Haven't read all the posts, but am I right in assuming that super sampling (SS) is the key to good looking VR? Has anyone tried turning other settings way down, and setting SS as high as possible for smooth play?
Increase pixel density - VIVE - improve text readability
- Thread starter ElectricZ
- Start date
The article <http://alex.vlachos.com/graphics/Alex_Vlachos_Advanced_VR_Rendering_Performance_GDC2016.pdf> is quite interesting and might answers why ss 2.0 works on cards with more than 4GB of RAM. If you take a look on page 47 and 48 a scale of 1.4 requires just below 4GB or ram to enable 4 rendering pipes. A scale of 2.0 then requires 6.5GB. Maybe amount of RAM on the card is key to the different performance we see. I have 6GB of RAM on my 980ti and the shuffling of the last 0.5 GB might be ok in this context??
- - - - - Additional Content Posted / Auto Merge - - - - -
I run with FDs launcher but start steamvr on the side (I thought this was mandatory and have not tried any other way). I.e. I do not start ED from steam. But make sure that you set the scaling from steamvr and not from inside ED graphics option as described in many posts in this forum. This is the key to good gfx.
- - - - - Additional Content Posted / Auto Merge - - - - -
Cheers Aio. So as usual, it's a matter of finding "the sweet spot".
When you say steamvr, what are your impressions with VR through FDs own launcher?
I run with FDs launcher but start steamvr on the side (I thought this was mandatory and have not tried any other way). I.e. I do not start ED from steam. But make sure that you set the scaling from steamvr and not from inside ED graphics option as described in many posts in this forum. This is the key to good gfx.
I'm really pleased to hear people are enjoying what they get; I am too! This really isn't about trying to take anyone's enjoyment out of their game - for me it's mainly about seeing what changes effect quality and whether I'm missing something (and of course as a resource so other people with similar setups can quickly find the best settings)! That's why I want to compare - but for me to do that I really need some concrete numbers to compare things with.
If Aio, Shadragon and J'Zargo could take some screenshots of the Frame Timings in a standard situation - either the lobby, or the icy rings in the first training mission - then we could sensibly compare. The 1070 FE should be similar (or slightly better) performance in VR to the 980Ti so if I'm getting considerably worse performance then I've clearly set something up wrong somewhere and can improve my experience!
Last edited:
I did some tests and I have some missing frames inside the station. I run the "welcome to VR" traning. I disabled reprojection which makes the frame timing more clear. When inside the docking I have quite a lot of missing frames, which is a bit strange to me. When I "go to surface" things gets better. I have some missing frames when the Annaconda flies over me but then inside the station I have quite solid 90 fps (usually no frame drop in the 900 frames) with idel time of ~1ms per frame. But I have no constant 45 fps as far as this limited test goes.
Ok, here we go. Beware a lot of images to follow.
I did some extensive tests and recorded all the steam vr timing graphs.
My setup is as follows:
i7-4770K overclocked to 4.2Ghz on all 4 cores
16GB RAM
MSI 980Ti Gaming with 6GB VRAM, default MSI clocking ( which is already an overclocked version)
I did tests for all in game graphics presets (VRLow, VRHigh, High, Ultra) , where I changed only the SS to 0.65 and always deactivated Antialiasing.
The VR Steam SS settings were set to 1.6.
1.) Docked ship inside the station of the VR training session.
vr_low:
VRHigh:
High:
Ultra:
VRHigh:
High:
Ultra:
2.) Inside a ring system, from the first training mission to destroy the canisters
VRLow:
VR High:
High:
Ultra:
VR High:
High:
Ultra:
3.) Hovering 2 km above a planet side station
VRLow:
VRHigh:
High:
Ultra:
VRHigh:
High:
Ultra:
4.) Planet surface in a SRV
VRHigh:
High:
Ultra:
VRLow and VRHigh are ok here. With High it is again the CPU which seems to have problems or at least cannot correctly sync with the GPU. It fluctuates always between 90fps and 45fps, but the GPU stays well below the 11ms. Ultra is always 45fps due to GPU limitation.
---
What can be seen is that there seems to be some CPU synchronization issues which might account for a lot of fps issues. The 980Ti has enough GPU power to drive the game almost always in High settings. It is the CPU which makes this impossible. The interesting thing is that the CPU utilization is nowhere near 100% more at 40%, so this CPU is spending its time more in waiting than pure processing, which might hint at a synchronization issue between CPU and GPU. This would also explain some observations I had, where for no obvious reasons the CPU time increased without any change in environment ( observed hovering outside a station).
Last edited:
Thank you for that, J'Zargo. Have some rep. 
Finally decided to tempt fate, and try to see if I can improve Elite visually.
Tried various combinations, but got a noticeable visual improvement only with 0.65/2.0... and there a bit of judder in the station menus that I found annoying. Plus, after about an hour of playing with these settings, I'm a little nauseous. Given that I've driven for the surface of planets in my SRV without locking my view to the horizon and never gotten sick, I'm going to set everything back to what it was, VR High/1.0, and wait to see if Frontier can fix the problem at their end.
Finally decided to tempt fate, and try to see if I can improve Elite visually.
Tried various combinations, but got a noticeable visual improvement only with 0.65/2.0... and there a bit of judder in the station menus that I found annoying. Plus, after about an hour of playing with these settings, I'm a little nauseous. Given that I've driven for the surface of planets in my SRV without locking my view to the horizon and never gotten sick, I'm going to set everything back to what it was, VR High/1.0, and wait to see if Frontier can fix the problem at their end.
My experience matched yours, Darkfyre99 - there wasn't much noticeable judder but I felt yucky out of the blue, and I have never had that with ED since getting my Vive. So I'm back to waiting for an 'official fix'.
I too noticed I got a little weird feeling after a bit when upsampling in Steamvr and downsampling in game, setting the game back to 1.0x and I'm feeling fine, still got the Steam setting at 1.4.
I read somewhere (can't remember where) that supersampling was a large hit on gfx ram, using up to and beyond 6GB, which could give some cards trouble. Not sure if this is any help, just thought I'd put it out there.
Quick question ( I haven't ready every post so I may have missed this):
What is the commenting format for the code used in this cfg file? Is it like C++ or what? I'd like to comment my file, because I go for long gaps between messing with this stuff sometimes and I may forget what is what.
What is the commenting format for the code used in this cfg file? Is it like C++ or what? I'd like to comment my file, because I go for long gaps between messing with this stuff sometimes and I may forget what is what.
It's JSON, which doesn't allow comments. You could add a bogus key/value pair containing comments and find out if their json parser sucks
.It's JSON and from what I could google, no there isn't any real-valid state o making comments in these types of config files.
On another matter, it appears I have been getting 45fps whenever I'm pretty much doing anything
But it is doing this regardless of fidelity setings, even vr low, and no supersampling setup it pretty much stays at 45fps if I'm looking at something interesting.
In fact the one thing that had most impact to improve my fps in game was setting the monitor window to 640*480.
it still dips but seem to come back up quicker.
I'm running an i5 4670K and cpu usage is barely 50% while playing, so if there is a timing issue there, it should be fixable (albeit with great difficulity I bet) on FD's part.
I'm not quite ready to go I7 just because of this when most things point it to be a software fix.
Oh and I might mention, the undock training I tested it seemed ok, it's when I load up an actual game it drops solo or group it starts dropping.
Why are people still using the in game SS, even down sampling? The in game SS seems pretty broken to me so why use it at all?, e.g. SS: 0.65 and render target of 2.0 rather than just disable the in game SS and use, say, render target of 1.65 or so?
I'm using 1.5 at the moment and finding it pin sharp with no adverse affects - also decided to wear my contact lenses for extra sharpness - I tend to wear my glasses outside of work and have been using my Vive with 'bear' eyes, which I thought seemed fine, but corrected it's even better.
I'm using 1.5 at the moment and finding it pin sharp with no adverse affects - also decided to wear my contact lenses for extra sharpness
- I tend to wear my glasses outside of work and have been using my Vive with 'bear' eyes, which I thought seemed fine, but corrected it's even better.
Kinda confused by all this.
I thought the Vive default render target was 1.4.
If that's the case how does switching to 1.6 or even 1.5 make it pin sharp.
I mean I'm not saying it doesn't, 1.6 looks miles better to me, it's just if the default is (as alleged 1.4) then surely it shouldn;t magically fix everything?
I can;t help wondering if ED was overriding this, but I'm not even sure that could be the case since if it was then surely changing the Steam config value would have no effect.
I thought the Vive default render target was 1.4.
If that's the case how does switching to 1.6 or even 1.5 make it pin sharp.
I mean I'm not saying it doesn't, 1.6 looks miles better to me, it's just if the default is (as alleged 1.4) then surely it shouldn;t magically fix everything?
I can;t help wondering if ED was overriding this, but I'm not even sure that could be the case since if it was then surely changing the Steam config value would have no effect.
Last edited:
As wierd as it may sound, .65/2.0 (inGame/Vive) simply looks better than 1.0/1.3.
There seems to be something broken within Elite: Dangerous' rendering engine. I don't think it's optimized sufficiently, which may be why down sampling ED then up scaling the Vive gets you better results than other combinations.
...
You know, I haven't tried 0.65/1.5. I wonder if it'll look better than 1.0/1.0
edit: Heh, replied to magic man, but was more addressing bitstorm.
There seems to be something broken within Elite: Dangerous' rendering engine. I don't think it's optimized sufficiently, which may be why down sampling ED then up scaling the Vive gets you better results than other combinations.
...
You know, I haven't tried 0.65/1.5. I wonder if it'll look better than 1.0/1.0
edit: Heh, replied to magic man, but was more addressing bitstorm.
Last edited:
I've been away from Elite for a couple weeks, so was excited to play with these new VR settings before getting back in the cockpit for a bit tonight. I didn't expect much difference, so was absolutely amazed at the huge increase in crispness and quality!
I'm running at 2.0 / 0.65, and my GTX 1080 handles that beautifully with mostly-Ultra settings (but only FXAA). I'd like to improve the AA a bit, so I'll have another play with the settings tomorrow, see what my 1080 can handle…
But yeah, absolutely amazing difference, thanks for all the experiments that have lead up to this! Cheers guys.
I'm running at 2.0 / 0.65, and my GTX 1080 handles that beautifully with mostly-Ultra settings (but only FXAA). I'd like to improve the AA a bit, so I'll have another play with the settings tomorrow, see what my 1080 can handle…
But yeah, absolutely amazing difference, thanks for all the experiments that have lead up to this! Cheers guys.