TL: DR what I have learned.
- My local benchmark table cannot be copied over to the forum in one click :,-(
If you want to add a bunch of results, I suggest creating the table in the forum and adding directly here while saving regularly if you keep it open for more than 10 minutes.
- Flatscreen resolution does not seem to affect performance (when in VR).
- SteamVR resolution and HMD quality in-game have comparatively the same effect on performance.
- Other SteamVR options did not influence performance but can affect comfort (motion smoothing, auto throttle).
- Default VR graphic profiles in ED are not necessarily “optimal” depending on your headset. They are used to get the same baseline if other CMDRs want to add their results.
- There is a bigger decrease in performance between VR medium and high.
- Station interiors are the most demanding of the scenarios tested.
Introduction
I wanted to see if my VR experience is similar to the average.With many different hardware combinations (computer and VR headset) as well as a lack of real VR benchmark, I went on to establish some protocol to get repeatable measurements.
Having a SteamVR native headset already put me on a specific use case but I think comparing with other compositers (hopefully that is the right word) can still be useful.
My configuration (using a mix of system info and HWiNFO)
- Intel i5 12600K @3.7 Ghz
- 32 GB RAM DDR4 3200 16CL
- Geforce RTX 4070 (GV-N4070WF3OC-12GD)
- XPG GAMMIX S70 BLADE NVMe 1TB
- Bigscreen Beyond VR headset, 3560x3560 100% SteamVR resolution 75 Hz refresh rate
Test protocol
- This is what I used and should ideally be followed to ensure consistent results across tests.
- Put your VR headset on something stable and levelled so that all UI elements are visible in the VR view, this will ensure the viewed scene is consistent across tests.
This will also make it much easier to alt-tab and note your results after each scenario. You will NOT wear your headset at any time.
- SteamVR: Click the three horizontal parallel lines on the mini SteamVR UI, enable Display Performance Graph. You should now see your frame time in the mini UI.
For example mine is at 2.0 of 13.3 ms (75 Hz) when there are no game launched and the Steam home is NOT launched.
- Game version started on 4.1.0.2 (Odyssey) and will be indicated if different.
- Use the various training scenarios to help with consistence, those are available at any time from the play menu, under training.
- Under Controls, Ship Controls, Miscellaneous, bind a key for "Reset HMD orientation" and press this once you have set settled your HMD on a flat surface.
This will ensure the view is fairly consistent across tests.
It is normal to have a skewed view on your monitor since ED only show you one eye, likely the right one.
What this reset doesn't address is the up/down tilt.
As a reference, I got to the CONTINUE menu where you see open play and such and make sure I see the word ODYSSEY at the top left and BACK TO MAIN MENU at the bottom left.
- Get the frame time after counting 10 seconds from the start of the scenario (after the loading screen).
- Surface Operation (on-foot tutorial) has two values, flatscreen and VR using the default external camera facing your Commander.
- Advanced SRV/SLF does NOT benchmark SRV or SLF performance, just the initial situation.
- "SRV pratice" scenario test SRV performance.
- Ideally, re-run a few scenarios to see if you get the same results. A complete test loop of all the scenarios took me between 15-20 minutes.
- In-game time when the tests were run has been indicated. I am not sure how much this can impact the performance in the various systems used in tutorials.
- SteamVR specific
- Motion smoothing off
- Throttling behavior fixed 90 FPS, 0 ms additional prediction
Conclusion
- A Geforce RTX 4070 is not enough to run ED with the Bigscreen Beyond at its native resolution.
The biggest impact frame time is the resolution, the performance at HMD quality 0.5 is good in all scenarios I have tested.
I can push it to 0.85 in mostly empty space (exploration scenarios, excluding landing on planets).
HMD quality can be changed on the go, whereas some graphic settings need a change of system.
This makes it a good candidate to adapt performance to the situation.
- The benchmark scenarios, while diverse, lack specific game scenarios (neutron star, repeatable multi-ship flight/combat, etc).
- From my experience, planet surfaces can have very different performance, likely based on the terrain.
- Hopefully people can contribute with their own benchmarks and we can compare results.
- It’s definitely not a perfect protocol, I am happy to take constructive feedback, o7
Last edited:
