TL;DR - I finally have what I think is a solid baseline of settings for this headset and computer hardware combo that works well as long as you're not interacting with ground settlements, so if you're using the same setup (or similar enough to it) see the screenshots on this post for my settings to use as a starting point for further tweaking your setup).
o7 again commanders, I'm back to necro this oldish thread one more time with some updates. After over a week of tweaking settings, testing, data logging, repeatedly, ad nauseam, I now have what I think is a solid baseline setup for anyone running my exact hardware combination or relatively close to it, with a VR headset of similar resolution, FOV and refresh rate. My hope is anyone with this setup that comes across this thread in the future will see it and can simply plug these settings in and go about playing without having to spend days of effort like I did. For a quick recap, here's my hardware combo as far as what's relevant to the performance of ED:
CPU: AMD 7800X3D (with EXPO and PBO enabled)
RAM: 32GB GSkill TridentZ DDR5-6000 (timings have been optimized with BuildZoid's "easy hynix timings" as well)
GPU: EVGA Nvidia 3090 Ti FTW3
Storage: 4TB PCIE gen 4 M.2 SSD
I. Recap
In previous posts on this thread we've established that it's IMPOSSIBLE (as of April 2025 at least) to get this game running without massive frame timing issues when near ground settlements, when using the Pimax 5K Super at 100% native resolution, even in "potato" FOV mode and even at 90Hz refresh. Even people with 4090's and other headsets with similar resolutions can't achieve it (I doubt 5090 can either). The settlements are just insanely difficult for the engine to render and anytime they're in view, the frame times just go to complete hell. So if you've stumbled upon this thread in the future and have this headset and hardware combo, abandon any hope of getting a smooth running experience at 100% resolution and wide FOV that works
everywhere in the game.
II. New direction
After our previous conversations here and all the original testing I did, I just accepted that my goals weren't achievable with this combo, so I thought about it for a bit and decided what's most important to me for my play style and ultimately settled on the following:
- 100% resolution - for me personally, I absolutely HATE resolution scaling that doesn't line up perfectly with the native resolution of the headset. My vision is still really good and it just looks "off" to me and annoys the hell out of me, so I decided to prioritize native pixel resolution if at all possible, and with as much FOV as I could get away with. I'm not using supersampling either, as this headset has enough resolution that it's really not that beneficial. The single biggest reason I got the Pimax 5K Super was for the resolution so scaling that back just as no-go for me if I could help it.
- I wanted to try to get it working at higher refresh rates than 90 Hz. I used to be ok with 90 but once I experienced 120, 144, 160 and 180 in the Pimax 5K Super (with other games, not ED) it's really hard to go back to 90 if I can help it (spoiler alert: don't even bother trying at 120+ at native resolution with ED, even in "potato" FOV).
- I wanted at least a 90 degree FOV. I got the Pimax first and foremost for the resolution and clarity, and the wide FOV thing was a very close second. If I could get at least the standard 90 degrees of most common headsets I can accept it, especially since as you get further out to the periphery of the Pimax the visual acuity begins to warp drastically. That wide FOV is really just there to give your peripheral vision enough information to pick up on (it's amazing for racing simulators where you're driving side by side with other cars and can sense their movements with your peripheral vision), but it's not something you'll make a conscious decision to look over at directly with your eyes as it's too blurry.
- Play style - I'm an explorer type of commander at heart, so in the end the settlements aren't really something I'll interact with much. I spend most of my time in space, and often out deep in the black doing exploration all over the galaxy, which is nowhere near as hard on the GPU as planetary landings and settlements. Plus, I absolutely SUCK at playing FPS combat games with a game controller anyway (I use an Xbox 360 controller for all my flying) so not being able to approach settlements on foot or in an SRV isn't really a deal breaker for me. If I want to interact with the on foot content, I'll come use my main computer and play it via keyboard/mouse outside of VR.
That last part is the compromise that allowed me to get a settings configuration that works well in all other situations for my VR activities.
III. Testing Strategy
Just to reiterate before going forward - the statements I'm making regarding performance on this post only apply to those with exactly my hardware setup and headset, or very close to it, so for other folks with sufficiently different configurations, don't take the following statements as truth for your situation and you'll need to do your own testing.
For this kind of tweaking exercise you need consistently repeatable test scenarios to validate the impact of your settings changes and that process took a bit of time to nail down. Regardless of how low I dropped the settings, as long as the headset was at 100% resolution (even in "potato" FOV mode) there was simply no eliminating the intermittent occurrence of some orange frametimes (times > 11.1 ms needed for 90Hz) in the fpsVR frametime graph. But it is possible to minimize their occurrence to once or twice every few seconds or less, and at that level they're not noticeable in headset so then it was just a matter of finding a "sweet spot" of settings that minimize them consistently to an acceptable degree in all game scenarios I care about. Apart from that, I wanted the settings such that the frametimes keep the GPU at 90% utilization or less, as you want some extra capacity available to absorb those sudden spikes of activity that require more processing power.
So with all that in mind I tested with the following scenarios (in order of increasing difficulty):
- Scenario 1 - Simply flying out and about in space through the Sol system
- Scenario 2 - Stopping in POV's and making sure the frametime performance was consistent during fighter combat and other ship to ship interactions in space. Being in the midst of combat is the LAST place you want graphical hitching and stutters in VR
- Scenario 3 - Driving around on planets in the SRV (further details on my testing of this in the "planetary testing" section)
- Scenario 4 - Space stations - flying into and out of, and while sitting parked inside.
NOTE: All my mentions of frame times only applies to the
GPU frametimes. The
CPU frametimes are consistently in the low 3ms range (except for when disk I/O interrupts it such as when jumping between systems or spawning on the surface of a planet) and it's not even using more than around 20% of my CPU most of the time and the CPU hovers around 65C in temperature. The 7800X3D is a beast of a chip and is frankly bored by this game, lol. Even running my racing games with 60 AI cars on track doesn't get anywhere close to maxing it out. As such I don't see upgrading to the 9800X3D improving my ED experience in any way, at least not with a 3090 ti. It's possible it might have some benefit for the 4090/5090 owners, but I have no way to test that unfortunately (I refuse to pay scalper prices on ebay for a video card, especially a USED video card like a 4090).
IIIa. Planetary Testing
The landable planets in the game can vary wildly in how graphically taxing they are, so proper testing is about finding a worst case scenario and testing your settings changes against it. For a long term and graphically intensive experience, I went to Europa and landed on one side of one of the deep cut canyons you see all over that moon, then got out in the SRV and captured a data log file with fpsVR while driving straight down oneside and up the other side, then turning around and driving back to the starting point the same way. This gives a long term test and a wide variety of scenarios as the terrain changes from super shiny icy and flat terrain to geometrically complex rocky terrain sections. And because I'm a rather crummy SRV driver with no regard for my virtual character's safety, there were many low grav bounces off rocks and wipeouts along the way which further taxed the GPU at times
.
I also tested by going back to "Kawle Genetics Complex" settlement on planet 2045 PC2 in the Alpha Centauri, and then driving the SRV away from the settlement. It was a bit choppy at first obviously but as I moved away from the settlement it quickly stabilized and eventually reached similar levels of frametimes as my testing on Europa.
IIIb. Station Testing
Stations are among the more taxing scenarios in the game, particularly the interior of coriolis stations. I did all my station testing at Mars High in Sol as that was the most graphically intensive one I know of offhand. At the settings I've provided you'll get some dropped frames and hitching as you get close to the mail slot and more as you enter the station but it's bearable for the short amount of time it takes you to get to your landing pad. Once you're landed and bring your ship down into the docking bay the issues mostly subside once the door shuts and you can no longer see into the station interior, and it becomes even less of a problem while you're in the station menu. Bringing the ship back out to the interior and just sitting on the landing pad facing toward the mail slot is still somewhat taxing, though it varies wildly based on how much of the station interior is in your view (it's worse on the pads toward the rear than the ones upfront). Performance is more than sufficient for safely taking off and leaving with other ships around so I consider it a worthwhile tradeoff for the higher visual fidelity everywhere else in the game.
IV. Results
Here's what I was able to get working well with consistency:
- 100% resolution on the Pimax 5K Super is doable, with some restraint on certain settings
- >90Hz refresh rate isn't gonna happen at 100% resolution. The next highest setting is 120Hz and that's 33% more frames to crank out, which the 3090 Ti simply can't keep up with at 100% resolution.
- The "small" FOV setting (which on the Pimax 5K Super is in the realm of 90 degrees) is as high as you can go if you'll be spending a lot of time on planets. If you stick strictly to space then you can push it to the wider FOV's at 100% resolution and 90Hz refresh rate. I didn't test it, but higher refresh rates MAY be possible when only in space.
Space only settings
The good news is that I was able to max out most of the settings within ED and have an acceptable playable experience in the majority of gameplay time. And
if most of your gameplay time is scenarios 1-2, then you can just max them all out get to playing.
Planetary Settings
If you're going to be spending a decent amount of time on planets then you'll have to dial back a few settings. Not surprisingly. the settings that make the biggest impact (apart from the obvious like supersampling or upscaling) are
Ambient Occlusion, Directional Shadow Quality, and
Spot Shadow Quality. These are generally the most computationally expensive graphical operations in the game so it makes sense. Surprisingly all the terrain related settings didn't seem to have much impact on performance so I just maxed them all out (I started at the lowest options and went up to the highest and didn't see much effect). A few are contingent on having sufficient VRAM but that's not a concern with the 24GB on the 3090 ti (highest fpsVR ever recorded was ~11.2GB peak VRAM usage).
Using the settings as shown in the screenshots, planets are pretty much fine for extended periods of gameplay. The incidence of orange GPU frametimes definitely increases and is less consistent; you'll see it more as you're driving when the game suddenly spawns more rocks or you get to more geometrically complex terrain like the rocky sections, but you should never see red frame times (a frametime that exceeds more than double the refresh rate of the headset) so the risk of VR sickness is low. Granted I have a pretty strong stomach when it comes to VR sickness so the intermittent late frames didn't really affect me at all, but if you're more sensitive to it then you may end up having to dial back the settings a bit further so your mileage may vary there, but what I have here should get you 95% of the way there.
Additional comments on some of my settings:
- Model Draw Distance - I didn't experiment at all with this. I like being able to see things far out so I didn't want to compromise on that if I could help it.
- Blur - I completely disabled this as I personally hate this effect, so I didn't do any testing with it but I don't imagine it would make an appreciable performance difference if enabled.
- Bloom - I tested all options while orbiting Jacksons Lighthouse as the neutron stars are where I usually see the most bloom. Even "ultra" was perfectly performant but I ended up dialing it back to medium simply because the Pimax 5K Super uses LCD screens and excessive bloom tends to wash out the image. Ultra looked amazing on my Vive Pro which has an OLED screen though. Tweak to your personal tastes.
- Upscaling - I didn't even mess with it. As we established on previous comments on this post, FSR 1.0 is hot garbage and not at all worth using.
- Antialiasing - Set it to MLAAX4 and forget it. I didn't find X2 made an appreciable difference in performance. SMAA didn't improve performance much either and FXAA looks like absolute garbage so definitely don't bother with it.
- Texture Filter Quality - Anisotrophy doesn't have any real performance impact so just max it at x16 and have fun.
- Depth of Field - personally I hate this effect too and disable it in all my games. It probably makes even less sense in VR since you're actually actively focusing your eyes while in the headset. If you like it, give it a whirl as I don't think it'll make much performance difference either.
- Terrain Checkerboard Rendering - I left this one on, as it reduces the rendering load a bit and I didn't find it made an appreciable difference in quality by turning it off so I felt leaving it on was worth the small bit of performance bump.
V. Other miscellaneous stuff
- I did some initial testing using Pimax XR instead of SteamVR and as expected it appears to be at least as performant as SteamVR or more. Unfortunately I don't have any data to back that up because fpsVR doesn't work with OpenXR so you're kinda flying blind there. I did just discover yesterday the existence of XRFrameTools and gave it an initial spin and found it looks promising. You can do frametime logging, have it display in real time on the desktop and of course run the logs through visual analysis post session. It's a great (and free) first step for profiling OpenXR and I hope the developer keeps improving it, but not having a live data readout in the headset to look at while you're in VR like you do with fpsVR makes testing much more difficult.
- When using PimaxXR you have the option to use their fixed foveated rendering. I gave it a shot and it definitely improves performance a bit, but you can't tweak the radius of where it applies foveated rendering levels and I found that Pimax sets it JUST too close in toward the center of my vision and so it feels like there's a constant "bubbling" of pixels in my peripheral vision that bugs the hell out of me so I gave up on it.
- I included pics of one of my planetary SRV logging sessions for those who are curious. You'll see it says 95% of frames were on time, which normally wouldn't be the greatest, but the distribution of those frametime dips is intermittent throughout the session so it doesn't feel as dramatic as the data makes it out to be in this case. And there are a few intervals in there where it momentarily went red for a few frames when the game loaded something into the world, which is unavoidable no matter how low your settings are (the Pimax software is just very sensitive to anything that interrupts the CPU like disk or network I/O). Also you'll notice it says the headset is "Pimax Crystal" but trust me, it's a 5K Super. Pimax is kinda infamous for forgetting to update the name of the headset in the driver config files (there's another setting in a different config file that calls it the 8K too).
- If you're not like me and are ok with some resolution downscaling, there's a very good chance you can completely max out the settings and perhaps increase the FOV and/or refresh rate. Might be worth spending time tweaking it if you're someone who prefers higher refresh rate to resolution.
Hopefully all of this helps at least one person have a good experience without hours and days of tweaking If anyone has any questions let me know. And if anyone has any other more difficult test scenarios they know of, let me know so I can test those too!
CMDR Darock
