RAGEdemon

[RAGEdemon]

Most games are indeed not coded for crossfire or SLi, but you can use the VR game engine as a good indicator, for example, if it uses Unity, UE4 etc. then you can use a profile from a non VR game using the same engine to see if it works. If HTC haven’t blocked it in some manner, it should work just fine.

For example, on the old oculus runtimes, SLi ran without problems.

It would be a cool test to do, if you have the time :)

I’m biased of course, i’m desperate to get SLi working, for obvious reasons. Then I can SLi 1080s for a hopeful 90fps @ ~1920×1440 – what our VR headsets were designed for!

[RAGEdemon]

Oculus have coded SLi/crossfire to be disabled because they claim it causes “lag”. I wish there was a way to bypass this arbitrary limit. I assumed it was the same with the Vive? Can you get normal VR games to work in crossfire too? I cant with Oculus. I was hoping that Vorpx did something to lift this limit with you.

Do you notice lag in head tracking? I feel doubling lag would be worth doubling the performance.

Ralf, does VorpX really require Oculus runtime to be running?

Thanks.

[RAGEdemon]

Hi,

Now I’m confused.

I have tried games such as Portal 2 and Mas Effect 3. All of which work flawlessly in SLi when VorpX is not running. With VorpX enabled, there is virtually no activity on the second card.

If indeed the game can render with both cards simultaneously and increase their performance, then I think that is something huge that needs to be looked into.

Perhaps it’s a case of Crossfire working but SLi not?

[RAGEdemon]

I have been doing some investigating:

The Oculus render resolution absolutely has a great affect on the image.

Please see here:
<unable to paste link; uncool>
Just google “Increase the render resolution of the Rift” – it’s the top link…

This is an extremely important tool because it dramatically changes the Rift’s resolution, sharpness, and antialiasing effects. It completely transforms the Rift IMHO.

I believe the render target resolution on the Rift when this tool is set to 1.0 is the native resolution of the display, i.e. 2160×1200.

Setting this to 1.4 (a render resolution of 1513×1680 I think) gives great image quality in games compared to default! (Some games use values lower than 1!).

With VorpX, We want to get this close to the game resolution, i.e. 1920×1440, so we want to set the debug resolution to 1.2, which brings the render target to 2592×1440.

I have tested this with various settings:

Setting of 0.4 = smooth (high game resolution) but blurry (low Render target).
Setting of 1.0 = smooth and sharp.
Setting of 1.2 = smooth and sharper
Setting of 2.0 = smooth and even sharper (Placebo?)

The interesting thing is that although in other games, this value heavily impacts performance, with VorpX, this setting doesn’t seem to impact performance probably due to the fact that the scene has already been rendered at the game resolution, and is only passed along to Oculus software for final warping. This means that this does not seem to impact performance! I would recommend 2.0 setting, if indeed it’s not a placebo above 1.2.

Ralph, a quick breakdown of the different stages of the rendering, with their respective resolutions would be immensely helpful in our understanding. Right now, I’m just taking stabs in the dark. This could potentially be a huge boon to VorpX Image quality.

For example, my understanding is that:

Image is rendered at Game Resolution then passed into –> Oculus Render Target Resolution which does the anti-lens distortion, which is then passed onto the display –> Rift Display 1080×1200. Am I correct?

Thanks :)

[RAGEdemon]

Thanks for the insight Ralph.

One more question: Does the render scale setting in Oculus Runtime (using the oculus debug tool) affect the quality on top of the resolution used in-game? i.e. if I set it to 2.0, will I get better image quality on top of setting the game resolution to 1920×1440?

[RAGEdemon]

Hi Ralph,

“All headsets warp the image because of the lens distortion which leads to larger pixels in the center. Hence the best render resolution is higher than the display resolution. ”

Yep, I understand that, hence my trying to find an optimal render resolution (‘I’m a huge fan of supersampling with GEDOSATO tool, as well as nVidia’s DSR).

Due to my experience with supersampling, I know how important it is to have exact 1:1, 2:1, 4:1 etc supersampling ratios, as in-between ratios will yield unsatisfactory results.

This is the reason I am trying to ascertain the correlation between the screen resolution of 1080×1200 and the render resolution.

So that I understand, 1920×1440 is indeed a 4:3 resolution. This just does not make sense to me if we are distorting onto a 9:10 screen. surely we would want a 9:10 supersampling resolution?

Now you are saying that there are black bars on the horizontal extremities, which somehow affect the rendering ratio?

I apologise if I am not getting this. Would you kindly elaborate, perhaps by using a diagram?

I am sure the more technical of us would be able to use the information to best calculate our preferred custom resolutions. I personally find this fascinating!

I should add that I use the VorpX cinema setting in games, rather than the VR setting, as it gives me proper head tracking.

[RAGEdemon]

Hi Ralf,

I am sure you are correct. Would you kindly explain the logic behind this resolution of 1920×1440?

The native horizontal resolution is 1080. Using 1920 gives sampling factor of 1.8.
The native vertical resolution is 1200. Using 1440 gives sampling factor of 1.2.

Surely for an optimal resolution, the sampling factors should be identical both horizontally and vertically?

[RAGEdemon]

How does one edit a post?

I have been doing some reading online, and it just gets more confusing as noone seems to know what the default/recommended rendering resolution of the CV1 actually is.

Some people are claiming that its 3024×1680, others are saying its 2700×1600.

Yet some say to half the resolution to 1512×1680, as it should be a “per eye” resolution as Ralph mentioned.

All I want to know is, everything else being equal, what is the optimal game resolution that should be set? Noone on the entirety of the internet seems to be able to answer that definitively :)

If I think logically with my limited understanding, I would double both the horizontal and vertical resolutions of one eye (1080×1200), which would yield 2160×2400 as the optimal rendering resolution, which is 466 million pixels per second at 90fps – pretty close to the “400 million pixels for VR” figure that everyone keeps regurgitating.

[RAGEdemon]

Unfortunately, this is still not clear to me. My curiosity is perked by ralf’s comment “The 100% ideal resolution would be the per eye render resolution of the given headset (which is higher the the screen resolution due to the lens distortion).”

Ralph only mentions that 1920×1440 is an approx to the optimal resolution.

Would someone kindly tell me what the 100% optimal render resolution is?

I can set my resolution to whatever I want using the Custom Resolution Tool, as well as use DSR etc.

[RAGEdemon]

Is this still Valid with v16?

I have a DK2, (waiting on CV1 to ship).

I have set the resolutions but when I go into the games, whether Geometry of Z-buffer, the horizontal line is indeed reduced but the 3D effect has disappeared -both separation an convergence.

Will this still be valid for CV1 which has dual screens?

[Author]

Posts