Sony PlayStation VR2 (PSVR2)

Yeah I agree it's a weird comparison given they're both designed to address the different use cases of their target markets.

The fact that Quest Pro is lower resolution and heavier is for example meaningless given the screen needs to be powered from the internal SoC and it has a battery pack.

For me the interesting (and exciting) things from that article were confirmation of the on-handset cameras, the expression tracking, and the eye tracking with foveated rendering. Those features bode well for the next gaming orientated headset. Do we know if they'll be producing a non-standalone version moving forwards?
 
Yeah I agree it's a weird comparison given they're both designed to address the different use cases of their target markets.

The fact that Quest Pro is lower resolution and heavier is for example meaningless given the screen needs to be powered from the internal SoC and it has a battery pack.

For me the interesting (and exciting) things from that article were confirmation of the on-handset cameras, the expression tracking, and the eye tracking with foveated rendering. Those features bode well for the next gaming orientated headset. Do we know if they'll be producing a non-standalone version moving forwards?
They are not going to produce any more non stand alone versions, according to what they said when launching Quest 2.
 
few years ago there were meta sketches of a standalone VR headset paired with a processing box, so the headset would be lighter, and at the same time offer more processing power, wonder where that went.
 
Yeah I agree it's a weird comparison given they're both designed to address the different use cases of their target markets.

The fact that Quest Pro is lower resolution and heavier is for example meaningless given the screen needs to be powered from the internal SoC and it has a battery pack.

For me the interesting (and exciting) things from that article were confirmation of the on-handset cameras, the expression tracking, and the eye tracking with foveated rendering. Those features bode well for the next gaming orientated headset. Do we know if they'll be producing a non-standalone version moving forwards?

The question WRT gaming is which of these new features will make it to a commercially priced Quest 3 which presumably comes out in 2023 sometime? The controllers can already be used with Quest 2 (300 USD for the controllers), but that's way more than an entry level VR device could support. Unless, of course, it's actually not that expensive to manufacture and it's that much because it's targeted at a traditionally high margin market (business and professional).

I'd imagine perhaps features that closely map to what PSVR2 supports while likely dropping many features that the PSVR2 doesn't support (to reduce the price). That would make it easier for developers to port PSVR2 games to the Oculus, er facebook, er... Meta storefront. :p

Regards,
SB
 
560g (devkit was 630g)

If somebody stuggle with setup
pnnxlH6.png
 
Last edited:
RE 8 Village is 4k 120 fps without RT on PSVR 2 for foveated rendering they use a functionnality of the GPU call Flexible scale rasterization.

from a japanese interview translated with DeepL



How advanced are the graphics of PSVR2?

 In conclusion, the PSVR2's visuals are quite beautiful. I would say that it is on a par with the binocular 4K-class VR HMDs that I have experienced.
 First of all, the resolution, even if it is an exaggeration to say that it is at the level where dots cannot be seen, it is at the level of a 24-inch display with full HD resolution. The resolution is 4K for both eyes and 2K for one eye, so it is understandable.

In addition, I observed that the furniture in the old castle looked unusually realistic in terms of texture, so I moved closer to them and closed one eye alternately to observe how the highlights of the specular reflection appeared. I found that not only were the 3D objects seen by the left and right eyes rendered separately to express the parallax of the 3D shapes, but also the lighting and shading were done separately for the left and right eyes.
 Therefore, when comparing metal vessels and reliefs with one eye closed, one can see specular reflections that take anisotropy into account. To put it more concretely, the highlights of the object we are looking at are different when we look at it with the right eye than when we look at it with the left eye. Needless to say, the same is true in the real world. This is one of the reasons why the texture of the scenes in the old castle in Resident Evil: The Village is so realistic.

When the author asked, "If you go this far, will you be able to get a stable 120 fps in the main game?" He answered emphatically, "We are doing our best by tuning the graphics elements of the original version. He answered emphatically, "We are doing our best by tuning the graphics elements of the original version.

 This makes me wonder how the graphics elements of the original version were tuned, and from what I have heard, there seem to be several key points.
 First, Resident Evil Village does not use the PS5's real-time ray tracing. The PS5's ray tracing capabilities are limited, and even in the original version of the game, the rays are spatially and temporally distributed. Therefore, rays were not fired from every frame and every pixel, and the source of firing was swapped in time or the direction of firing was changed. When rendering, the information acquired by a certain number of rays fired in the past frames is collected, retrieved, and rendered, but when the completed frame generated is viewed simultaneously by the left and right eyes, it appears flickering and noisy. They did not use ray tracing in the PSVR2 version because they did not like this.

 In addition, the pseudo-shadows created by Ambient Occlusion are also said to flicker noticeably when viewed in stereoscopic VR, and the PSVR2 version of Resident Evil Village is being developed with post-effects expressions removed or adjusted. This reminds me of the efforts made in the past to develop games that support 3D stereoscopic viewing.

 The second is tuning the rendering system to optimize it for VR, which is a combination of two technologies.
 The first is the use of "Flexible Scale Rasterization" (Flexible SR), a special rasterizer feature of the PS5 GPU.

 A rasterizer is one of the functional blocks in any GPU that is responsible for breaking down polygons into pixels. On PS5 GPUs, the rasterizer can intentionally break polygons into an unbalanced balance of pixels.

Since the PSVR2 eyepiece has the highest optical resolution near the center, and the resolution decreases as you move outward, SIE encourages PSVR2 titles to take advantage of Flexible SR to rasterize at a lower pixel resolution as you move outward in the video frame. SIE encourages the use of Flexible SR too.



The third is Foveated Rendering, which utilizes eye tracking technology.
 PS VR2 can track the wearer's line of sight in real time, which can be used to create lighting and shading that is not uniform in resolution. This is the basic concept of Foveated Rendering (related article).

 Foveated Rendering is generally implemented by utilizing Variable Rate Shading (VRS), a non-uniform resolution shading technology supported by most current-generation GPUs. However, since the PS5 GPU does not support VRS, a different approach is required.
 Specifically, we use uneven resolution rasterization with Flexible SR to aggressively reduce pixel resolution in regions outside the line of sight. This allows us to achieve rendering similar to that achieved when using VRS.

 In summary, the PS VR2 GPU achieves GPU performance by controlling the non-uniformity of pixel resolution using Flexible SR in response to the user's line of sight. Thanks to these innovations, the PS VR2 can stably display high frame rates of 90 to 120 fps even at 4K resolution.

 In Resident Evil Village, we placed a complex chandelier in the center of our field of view and observed how the chandelier looked different when we moved our eyes to look at the periphery of the screen. For example, when the viewer gazes at the chandelier, the gaps between the many candelabras are clearly visible, but when the viewer shifts his/her gaze from the chandelier, the gaps disappear and the image is blurred.

 The binocular 4K image and 120 fps of PSVR2 were achieved through the accumulation of such diligent efforts. Hats off to the ingenuity of the development team.
 
Is Flexbile RS standard in RDNA2? It's a better solution for foveated rendering than VRS as you are reducing every aspect including pixel count, whereas VRS just reduces shader quality (although you could match VRS with other resolution reductions etc via other techniques). This is the first I've heard of it.

Googlage throws up these


Seems Sony included PSVR focussed hardware tweaks, demonstrating a real commitment to VR.
 
Is Flexbile RS standard in RDNA2? It's a better solution for foveated rendering than VRS as you are reducing every aspect including pixel count, whereas VRS just reduces shader quality (although you could match VRS with other resolution reductions etc via other techniques). This is the first I've heard of it.

Googlage throws up these


Seems Sony included PSVR focussed hardware tweaks, demonstrating a real commitment to VR.

No it isn't a standard RDNA 2 feature, something Sony asked to AMD like the cache scrubber.
 
But is this hw feature and not software way of beneficial use of eye tracking ?

It is an HW feature like the cache scrubber or the ID buffer. Many people misunderstood the cache scrubber it does not give to PS5 better performance but better memory management. It means you don't need like on on PS4 or all other hardware to separate in two memory pool GPU data in memory and streamed data. For example on KZ Shadow Fall GG gave in a GDC presentation, the RAM allocation and it is separated.

This is funny in HFW deffered texturing they explained they did not use the ID buffer for creating the visibility buffer because they need to do a PS4 version. The hardware is there on PS4 Pro and PS5.
 
It is an HW feature like the cache scrubber or the ID buffer.
How do you know this ? I briefly checked this patents and hard to say, only found things like:
in https://patent.nweon.com/19075
[0034] Hence in summary FSR is a feature (software or preferably hardware) which allows the resolution of the rendering to be adjusted across render buffers dynamically so that areas requiring more resolution can be rendered in more detail. As noted above, an example application of this is foveated rendering, which concentrates high resolution rendering to the area of the image corresponding to the fovea (e.g. the centre) of the viewer, matching the human optical system which has a higher density of photoreceptors in the foveal region. Gaze tracked foveated rendering systems utilise this further by moving the high resolution area to track the part of the image that the eye is looking at.
[0064] It will be appreciated that the above methods may be carried out on conventional hardware suitably adapted as applicable by software instruction or by the inclusion or substitution of dedicated hardware.
[0076] Accordingly, in a summary embodiment of the present description, an entertainment device (10), such as a Sony PlayStation 4 .RTM. or 5 .RTM., may comprise virtual model generation circuitry (such as for example CPU 20A and/or GPU 20B) adapted (for example by suitable software instruction) to generate a view of a virtual model having a locus of interest; flexible scale rasterization circuitry circuitry (such as for example CPU 20A and/or GPU 20B) adapted (for example by suitable software instruction) to perform flexible scale rasterisation with a first bin distribution for a luminance channel, and to map the results to a first pixel space;

or in https://patents.google.com/patent/US9710881B2/en
Metadata associated with the subsection a primitive overlaps determines a pixel resolution for the subsection. The metadata is used in processing pixels for the subsection to generate final pixel values for the viewport of the scene that is displayed on the display device in such a way that parts of the scene in two different subsections have different pixel resolution.
 
How do you know this ? I briefly checked this patents and hard to say, only found things like:
in https://patent.nweon.com/19075




or in https://patents.google.com/patent/US9710881B2/en

Because in the Japanese article about RE8 article they said it is a feature of the PS5 GPU hardware rasterizer.

 The first is the use of the PS5 GPU's rasterizer special function " Flexible Scale Rasterization " (Flexible SR).

 A rasterizer is one of the functional blocks installed in any GPU, and it plays the role of decomposing polygons into pixels. And with the PS5's GPU, the rasterizer can decompose polygons into pixels with an intentionally imbalanced balance.

And this is not new, PS4 Pro had it too but not as advanced but it seems an improved version in PS5.


Since the Pro - PS has built-in support for multi-resolution render-targets. The real thing - none of the 'splitting/culling views' shenanigans involved. If patents are to be believed, PS5 could even have improved versions of that.
 
Is Flexbile RS standard in RDNA2? It's a better solution for foveated rendering than VRS as you are reducing every aspect including pixel count, whereas VRS just reduces shader quality (although you could match VRS with other resolution reductions etc via other techniques). This is the first I've heard of it.

Googlage throws up these


Seems Sony included PSVR focussed hardware tweaks, demonstrating a real commitment to VR.
Yup, I still firmly believe that the PS5 was developed with VR heavily in mind, with features to increase immersion trickling down to non VR like the haptics, force feedback triggers, 3D Audio etc. It's just so much to increase immersion and most of them only achieve their full potential in a VR environment.
 
Back
Top