Reverb G2 lensens investigation MRTV

[SCG_Fenris_Wolf]

That's true, the G2 was overhyped.

 

Currently, there's no proper headset that fulfills all wishes, which for me are: High-Res, edge to edge clarity, FOV ~140° horizontal ~130° vertical, good speakers, <800g weight, eye-relief, no distorsion.

 

Well, at least he warned us about the 8KX. I didn't believe about the distortion and broken world scale, thought it was his fallout with Pimax that had led to it. When I got it, it made me lose my faith. It should get downright banned from getting imported to Europe, what a health hazard.

Edited January 25, 2021 by SCG_Fenris_Wolf

[ICDP]

7 hours ago, SCG_Fenris_Wolf said:

That's true, the G2 was overhyped.

 

Currently, there's no proper headset that fulfills all wishes, which for me are: High-Res, edge to edge clarity, FOV ~140° horizontal ~130° vertical, good speakers, <800g weight, eye-relief, no distorsion.

 

Well, at least he warned us about the 8KX. I didn't believe about the distortion and broken world scale, thought it was his fallout with Pimax that had led to it. When I got it, it made me lose my faith. It should get downright banned from getting imported to Europe, what a health hazard.

 

I had heard about him giving bad reviews of the 8KX but by then I had already orderd an 8KX and had stopped listening to anything MRTV said.  Thankfully my 8KX is perfect and I get no serious distortion.  Can I ask if you got the thin or the thick face foam on your 8KX, we were meant to get both but I only got the thin one and others reportedly got the thick one.  When I finally got my thicker foam delivered it was absoutely horrible, the HMD became totally unusable with massive distortions and blurry focus.  So I believe many people simply got the wrong (for them) facefoam and declared the 8KX a pile of crap.

 

So once agian it is a case of Pimax as a company repeatedly shooting themselves in the foot with their dodginess/ineptitude, or both.  People can't even use the "Pimax are new and still learning" defence, becasue they have been at this for years now and messing up for them is second nature.

 

As I said above, for me the 8KX is a great HMD techincally and it feels well built but Pimax as a company suck and the price is a joke.  So I can only recommend the Pimax 8KX for those who are prepared to pay a lot to own arguably the best HMD on the market from a techincal aspect, but with lots of risk and caveats.

Edited January 25, 2021 by ICDP

[Alonzo]

On 1/24/2021 at 3:10 AM, coconut said:

The claim that you need 100% supersampling in the G2 in SteamVR to achieve whole-lens clarity is IMO incorrect. It's neither necessary nor sufficient. Supersampling doesn't affect lenses. You will gain in picture quality, which of course plays a role in what people call "clarity", but it's just wrong terminology.

If a HP representative stated that, well, they are wrong too. Maybe trying to blame a "weakness" of the lenses (which I actually think are more than decent) on the user's system, because running at 3100p is out of reach for most hardware and games. Or they never actually said or meant exactly that, and the person who relayed that misinterpreted them.

 

This is my thinking. What you put on the pixels does not affect whether the lens gives you a sharp rendition of those pixels, surely?

 

On 1/24/2021 at 7:52 AM, SCG_Wulfe said:

I want to clear up any misunderstanding and because I think it will very much help people achieve the most out of the G2. What the rep stated is that the supersampling is applied exponentially in the centre of the lenses. This means at lower settings it also falls off exponentially at the edges. I have noticed a marked difference in how well I can read text and see things on the edges at higher SS. 
 

There are two factors going on. I agree that changing SS does nothing to alter the sweet spot as it is impacted by  lens construction.

 

On 1/24/2021 at 2:54 PM, SCG_Fenris_Wolf said:

Exactly. It's built-in Fixed Foveated Rendering, basically.

 

With a ludicrously high 100% though.

 

This is the thing I don't understand, though. Are they still rendering a 'square' image like we see in game? How can this achieve an exponential falloff result? In all the things I've read about the way the image is warped and then corrected by the lenses, it would seem the least panel pixels are used in the center of the lens because it's the least warping.

[coconut]

On 1/30/2021 at 2:45 PM, Alonzo said:

This is the thing I don't understand, though. Are they still rendering a 'square' image like we see in game? How can this achieve an exponential falloff result? In all the things I've read about the way the image is warped and then corrected by the lenses, it would seem the least panel pixels are used in the center of the lens because it's the least warping.

 

Lenses in headsets typically unzoom the center, giving you a higher pixels-per-degrees resolution in the center (this is good). This higher angular resolution is of course wasted if it isn't matched by a matching resolution in the center of the image produced by the game. And because traditional rasterization is "straight", it also means you are producing unnecessarily many pixels at the image's sides (which is bad).

 

On 1/30/2021 at 2:45 PM, Alonzo said:

On 1/24/2021 at 3:52 PM, SCG_Wulfe said:

I want to clear up any misunderstanding and because I think it will very much help people achieve the most out of the G2. What the rep stated is that the supersampling is applied exponentially in the centre of the lenses. This means at lower settings it also falls off exponentially at the edges. I have noticed a marked difference in how well I can read text and see things on the edges at higher SS. 

 

I don't see that.

 

On 1/30/2021 at 2:45 PM, Alonzo said:

On 1/24/2021 at 10:54 PM, SCG_Fenris_Wolf said:

Exactly. It's built-in Fixed Foveated Rendering, basically.

 

With a ludicrously high 100% though.

 

This is the thing I don't understand, though.

Neither do I. If they've managed to get FFR working in the jungle that is PC games, that would be a major achievement. I remain skeptical.

 

If @SCG_Fenris_Wolfwas referring to the distortion of the lens as "built-in Fixed Foveated Rendering", I would object to that wording. Most headsets have it to varying degrees, and it offers an increased angular resolution in the center, but it doesn't come for free.

 

Now if the rendering pipeline implements FFR, meaning it skips rendering of pixels outside of the foveated area, and is a bit too aggressive, you would need to compensate that with high supersampling. It would be somewhat self-defeating though.

[SCG_Fenris_Wolf]

That's exactly what they have stated. I only go by what they've stated.

 

By first principle, this actually is fixed foveated rendering.

If the rendered pixel density falls off to the outer degrees of the picture in a function, whole it remains steady in the center, that is fixed foveated rendering in its base-form. Let's write it out and it's easier to understand:

 

If pixel density 

 

At the outer ring of FOV:

Falls off by X when going to 60% from 100% SteamVR's SS,

At the inner cone of FOV:

Falls off by Y when going to 60% from 100% SteamVR's SS,

 

And Y < X is true,

And definition of FFR to have different pixel densities based on fovean distance to center is true,

 

Then as Y < X is true in Reverb G2, and as the definition is true, you have de facto built-in fixed foveated rendering in the HP Reverb G2.

 

 

Whether the characteristic is acquired through a distortion profile function, or the slap on a button (like Pimax'), does not matter to the question of whether it is not fixed foveated rendering or it is.

[VR-DriftaholiC]

You need 100% SS to have proper lens distortion correction. 

[BlueHeron]

Seems MRTV was right, I'm having a much better experience after using prescription lens inserts and ensuring a proper fit.  I barely notice the sweet spot at all now.  Suddenly it's a hug step up from the Rift S.

[coconut]

On 2/3/2021 at 2:37 PM, SCG_Fenris_Wolf said:

That's exactly what they have stated. I only go by what they've stated.

May I ask for a source? A google search gives lots of links to the Omnicept edition, but very little on the headset itself.

Only thing of interest I could find was this:

Unsurprisingly, it's not every application that supports FFR. The cliff house does, and for that case I can understand that lower than 100% would give you pixelated sides.

Not relevant for IL-2 GB though. I've seen the comment referring to some discussion with the devs on some discord, but without seeing the discussion, I'm not inclined to believe it.

 

On 2/3/2021 at 2:37 PM, SCG_Fenris_Wolf said:

Let's write it out and it's easier to understand: [...]

I'm not sure what you mean by pixel density, because I don't know which stage(s) of the VR rendering pipeline you are including.

 

If you mean the angular resolution, then it's a factor of the lens and the panels. Rendering resolution by the game isn't relevant.

 

If you mean per-pixel supersampling factor due to lens distortion correction, then supersampling actually goes up as you move to the sides, not down. See the page from Doc_OK posted earlier in this topic. Moreover the relation between supersampling on the sides and supersampling in the center shouldn't change with the rendering resolution, because that's entirely determined by the distortion profile.

 

If you are talking about the rendering done by the game, then there is no notion of variable pixel density because the game doesn't implement FFR.

 

Which makes me think. Has anyone tried using Nvidia's VRSS by forcing it with NVInspector? Results were dubious when people tried it with the PiMax, likely because of the angled/wide FOV, but it might work with the G2.

 

7 hours ago, VR-DriftaholiC said:

You need 100% SS to have proper lens distortion correction. 

 

I don't.

[I.JG3_CDRSEABEE]

40 minutes ago, BlueHeron said:

Seems MRTV was right, I'm having a much better experience after using prescription lens inserts and ensuring a proper fit.  I barely notice the sweet spot at all now.  Suddenly it's a hug step up from the Rift S.

What brand inserts are you using? I see many companies offering lenses. Some with Blue Light filter????  Anti glare????  Are these needed?

 

When I wear my glasses normally and focus on an object while turning my head the object gets blurring. I guess like people are describing in the G2.  Do the inserts make this effect go away?

 

 

[Alonzo]

7 hours ago, coconut said:

Which makes me think. Has anyone tried using Nvidia's VRSS by forcing it with NVInspector? Results were dubious when people tried it with the PiMax, likely because of the angled/wide FOV, but it might work with the G2.

 

I think I tried VRSS when I had the Pimax Artisan, but it was a while ago. As far as I recall, the problem with VRSS is that it varies the MSAA which is being applied to the image, based on a mask/GPU usage (more headroom = bigger mask and/or higher MSAA). So it's super-sampling pixels in the middle of the display when you have headroom to do so.

 

The problem is that the G2 has a huge number of pixels, so everyone has MSAA switched off already and we're really just trying to get the pixels rendered at 0x MSAA, SteamVR 100%. I guess the picture might look better with 4x MSAA VRSS + SteamVR 50% (or something).

 

We really need FFR to be a SteamVR / OpenXR native driver-level feature. Then games would "just work" and we could get good performance out of a fancy headset.

[BlueHeron]

21 hours ago, CDRSEABEE said:

What brand inserts are you using? I see many companies offering lenses. Some with Blue Light filter????  Anti glare????  Are these needed?

 

When I wear my glasses normally and focus on an object while turning my head the object gets blurring. I guess like people are describing in the G2.  Do the inserts make this effect go away?

 

 

 

The inserts reduce this affect to an acceptable level, but still not as good as the Rift S sweet spot.  My lenses are from vr-wave.  I think the blue light filter and anti glare were included in the price.  Supposedly this reduces eye strain.

[VR-DriftaholiC]

On 2/4/2021 at 11:54 AM, coconut said:

On 2/4/2021 at 4:14 AM, VR-DriftaholiC said:

You need 100% SS to have proper lens distortion correction. 

 

I don't.

 

Let me put it this way, If you turn SS down to match the 2160x2160 resolution you will be below the pixel resolution of the panels in the portions of the screen where the image is stretched by the lens distortion correction. If you want a minimum 1:1 panel pixel to rendered pixel resolution in 100% of the viewable area then you absolutely need it set to 100%

 

It's touched on here:

http://doc-ok.org/?p=1694

 

Not only will the center of the lenses render below 1:1 you will also sacrifice chromatic aberration correction if you try and set the steam SS resolution to say 50% to match the panel resolution.  Better to lower the resolution in game.

Edited February 7, 2021 by VR-DriftaholiC

[coconut]

1 hour ago, VR-DriftaholiC said:

Let me put it this way, If you turn SS down to match the 2160x2160 resolution you will be below the pixel resolution of the panels in the portions of the screen where the image is stretched by the lens distortion correction

Yes, I agree with that.