SU11 is here - First test with performance comparison with SU10

Hello! Before the servers went offline here, I managed to run a couple of test runs.

I compared two DX12/HAGS modes between SU10 and SU11.

As usual in my tests, all VR, 4090, 13900kf, 32 GB DDR5 @ 6400 etc.
Ingame render scale 100, OpenXR render scale 200. DLSS Performance.
All VR options are on Ultra, but Occlusion, Light Shafts and Reflections (on Low).
All tests are an aggregation of 5 different runs (median three).

Let’s begin with DLAA, the new DLSS mode: for me it’s not usable, FPS go down to 25 (max) and 10 (low). So I scrapped that entirely, for now.

Going back to the test, the first interesting thing I can flag is that, in all the 4 scenarios, the max FPS remains roughly the same, while the low percentile ones have greatly improved in SU11 - this should translate in more smoothness!

I really like the 58 FPS locked scenario in SU11, with roughly 97% of frames with a variation under 4ms!

Looking at the analysis, the 58 FPS locked one shows great improvements: P0.2 went from 31.1 FPS to 44.1 FPS, and P0.1 from 30.5 to 42.6! Also, no more spikes above the 30 ms, at all.

The Free FPS shows a similar pattern, albeit with minor improvement. Again, no spikes above 25 ms, while with SU10 we had plenty of those.

Please note these tests are done with a replay, so they’re “on a track”, let’s say. This evening I’ll try flying for real (moving head, and so on).

Furthermore, they are a very rough first impression, I will try to complete the thread with more info later today, or during the weekend.

Final notes:
58 = 58 FPS blocked (H: with HAGS)
U = free FPS (H: with HAGS)
NT = No OpenXR Toolkit (T = Toolkit activated, only Fov rendering)
NRPO, NRPG = no motion retroprojection of any kind
NH = Holographic Shell off (to solve the W112 22H2 issue)




5 Likes

I wish I knew how to interpret this and conduct similar tests with my DX11 3080 5800x setup.

1 Like

Well about how to conduct… first get CapFrameX (it’s free). Then fly and save a replay (1,5/2 mins will suffice), for example with SkyDolly. Then repeat this replay using always the same meteo/time of the day, and adapt the various options as you want to test the various things.

If you have time, repeat 5 times each test and aggregate, keeping the medium three. Otherwise, just go with 1-time testing.

About interpreting the results, it’s pretty self-explanatory in CapFrameX, there are plenty of hints.

3 Likes

If I don’t want to benchmark per se, but want to do optimization of a given live weather environment, etc., does running CapFrameX while just flying normally give me anything useful to that end?

Well if you “record” the flight, you can analyze it after landing, to see how smooth it was, when problems occured, and so on.

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.