8KX, Elite and 3090

Well, after a VERY long wait, my ROG Strix 3090 24GB OC is finally here, to replace my old 1080ti. What a monster. It took me a few days to summon up the courage to attempt installing it. Fitting it into my ATX case next to a big radiator for the watercooled CPU was a bit tricky, but it went in eventually without needing a hammer.

The reason I bought the 8KX was to play Elite: Dangerous in VR. Elite was a bit marginal with the 1080 though, hence the 3090. I’d be interested to know what settings other 8KX owners with 30-series GPUs are using for Elite, both in-game and in Pimax Experience. Unfortunately PE doesn’t currently recognise Elite if it was bought directly from Frontier, so I can’t use the cloud profile.

6 Likes

I pushed a profile with 3090 for this game and 8KX.

It has to work and be recognized.

Please restart everything, it has to work.

4 Likes

Thanks, this will be very useful.

Unfortunately PE currently only recognises Elite if it was purchased from Steam. Armin acknowledged this in another thread - it is on his list of things to address.

I continue to experiment with this. I would be interested to know how someone with this card sets their in-game and NVidia Control Panel settings, as well as in PE.

One thing that is bugging me. When I play Elite on my 4K monitor I set the in-game graphics settings to 3840 x 2160 fullscreen. But if I start Elite in VR (from PE, with the settings provided above by @Timo.H1) I find that the in-game settings report that the display has been set to 1280 x 768 windowed. Is this right? If not, what is causing it?

Thanks

the settings in the Windows window should be lower.

it saves performance.

the 3090 is at its performance limit anyway

3 Likes

Thanks for the reply, but there are still many things that I do not understand about the way that Steam VR, Pitool, PE and the in-game settings all interact.

I guess you are saying that 1280 x 768 windowed is the maximum that Elite can be set to in VR.

Is that actually what gets sent to the 8KX? Or is the actual resolution sent to the 8KX determined by other factors?

That setting only sizes the window that mirrors your view on the desktop, and has nothing to do with what you get in the headset.

Resolution settings in the various stages along the VR pipeline do not “interact”, as such - they are just “inherited” from start- to end point:

  • The resulting resolution given by your PiTool settings (Quality, and PP on/off), is the headset resolution piserver tells to SteamVR…

  • SteamVR then applies its own modifying factors (global and per-game) on this, producing the resolution it recommends the game to use (…which the game is perfectly free to ignore, and which e.g. many UnrealEngine4-based games infamously do).

  • The game in turn can multiply the size of the bitmap rendered, as recommended to it by SteamVR, using its own “HMD Quality” setting.

  • The frame pair, information on how large they ended up, attached, is then handed back down the line, from the game, thru SteamVR (…which may add some overlays, if active)…

  • …to Piserver, which pre-distorts them, so that they come out right when being seen through the opposite distortion caused by the HMD lenses; Resampling them for native screen output in the process (possibly with optimised texture filtering depending either dynamically determined on the the size of the incoming frames, or on the Pitool Quality setting at the bottom of the stack, but this possibility has only ever been implied, so how knows). .

Elite also has a “Supersamling” setting, which basically (…for this game - others titles may call things differently) does the same thing as “HMD Quality” - rendering smaller or larger images than requested, but, crucially, resamples to the requested resolution on its own, before handing the finished frame pair over to SteamVR. This means any extra detail rendered will be “baked in” into the picture, so to speak, and can therefore detrimentally not be of benefit when piserver does its pre-distortion pass (it is locked in place as a downsampled part of a certain texel, and can not be spatially moved individually, to its new location on the output bitmap, which loses some precision).

That said: This ED option; “supersampling” (…other than it working when playing on a monitor as well…) does have its place, because most fast sampling algorithms will get to the point where they start to skip source-image pixels, if their incoming source pictures are overly large - often around the region where the input is twice the size (twice width and height - not twice the area) of the output.
If one’s final render resolution (PiQ x SteamVR x HMDQ) surpasses that size, more detail will be preserved, rather than rendered and then just discarded, if pre-sampled-down to the 2x point, before we get to the final resampling step to native screens.

In the past, some have actually professed a preference for using the slight blurring cused by using a little bit of the Elite Dangerous “supersampling” (typically a value less than 1, rendering smaller images, which are then up-scaled" to game output resolution (which may include a countering greater than 1 HMDQ), to somewhat blur out the strong aliasing associated with the game.

I’ve never agreed with this myself, hating blur even more than I do aliasing, but may have to reconsider my stance come ED:Odyssey, some of whose new distant terrain detailing comes very high in both frequency and amplitude, which (at least in the alpha - we’ll se about release) produced a lot of discretely light and dark pixels blinking in- and out in a ludicrous turmoil of aliasing, as one flew over it - here I found that tuning down some of one’s HMDQ, and bringing in a bit of SS in its place, produced some more natural-looking mid tones, and began to eliminate the sense that the ground looked like TV static. :stuck_out_tongue:

6 Likes

@jojon Thank you very much for this detailed explanation.

To be clear: I’m not looking for a “plug and play” experience (although what would be wrong if I did?) but it is hard to start experimenting with settings when their relationship and hierarchy remains a deep mystery.

I bought the 8K X for mainly Elite Dangerous.
If your top priority is plug and play, in all honesty, I’m not sure if any VR headset meets that criteria. People who have been through several headsets find out that each headset has an estimated face shape/eye position and if your face is off from it you need to adjust it a lot (try and error) and if it’s way off, you need to do some modification to the face cushion and such. And the drag is that if you don’t have anyone that you can borrow/try the headset, you need to get one to find out how it fits your face.
But I’d still like to really really encourage you to go through the hassle since the experience is totally mind-blowing. It literally looks like you are flying through NASA footage when it is working as it should. Something only astronauts can experience.

Edit: Where are my manners! o7

3 Likes

o7 to you too, Commander!

I also bought the 8KX for Elite.

As I said, my top priority is NOT plug and play, but even for those who are prepared to go through the process of fine-tuning the system, the complexity is frustrating. While I have better results with Elite than I did at first, I do not really know if what I am experiencing is as good as it can be with my CPU and GPU.

I would not describe it as “totally mind-blowing” yet, with stuttering video and aliasing artifacts getting in the way of immersion.

1 Like

Unfortunately, ED is about the worst scenario for VR. It requires Parallel Projection mode and it suffers badly from aliasing (so you want the highest supersampled res you can get.

Have you run UserBenchmark to verify that there are no bottlenecks in your system? Are you running ProcessLasso, to optimize your CPU thread usage? They both helped me optimize my overclocked system.

I have an 2080, so I’m running my 8KX at Small FOV, so that I can get crisp visuals. I pretend that my space helmet is obscuring the edges of my vision. It’s far from ideal, but I get a playable framerate.

3 Likes

Thanks Neal, I’ll try both of those.

ProcessLasso can be confusing to setup. The main settings you should care about are Main/ProBalance Enable and Options/CPU/CPU Priorities: Set games like MS Flight Sim and Elite Dangerous to High. The online doc has a lot of info, so be sure to read it if you care about the details.

Turns out I already had UserBenchmark. It reports that my CPU and GPU are both well into the green zone, while my M.2 SSD and RAM are below expectations due to not having dual-channel XMP enabled in BIOS. Will investigate.

Have downloaded and installed ProcessLasso. Will start playing with it. Thanks for the tip.

1 Like

UserBenchmark determined my RAM was not installed in the correct pair of slots. I got about a 20% framerate boost in most games, once I moved a RAM stick.

It’s a great tool for finding issues and the suggestions are helpful without being too technical. I recommend it a lot. Even if your system has no configuration issues, it’s nice to have that confirmed.

2 Likes

Thanks for this Neal, I hadn’t tried either before. System is running really well according to UB.

2 Likes

As an illustration of how ProcessLasso works, I find that if you run UserBenchmark with ProcessLasso enabled, you get a lower score. But presumably that is because PL will detect that UB is hogging the CPU and will throttle it back.

The reason to run ProcessLasso is to reduce stuttering, by preventing other apps/processes from preempting your game.

You need to set the app/game exe priorities under the various ProcessLasso > Options settings. When I first tried benchmarking w/ and w/o ProcessLasso, I initially saw slightly lower UB performance, but once I configured it properly, I did see a slight performance increase. By configured properly, I mean that I used the same ProcessLasso settings for UB as I used for MS Flight Sim and Elite D.

Oddly, the biggest performance increase was with my NVMe drive. I assume that it’s because the SSD drivers are also prioritized when called from an app running at High Priority.

w/o ProcessLasso: Asus TUF Z390-PLUS GAMING (WI-FI) Performance Results - UserBenchmark
w/ ProcessLasso: Asus TUF Z390-PLUS GAMING (WI-FI) Performance Results - UserBenchmark

Emboldened by this thread, and UB results causing me FOMO, I investigated my BIOS settings (ROG Strix Z370-E motherboard) and found that it had a fairly simple overclocking wizard that recommended a 39% boost to my CPU (i7 8700K). Results are encouraging:

https://www.userbenchmark.com/UserRun/43096994

Don’t really understand why the M.2 SSD and RAM results are comparatively low, given that I have enabled XMP in BIOS, and all the RAM slots are populated.

At the moment my test for CPU bottlenecks is Euro Truck Simulator 2 on a 4K monitor, as I was previously getting stuttering while the 3090 was barely working hard enough to start the fans. This OC setting certainly improved things.

Then I went on to Elite in pancake mode in 4K at 144Hz - very smooth, with a solid 144 FPS most of the time, not dropping below 120 FPS in stations.

Will try Elite in VR next.

Then I will look at overclocking the 3090. although I don’t think it is going to be the bottleneck any time soon.

1 Like