Playstation 5 [PS5] [Release November 12 2020]

Are we sure that RDNA2 has VRS ?
DX12 Ultimate and AMD says RDNA 2.0 does. We essentially asked SONY whether the primitive shader is functionally the same as the mesh shader described by AMD (since primitive shaders are not the same), whether any of the above extensions of sampler feedback are there, and whether or not it supports VRS as AMD has outlined it. Those all require HW components. We wait!
 
I mean, there shouldn't be anything stopping Sony from getting the VRS right? I know MS has patented their own version but at the very least Sony could just use the one from AMD's RDNA 2.0 right?

IIRC Cerny did mention it was custom and that if they didn’t feel it would fit with their strategy/requirement then they wouldn’t have it.
 
DX12 Ultimate and AMD says RDNA 2.0 does. We essentially asked SONY whether the primitive shader is functionally the same as the mesh shader described by AMD (since primitive shaders are not the same), whether any of the above extensions of sampler feedback are there, and whether or not it supports VRS as AMD has outlined it. Those all require HW components. We wait!

https://www.windowscentral.com/amds...ecture-will-fully-support-directx-12-ultimate

There are more links from other sources, but I didn't want to be repetitive.

Thx, I'm a little out of the loop these days.
 
I'd be absolutely gobsmacked if Cerny said 'no' to VRS. It's like, free performance. On a silver plate.

Maybe they started ps5 design before rdna 2 ? In 2019, their was a rumor about a ps5 with jaguar based cpu, but higher speed, and a new gpu, but it was scrapped. OC it was just a rumour .
 
The theoretical specs existed in design only but never made it beyond that to physical chips.
 
I'd be absolutely gobsmacked if Cerny said 'no' to VRS. It's like, free performance. On a silver plate.

Well it's not 'free' as such (or is it!?) maybe they dedicated that part of their budget to something else (cache scrubbers? or something else which he said was exclusive to PS5) - I don't know, just throwing it out there.
 
I'd be absolutely gobsmacked if Cerny said 'no' to VRS. It's like, free performance. On a silver plate.
VRS seems to be more software than hardware. Figuring out which section of the image can get away with less resoluton sounds like a difficult problem. They were doing a simple form of VRS on PSVR and planned it for foveated rendering, so it's a no-brainer they continued in that direction.

I'm curious what physical hardware is required for helping with VRS though.
 
VRS seems to be more software than hardware. Figuring out which section of the image can get away with less resoluton sounds like a difficult problem. They were doing a simple form of VRS on PSVR and planned it for foveated rendering, so it's a no-brainer they continued in that direction.

Isn't figuring out what sections of the images can get away with less resolution a pretty simple problem from a perceptual point of view?. From what I can tell you could easily have some heuristics related to how different a 'block' within the screen is to another based around the colour values, brightness, etc, and when you fall below a given value you start doing VRS on those areas.
 
DX12 Ultimate and AMD says RDNA 2.0 does. We essentially asked SONY whether the primitive shader is functionally the same as the mesh shader described by AMD (since primitive shaders are not the same), whether any of the above extensions of sampler feedback are there, and whether or not it supports VRS as AMD has outlined it. Those all require HW components. We wait!

Did you also happen to ask about the power profile choices given to developers? I'm interested to understand how it limits computational usage for a given frequency choice since power consumption would be application dependent. If I choose the profile which provides me the maximum clock speed, then presumably I could write some code that would exceed that power budget. Is the scheduler throttling or are additional hardware blocks deactivated? All very curious indeed. Not to mention the source of a lot of confusion and misinformation.
 
VRS seems to be more software than hardware. Figuring out which section of the image can get away with less resoluton sounds like a difficult problem. They were doing a simple form of VRS on PSVR and planned it for foveated rendering, so it's a no-brainer they continued in that direction.

I'm curious what physical hardware is required for helping with VRS though.
VRS is slightly tweaked MSAA.

Not entirely sure what allows to feed the change in subsamples shading to hardware.

Pretty sure the generation of the image which controls MSAA/VRS rate is pure software.
 
Listen to this with headphones. A taste of what will be possible with Tempest engine!:

Impressive, abit like this binaural audio. The above clip sounds alot different then real life though... abit too much good maybe, or its my ears, too much fighter jets at close range :p Sound like that in games would be a game changer though, esp in MP games. Hope we see much more on the audio side of things 2020 and later. Would something like that even work in online games though? So much data to transmit?
Also, audio like that probably won't be limited to just a tempest engine ;)
 
I'm not an audio fan and probably I'm missing something, but I clearly remember this kind of demo running on my glorious creative 128.
 
Back
Top