OCULUS RIFT DK2 2560x1440 downscaleing SO Amazing

Are you talking with graphics maxed out, and in Rift native resolution and maximum framerate ? While framerate has always been a rather secondary issue for me, I think in the case of the Rift, you really need the game to run at very smooth rates.
 
I use Downscaling too and not only on ED. But I use a bit higher Resolution:

The DK2 could run at 2688 x 1512 @75 Hz.

That is the highest Resolution at 75 Hz that I know so far. Didn't try higher one.
____________

One Thing is on VR really important beside Latency: perfect Anti Aliasing.

Because the Eyes get two different Images the step effect is also different on each Eye and that looks much more worse than on a Monitor and it also ruin the Immersion. Downscaling is the best of all Anti Aliasing Methods and thats why NVidia List her new Downscaling (DSR) Feature under their new "VR Direct" brand.
 
Last edited:
So what is the difference between making your Gcard work harder to produce a higher resolution, then make it downscale the image to a lower resolution and just running at the lower resolution in the first place?
 
5 Example Pics:

1. 1080p no AA: Scene 1 - Scene 2

2. 1080p no AA Downsampling (from 3200x1800): Scene 1

3. 1080p no AA "perfect" (from 3200x1800 with alt+F10 = 12800x7200): Scene 1 - Scene 2

Please watch in 1:1 Size for correct Display what Downsampling looks like

You can see on 1 much stair effects on many Parts of the Image, on 2 there is much less stair effect and that came only from resizing the Image to the lower 1080p Resolution without to lost the Sharpness of the Image. So it is better than any AntiAliasing Technique that is often not accurate enough, smooth the Image (like FXAA) and/or work only on Parts of the Image and not on every Object.

Players who plays Games like FarCry with many Trees in a Forest knows how bad it looks without good AntiAliasing when you look up into the sky between Trees.
 
Last edited:
As it stands right now, a GTX 980 will not run Elite at a locked 75fps at 1080p on a DK2.

So no chances of a GTX 970 running a CV1 at a locked 90fps+ at 1440p+... :p

I'm not sure I agree, currently there is optimization to come on both the graphics engine and the Rift implementation. The Rift implementation is still only based on the OR SDK 2.x version though that will change in two days time. The Rift 4.x drivers are also in need of some work yet.

There's also a big difference visually between what's "good enough for the experience" and "maxed out". I'm quite surprised that a GTX 980 can't run at 75fps+ on a DK2 though as my Radeon 290X does this with ease with everything on high and full draw distance except for shadow quality and lens flare (medium for both) and that was mostly because I found very little difference in shadow quality and the lens flare was a bit distracting.

The GTX 980 and very probably the 970 should beat the 290X unless the Nvidia driver isn't working as well in the current build (back to future optimization).

Could you run it well enough for a great experience in CV1 on a GTX 970, with some tweaks for an acceptable framerate? I'd be surprised if you couldn't.
Could you run it with everything maxed - of course not.
 
I'm not sure I agree, currently there is optimization to come on both the graphics engine and the Rift implementation. The Rift implementation is still only based on the OR SDK 2.x version though that will change in two days time. The Rift 4.x drivers are also in need of some work yet.

There's also a big difference visually between what's "good enough for the experience" and "maxed out". I'm quite surprised that a GTX 980 can't run at 75fps+ on a DK2 though as my Radeon 290X does this with ease with everything on high and full draw distance except for shadow quality and lens flare (medium for both) and that was mostly because I found very little difference in shadow quality and the lens flare was a bit distracting.

The GTX 980 and very probably the 970 should beat the 290X unless the Nvidia driver isn't working as well in the current build (back to future optimization).

Could you run it well enough for a great experience in CV1 on a GTX 970, with some tweaks for an acceptable framerate? I'd be surprised if you couldn't.
Could you run it with everything maxed - of course not.

There is the EVGA G1 970 out that has performance on par with the 980 in some games and it overclocks considerably, I think even a 970 will run Elite good enough on CV1, but a 970 G1 seams to be the way to go.
 
5 Example Pics:

1. 1080p no AA: Scene 1 - Scene 2

2. 1080p no AA Downsampling (from 3200x1800): Scene 1

3. 1080p no AA "perfect" (from 3200x1800 with alt+F10 = 12800x7200): Scene 1 - Scene 2

Please watch in 1:1 Size for correct Display what Downsampling looks like

You can see on 1 much stair effects on many Parts of the Image, on 2 there is much less stair effect and that came only from resizing the Image to the lower 1080p Resolution without to lost the Sharpness of the Image. So it is better than any AntiAliasing Technique that is often not accurate enough, smooth the Image (like FXAA) and/or work only on Parts of the Image and not on every Object.

Players who plays Games like FarCry with many Trees in a Forest knows how bad it looks without good AntiAliasing when you look up into the sky between Trees.

Would like to have seen a sample at 1080p with 4 x AA as well.

Whilst I can see that running a 2560 or 4k monitor and game at that resolution has obvious advantages, I am not entirely convinced it is worth downsampling (yet).

Is this something that makes a bigger difference with OR than on a single screen?
 
There is the EVGA G1 970 out that has performance on par with the 980 in some games and it overclocks considerably, I think even a 970 will run Elite good enough on CV1, but a 970 G1 seams to be the way to go.

Yeah, there looks to be a lot of headroom even in the early Maxwell 9x0 GPUs and the pricing is competitive for a change too. The 970 already looks like one of those cards that comes along every now and then that hits a sweet spot, like the GTX 460 did a few years ago.
 
Is this something that makes a bigger difference with OR than on a single screen?

It makes a huge difference on the DK1 Rift.

I have no experience with it on a screen though.

There's also a big difference visually between what's "good enough for the experience" and "maxed out". I'm quite surprised that a GTX 980 can't run at 75fps+ on a DK2 though as my Radeon 290X does this with ease with everything on high and full draw distance except for shadow quality and lens flare (medium for both) and that was mostly because I found very little difference in shadow quality and the lens flare was a bit distracting.

The GTX 980 and very probably the 970 should beat the 290X unless the Nvidia driver isn't working as well in the current build (back to future optimization).

Could you run it well enough for a great experience in CV1 on a GTX 970, with some tweaks for an acceptable framerate? I'd be surprised if you couldn't.
Could you run it with everything maxed - of course not.

The Beta 2 build does seem to run things better. But that's only with eyeballing my hands-on experience - and not having any fps counters to hand etc. So yes, there is certainly a lot more room for optimizations.

Interesting about the 290X. Take a look at this post - I would be interested to know how this compares to your performance:

http://forums.frontier.co.uk/showpost.php?p=787238&postcount=3562

Have my 980 overclocked to 1355Mhz base, 1444Mhz boost. Memory left at stock.
Playing on the High preset with Vsync off.

In stations I still see drops below 75fps, the minimum I saw it drop to was 60fps but that was only for a second, mainly it fluctuated between 80 and 65fps depending on where I was in the station (i.e. docked and looking at the station from the outside seemed the worst).
Ring systems seemed OK though, saw one drop to 70fps for a second but it mainly fluctuated between 80 and 75fps.

You have to really look for judder though, if you weren't looking for it I think you wouldn't really notice. with AA off/Shadows lowered etc.... I'd be surprised if it dipped below 75fps at all.

I did record a video but it's a bit wonky due to it being recorded in portrait, tried rotating it and uploading to youtube but the output wasn't that good and you couldn't really make out the FPS so I abandoned it.

Will repeat the process on Tuesday/Wednesday with Beta 2 and report back.
 
I'm not sure I agree, currently there is optimization to come on both the graphics engine and the Rift implementation. The Rift implementation is still only based on the OR SDK 2.x version though that will change in two days time. The Rift 4.x drivers are also in need of some work yet.

There's also a big difference visually between what's "good enough for the experience" and "maxed out". I'm quite surprised that a GTX 980 can't run at 75fps+ on a DK2 though as my Radeon 290X does this with ease with everything on high and full draw distance except for shadow quality and lens flare (medium for both) and that was mostly because I found very little difference in shadow quality and the lens flare was a bit distracting.

The GTX 980 and very probably the 970 should beat the 290X unless the Nvidia driver isn't working as well in the current build (back to future optimization).

Could you run it well enough for a great experience in CV1 on a GTX 970, with some tweaks for an acceptable framerate? I'd be surprised if you couldn't.
Could you run it with everything maxed - of course not.

I was playing tonight on my overclocked 980 (1368mhz/1457Mhz boost, overclocked it slightly more than the post Granite references ;) ) @1080p on the DK2 and was still getting a small amount of judder at chango and beagle 2. Dropped the settings down to low to try and get rid of it but there was still a slight amount. Checked the FPS and it was dipping below 75fps slightly.

I think certain ships have an additional effect on FPS. I tried the viper, hauler, cobra and lakon 6 tonight and the cobra seemed the worst.

I think maybe playing online has an effect too, played online for a while and had what seemed like more judder than playing singleplayer (which I switched to as online seemed extra crashy tonight)

Like you say though hopefully performance increases and driver improvements will iron this out.

Here's a screenshot of GPUz so you know I'm not talking rollocks.
25jyzoz.png
Here my Heaven benchmark score so you know there's not something up with my PC.
2mfhwfc.png
Just benchmarked it using fraps to store the fps while playing.
I flew out of chango dock, rotated the ship, flew back in and docked.
I was in an Lakon 6, default low settings.
Here's the FPS fraps recorded.
FPS
68
70
66
68
72
74
76
74
76
67
67
68
66
68
67
66
68
66
65
66
64
65
65
67
72
72
72
66
66
66
63
63
68
70
67
69
71
70
68
69
66
68
69
68
67
67
66
63
63
64
65
63
64
64
64
64
64
64
65
65
65
68
66
68
65
67
66
66
67
71
75
76
75
74
74
74
74
73
76
73
75
73
73
74
75
74
74
71
73
73
73
75
76
73
74
73
74
75
75
73
73
75
74
75
73
71
76
74
75
74
75
70
68
71
70
66
68
67
69
68
68
72
74
70
71
74
72
74
75
72
69
70
68
68
68
72
66
67
67
65
69
65
66
70
71
72
74
73
76
74
75
73
71
70
69
67
69
69
65
71
69
65
66
70
75
67
68
66
67
72
74
70
65
70
68
67
66
68
66
68
 
Last edited:
So what is the difference between making your Gcard work harder to produce a higher resolution, then make it downscale the image to a lower resolution and just running at the lower resolution in the first place?
Supersampling (which is what downscaling is more correctly called) gives you a better representation of the underlying image that you're trying to display.

Consider 2x supersampling, that is you're rendering double the resolution in both x & y. For every pixel in the display image, you're essentially rendering 4 pixels internally. This equates to giving you 4 samples per pixel of the underlying image instead of just the 1.

So let's consider a simple example of a polygon edge slowly moving across a pixel, without supersampling, the pixel is either on or off as the edge crosses the center of the pixel. With 4x supersampling, you have 4 sample points within the pixel, so as the polygon moves across, you'll have off->25% on->50% on->75% on -> 100% on.

Consider this also with a very small object that may not even take up the entire pixel - say a ship at long range - with more samples, you get more of a chance that one of your samples will hit the object, so instead of black, you get some degree of color, to give you an indication that there's something there.

Consequently you get better fidelity of the scene being sampled, which gives you more apparent detail, even though the final image is the same resolution, this means that small details (such as the edges of text) look a lot better and it all becomes more readable.

The down side is that supersampling is REALLY inefficient, you're using a ton of samples whether you need them or not - you only really need the additional samples where you have high frequency detail that isn't getting sampled correctly, ie. sharp edges, which is where antialiasing techniques like MSAA come in, which in theory distribute your rendering samples where you need them.

It's worth noting that some antialiasing techniques like FXAA don't actually increase the fidelity of the image, that's really just a technique for removing objectionable details rather than giving you more information.

It's also worth noting that supersampling is handled by the Elite engine and can be enabled by adding <SSAA>n</SSAA> in the graphics.xml file, but it's not officially supported and the engine tends to delete the entry every time you quit, hopefully this won't be the case as development progresses.
 
Last edited:
Just benchmarked it using fraps to store the fps while playing.
I flew out of chango dock, rotated the ship, flew back in and docked.
I was in an Lakon 6, default low settings.
Here's the FPS fraps recorded.
FPS
68
70
66
68
72
74
76
74
76
67
67
68
66
68
67
66
68
66
65
66
64
65
65
67
72
72
72
66
66
66
63
63
68
70
67
69
71
70
68
69
66
68
69
68
67
67
66
63
63
64
65
63
64
64
64
64
64
64
65
65
65
68
66
68
65
67
66
66
67
71
75
76
75
74
74
74
74
73
76
73
75
73
73
74
75
74
74
71
73
73
73
75
76
73
74
73
74
75
75
73
73
75
74
75
73
71
76
74
75
74
75
70
68
71
70
66
68
67
69
68
68
72
74
70
71
74
72
74
75
72
69
70
68
68
68
72
66
67
67
65
69
65
66
70
71
72
74
73
76
74
75
73
71
70
69
67
69
69
65
71
69
65
66
70
75
67
68
66
67
72
74
70
65
70
68
67
66
68
66
68

I think the Lakon 6 is a performance killer. I have had a lot of frame rate drops with that ship. I'm flying a Lakon 9 at the moment.

On my GTX 780, I am now getting 60+ fps outside the stations and around 40fps inside the stations. This is with default high, and medium textures and shadows no AA at 1920x1200 on a DK1.

If I set it to 1280x800 I get 60+ fps inside the station again on default high (medium textures, medium shadows, no AA).


-------------------
------

I just tried it on low settings.

1920x1200 - 44-46fps inside Chango Dock
1280x800 - 68-72fps inside Chango Dock
 
I plumped for the DK2 now, on the basis that when CV1 ships, I can sell the DK2 or gift it on or something, and I wants my 3D cockpit before Christmas, so I can nerd away the whole holiday in the howling vasty dark.

So thanks for this tip, I'll definitely do that. This is also a trick used a lot in visual effects - sometimes, it's faster to render your frames at 1.5xHD, then downscale, than to try to anti-alias edge artefacts away.
 
you wont regret it, i just cant play ED on my 27 inch monitor after getting my DK2, your there its hard to explain to people that dont have one, how much of a difference it makes, give it 3 yrs your have 4k versions and the cards to run them, your see nvidia and ati realeasing cards that are geared towards VR. I use DSR in the nvidia control panel, and it does help, would help if there was an option to adjust the text in game, not sure if it can be done.
 
you wont regret it, i just cant play ED on my 27 inch monitor after getting my DK2, your there its hard to explain to people that dont have one, how much of a difference it makes, give it 3 yrs your have 4k versions and the cards to run them, your see nvidia and ati realeasing cards that are geared towards VR. I use DSR in the nvidia control panel, and it does help, would help if there was an option to adjust the text in game, not sure if it can be done.

What res are they at now ?

I am really thinking hard about it, not sure. I have a 42" monitor and sit only around 3ft away from it, so its pretty good 120hz. Just a LED, but it looks really good the game. I cap the fps @120 as I am lucky to have high rates, so the game is ultra smooth really. But this is all @ 1920x1080 so not a huge res by any means, but does for me at present.
But if OR is at a good res and smooth, then I might just get one.. The rates have to be good though, can't do with rough gaming, got to be smoooooth..
 
Last edited:
It does work great, and it's something for the future when our systems have more power and consumer Rifts are available.

I used the nvidia control panel and that new DSR feature to force the rift into 4K mode.... Everything looks mind blowingly amazing.. Obviously this was a complete slideshow so didn't really play, just admired sat in a station and an asteroid field. 10 FPS in DK2 is migrane material so didnt stop in for long.

As OP is doing, I run Elite at the same Res as him.. the sharper edges, does make a nice difference, drops a few frames at times but managable.
 
Back
Top Bottom