplay is more taxing on your system than running a replay track. 

So just running a replay unfortunately doesn’t give an accurate performance benchmark. 

 

You are absolutely correct. In some instances, like the VR benchmark , the replay track would not even boost my GPU, whereas when actually playing the game - which to me is what counts - my GPU was running full boost, even with a ton of activity going on around me. And temp was not an issue either in the benchmark, it was very low. The benchmark does not account for things like what  GPU speed is running at the time of running the trk file.

 

What matters to me is how it performs while actually playing the game.  This is how I will gauge how my investment in the 2080 Ti FTW3 Ultra will pan out soon, along with hopefully a complete new build for rest of the system in couple of months.

 

 

Edited September 18, 2018 by dburne

[SharpeXB]

4 hours ago, chiliwili69 said:

 

Interesting.

How did you measure the difference (live vs replay)? How large is it?

 

I would run through a quick mission live, with Fraps measuring. And then run a track of the same gameplay with Fraps also and then just compare the results. I think it might have been about a 10 FPS difference. I can’t recall exactly. But it’s an easy thing to do. Play once live and then do a replay, recording both. 

Note, there’s an FPS drop when you start a mission and look around, almost like the game needs to load everything, so look around first before starting Fraps. 

Later on I made my tests very simple because I realize what’s CPU or GPU limited. I’ll do a “GPU test” using a free flight with no other aircraft. And a “CPU test” on the runway with lots of other big aircraft. Just stationary. That way it’s easy to run nearly identical live missions if they’re very simple. Not like trying to use a dogfight. 

[chiliwili69]

1 hour ago, SharpeXB said:

I would run through a quick mission live, with Fraps measuring. And then run a track of the same gameplay with Fraps also and then just compare the results.

But How do you run the quick mission? With autopilot? How do you know that those quick missions are identical (live vs replay)?

 

If you are recording during live mission (to record the track) then you are demanding more to the CPU because you are recording.

 

You should know that even if you run a live mission with exactly the same settings and planes, the planes appears in different directions and locations everytime you run the quick mission.

[SharpeXB]

2 hours ago, chiliwili69 said:

But How do you run the quick mission? With autopilot? How do you know that those quick missions are identical (live vs replay)?

 

If you are recording during live mission (to record the track) then you are demanding more to the CPU because you are recording.

 

You should know that even if you run a live mission with exactly the same settings and planes, the planes appears in different directions and locations everytime you run the quick mission.

The in-game recorder has no performance penalty. Just start it in a static situation with a constant FPS. I don’t see a drop from switching it on. 

But yes, I was running the mission, recorder and Fraps all at once. Then replaying the same mission and using Fraps on that. 

But running nearly identical live missions isn’t hard if they’re very simple. Like a free flight or sitting on the runway, which I’ve actually found to be the most CPU demanding situation in this and DCS. Using complex dogfights as benchmarks is confusing because, now that I more or less understand that the bottleneck can be either GPU or CPU I wanted to test each of those individually to better understand which setting was causing a problem or to discover some that I can set higher with no trouble. 

It would make sense that live gameplay is more demanding because I’m guessing that it involves all the flight model physics whereas the replay is just that, it’s not running anything other than a playback. 

Edited September 18, 2018 by SharpeXB

[chiliwili69]

8 hours ago, SharpeXB said:

But yes, I was running the mission, recorder and Fraps all at once. Then replaying the same mission and using Fraps on that.

Yeah, in that case the mission is identical then. But you have to be sure to don´t move the head around to have exactly the same views. You can fix the headseat to a fake head and use the monitor view to control the plane while you are recording. Also you need to start Fraps at the same time you start to record.

Can you show us the fps for live vs track along the time? (you can record that with Fraps).

8 hours ago, SharpeXB said:

Using complex dogfights as benchmarks is confusing because, now that I more or less understand that the bottleneck can be either GPU or CPU I wanted to test each of those individually to better understand which setting was causing a problem or to discover some that I can set higher with no trouble. 

Yes, ideally we could have a track for runway, another for dogfight (smoke and fire), another for mountains, etc but it will be more complex to analyze how every setting affects every track.

But you say that the bottleneck can be the GPU and you have a 1080Ti. Could you tell me what mission/plane/settings is clearly bottlenecking your GPU? I have no found that in any fighter in any circumstance with SteamVR SS=150%

8 hours ago, SharpeXB said:

It would make sense that live gameplay is more demanding because I’m guessing that it involves all the flight model physics whereas the replay is just that, it’s not running anything other than a playback

Yes, this could be a very plausible reason for the difference. But only the Dev team could tell us the true reason for that (we don´t know if the track records actions of pilot or position of plane) 

[SharpeXB]

4 hours ago, chiliwili69 said:

But you have to be sure to don´t move the head around to have exactly the same views. You can fix the headseat to a fake head and use the monitor view to control the plane while you are recording.

When using TrackIR the in-game recorder will record your head movements exactly as they were in the live gameplay. I don’t know if that’s the same with VR

5 hours ago, chiliwili69 said:

But you say that the bottleneck can be the GPU and you have a 1080Ti. Could you tell me what mission/plane/settings is clearly bottlenecking your GPU? 

These tests I had done before I got the 1080Ti. Yes I don’t think there is a situation currently where I could bottleneck that card in BoS

i had been testing GTX 980s and Titan X cards in 4K.

[CanadaOne]

Il2-46 had "The Black Death" track as the standard hardware test. The community set that up I think.

 

But it would be great if the devs could set up a standard track to test hardware.