Sony VR Headset/Project Morpheus/PlayStation VR

[Deleted member 86764]

Taking into consideration the three possible configurations of Morpheus, it’s quite easy to calculate the pixels per second of each of the devices and they’re as follows:

Device: Morpheus
HZ: 120
Horizontal: 1920
Vertical: 1080
Pixels/second: 248832000

Device: Morpheus
HZ: 90
Vertical: 1920
Horizontal: 1080
Pixels/second: 186624000

Device: Morpheus
HZ: 60
Vertical: 1920
Horizontal: 1080
Pixels/second: 124416000

Device: Vive
HZ: 90
Vertical: 2160
Horizontal: 1200
Pixels/second: 233280000

Device: Rift
HZ: 90
Vertical: 2160
Horizontal: 1200
Pixels/second: 233280000

So it’s easy to conclude that for games that are native 120hz for Morpheus and 90hz for Vive/Rift (even if there’s very few of them) will actually give a better experience on the Morpheus. We can also conclude that both the Vive and the Rift will perform better on most higher spec games providing the hardware powering the devices are of a good enough quality.

If we consider the Morpheus at 60hz is pushing 124 million pixels, it’d take a PC GPU that’s double the performance (as 233m is about 2x 124m) for exactly the same graphical settings (bar resolution) to compare. Obviously these games would look better on PC since you have native 90hz and a higher resolution. But then the reprojection will go some ways (even if small) to negating some of that deficit.

I can’t help but think Sony have played a very smart game when creating their device, it can both perform better than the PC headsets under particular scenarios (no matter how powerful the PC). And also keep up with the others when they’re combined with a GPU that’s more than twice as powerful.

It’s only really the 90hz configuration of Morpheus that’ll really give a truly under par experience when up against the others. I do agree that options are good though, but this is the least viable option to my mind.

 

[Shifty Geezer]

PIxels per second is a spurious metric. 320x200 @ 10kHz would give 640,000,00 pixels per second, but there's no way that'd be better than any of the current headset configurations. If the pixels are more noticeable on Morpheus and the higher framerate doesn't make a massive difference, 1080p at 120 Hz might not be better than 90 Hz at 1200p. I don't believe 90 Hz is inadequate. I think the only reason Morpheus is 120 Hz is because that's a straight forward interpolation of 60 Hz. More important is low latency motion input, optics, and everything else.

 

[Deleted member 86764]

I'm not so sure. Oculus themselves have actually used pixels/second as a new metric when describing improvements. I guess if further iterations of the Rift/Vive/PS VR have increased hz, then it would be evidence enough that increased (i.e., >90) hz are important.

How much greater than 90hz (if at all) is necessary remains to be seen. But I wouldn't use an extreme example (320x200 @ 10kHz) as a way of describing that it's not important. You could same the same about framerate, but obviously a higher framerate is always preferable (unless it's the extreme).

There need to be a balance and Sony have created an incredible balance for the limiations in the hardware of their box.

 

[DieH@rd]

And there is also a "visual difference" between PC and PS4 VR panels. Sony promoted that they are using full RPG subpixel arrangement, while Vive/Oculus are probably using AMOLEDS with pentile subpixel arrangement. There was a talk while back that Sony's approach gives better "percieved" resolution.

Maybe full RGB was only in Morpheus v1 [the one that used LCD]. V2 switched to some unknown type of OLED.

 

[Shifty Geezer]

I'm not so sure. Oculus themselves have actually used pixels/second as a new metric when describing improvements.

Because it gives them a marketing number. But as not all pixels per second are equal, it doesn't provide a meaningful comparison. Another hypothetical example in contrast to my very low res, high framerate, would be 8k at 10 fps. 7680x4320 x 10 == 331,776,000 pixels per second. Is that 1.7x better than PSVR at 1080p90 and 186,624,000 pps?

Edit: For tracking progress of a platform like OR, and seeing the improvement of each iteration, pps makes some sense as a nice all round reference value. But it's no good for determining the quality of a VR experience between different resolutions and framerates.

 

[hesido]

I think pixels/sec tell more about computational requirements than the perceived quality.

 

[MrFox]

Is there evidence of this?

The obvious advantage of the 60 -> 120 is that it requires less performance for an experience that doesn't suck. I've seen no evidence yet though that its a fundamentally better experience - even in just slow moving games - than 90 native fps.

Reprojected or not, 120Hz provides both a lower latency and lower persistence. Latency is king.

The question of which is better is about the artifacts. If there's a lot of reprojection artifacts, the experience will be inferior to 90. If there are very few, or if they are imperceptible, the experience will be superior to 90, and effectively doubling the processing power available.

You can't make blanket statements about this without understanding what these artifacts are, and what aspect of the scene makes them more or less perceptible. We know that these situation are directly related to translation velocity.

If the reprojection is improved with a more extensive algorithm, using a Z buffer and motion vectors, it could be closer to perfection in all cases. I'm assuming this would require a lot of work on the game engine to support this, while currently I think they said Sony provides a very plug-and-play method (mostly just shifting rotating and deforming the output buffer based on the headset sensors).

 

[MrFox]

And there is also a "visual difference" between PC and PS4 VR panels. Sony promoted that they are using full RPG subpixel arrangement, while Vive/Oculus are probably using AMOLEDS with pentile subpixel arrangement. There was a talk while back that Sony's approach gives better "percieved" resolution.

Maybe full RGB was only in Morpheus v1 [the one that used LCD]. V2 switched to some unknown type of OLED.

They said the V2 OLED is a custom pixel arrangement with full RGB per pixel.

I would guess both Vive and Oclulus have already switch to full RGB too, or will do so at release. Previous prototypes were pentile because they would need a custom screen for full RGB. All existing smartphones screens are pentile at these resolutions.

 

[Deleted member 86764]

I think pixels/sec tell more about computational requirements than the perceived quality.

Then they'd better all drop their display hz down to 50.

 

[Deleted member 13524]

For the VR, graphics will be 3D and with frame rate more important than resolution, native res is going to be rare anyway, with purposeful blurring to hide physical pixels on purpose.

I'd guess that 1080p resolution and >=60FPS are the only things that are going to be kept at all costs with Playstation VR. I've been hearing a lot of people saying that 960*1080p per eye is already bad enough as it is. And even then they're probably going to FXAA it to a pulp.

 

[pjbliverpool]

I think the original assertion was that Morpheus would give an inferior experience because games will always run at a lower framerate. That is fundamentally untrue since the device actually has the highest hz of any headset.

That certainly wasn't my assertion. I was suggesting that all other things being held equal, 90fps native is likely a better experience than 60fps re-projected to 120fps. Naturally if Morpehus is running at a native 90fps it will offer a similar experience (from a frame rate standpoint) to OR/Vive and at 120fps it will be a better experience. Obviously at a huge cost to core graphics or resolution though.

Unless the minor resolution increase negates it - which I doubt.

It could be argued that one you have sufficient frame rate to give 'presence', any additional power is better put into increasing resolution. I haven't tried any of the head sets myself so I'm not in a position to comment from first hand experience, but one of the main compliants about VR to date is about the resolution with most people complaining that 1080p isn't enough (and even the higher resolutions of Vive/OR not necessarily being enough yet).

So it’s easy to conclude that for games that are native 120hz for Morpheus and 90hz for Vive/Rift (even if there’s very few of them) will actually give a better experience on the Morpheus.

This pre-supposes that 1080p is not a problem but 90fps is. And that's pretty much the opposite of most first hand reports I've read.

 

[pjbliverpool]

Reprojected or not, 120Hz provides both a lower latency and lower persistence. Latency is king.

The question of which is better is about the artifacts. If there's a lot of reprojection artifacts, the experience will be inferior to 90. If there are very few, or if they are imperceptible, the experience will be superior to 90, and effectively doubling the processing power available.

I'm not sure how much sense the concept of fast or slow paced games makes in VR anyway. The artifacts are going to be caused by fast head movements and that's possible in any genre of VR game. I guess we'll have to wait and see how truly effective the 60 -> 120 Hz is. In any case it's a good solution to the limited performance available in the PS4. Maybe if it does keep artefacts to a near zero then we'll see 180hz displays in the next generation of PC headsets that allow re-projection from 90fps. That would certainly be a nice "free" way to double performance.

 

[MrFox]

I'm not sure how much sense the concept of fast or slow paced games makes in VR anyway. The artifacts are going to be caused by fast head movements and that's possible in any genre of VR game.

No, the reprojection is correcting fast head rotations almost perfectly. It's what it does best.

 

[Deleted member 86764]

I was suggesting that all other things being held equal, 90fps native is likely a better experience than 60fps re-projected to 120fps. Naturally if Morpehus is running at a native 90fps it will offer a similar experience (from a frame rate standpoint) to OR/Vive and at 120fps it will be a better experience.

[snip]

This pre-supposes that 1080p is not a problem but 90fps is. And that's pretty much the opposite of most first hand reports I've read.

I think the best to look at the combined resolution and framerate, hence pps. Providing you're not looking at the extremes.

People suggest that 120fps is noticeable on a normal display. Now if you have that display strapped to your face and registering your movements, I'd say it's even more noticeable.

I have no idea what the upper limit of hz is for a VR display yet, only I suspect the refinements of these devices to go higher and higher suggests we haven't met that threshold yet. CV2 will be higher than CV1, as Vive 2 will be higher than Vive 1.

On a three dimensional scale of horizontal, vertical, and framerate the easiest calculation to compare differences is pps. I suspect I'm alone with that view in this forum though.

 

[Shifty Geezer]

I'm not sure how much sense the concept of fast or slow paced games makes in VR anyway.

Whether the game requires you to have fast head movements, or whether your throwing your head around just trying to make the reprojection break. But it's moot because...

The artifacts are going to be caused by fast head movements and that's possible in any genre of VR game.

It was calculated in one of these VR discussions on this board how many pixels the fastest head movement could displace in a single frame. It was very low and very much in the periphery, so basically it should be unnoticeable. Also you can't see squat when engaged in fast head movements. Unless I'm a defective example, my inability to track a point on this screen and keep the world unblurred when violently moving my head shows that visual artefacts while engaged in fast head movements are the least of your worries. I expect one will be puking merrily if one's head actions are enough to break the interpolation of a few shifted pixels each frame.

 

[Shifty Geezer]

On a three dimensional scale of horizontal, vertical, and framerate the easiest calculation to compare differences is pps.

It is indeed the easiest comparative calculation, but the comparison isn't qualitative. Take 1080p120 versus 2160p60. Both have exactly the same pps. Which is the better experience? Are they identical? They are clearly different - one having better fidelity but inferior motion. Depending on user and use case, either option could be the better choice. Hence the pps number means squat. there are two different qualities in effect, fidelity and motion, and they require a separate metric each to compare. There's no metric that sanely serves to convey the 2D quality matrix, where a display fits in a matrix of resolution and framerate.

 

[Deleted member 86764]

Take 1080p120 versus 2160p60. Both have exactly the same pps. Which is the better experience? Are they identical?

I realise you meant this as rhetorical, but I think it's important to actually consider. In the VR space I suspect (I can't say for certain as I haven't tried anything since Virtuality) that 1080x120 would be preferable. On a PC or home screen? Definitely 2160x60.

 

[MrFox]

Anyone remember when the previous CTO talked about future roadmap for playstation, saying their goal is to have a playstation that can do 300Hz in 8K?

It's funny how everyone was pointing and laughing saying both that 300Hz is stupid, and that 8K is stupid... But he said that when they were already 2 years into Morpheus development and Oculus was about to start their kickstarter. 300Hz will eventually be achievable with reprojection and it will be important to lower the latency even further (3.3ms instead of the current 8.3ms for Morpheus), and the 8K means it's 4K per eye with a very wide FoV, making this a reasonable target for an ideal VR technology. In retrospect there's was nothing overkill about his statement.

 

[DuckThor Evil]

It is indeed the easiest comparative calculation, but the comparison isn't qualitative. Take 1080p120 versus 2160p60. Both have exactly the same pps.

Might want to check your math on that 2160p is four times the pixels per frame compared to 1080p.

 

[DieH@rd]

Sony is preparing another minigame [as a part of Playroom VR] that is another crack at VR + 4 player couch coop. VR player is a cat, and 4 coop players are mices that are trying to snatch as much cheese as possible.
https://www.jp.playstation.com/blog...tkgpscom=dc_tgs2015_playroomvr_ps_tw_20150918

Did we talk about this tech of Sony for showcasing totally different viewpoint for couch players? How are they doing that? Brute-forcing rendering of both viewpoints on a single console [60fps VR render interlaced with 60fps TV render into 120fps stream] with PSVR splitter handling the conversions of video streams, or via 2 consoles [essentially a MP match with asymmetric controls]?

They said the V2 OLED is a custom pixel arrangement with full RGB per pixel.

I would guess both Vive and Oclulus have already switch to full RGB too, or will do so at release. Previous prototypes were pentile because they would need a custom screen for full RGB. All existing smartphones screens are pentile at these resolutions.

Thanks for the clarification.