Fixed Foveated Rendering in new AMD 6000

Fixed Foveated Rendering not working with radeon 5000 series but AMD claims their new 6000 GPU will have this option working. So my question is - do we need pitool update and pimax need to implement new FFR for amd or will current FFR work with any card supporting it?

I wanna get 6900xt but I don’t want to wait a year for new software. I wanna use FFR same day when I get 6900xt.

3 Likes

fairly sure the version ffr that pimax has implemented atm is nvidias rtx version not some standard dx12 version. so yeah, unless someone can tell you otherwise you’ll probably have to wait for a pitool update.

7 Likes

@Alex.liu might be able to answer that?

3 Likes

Well, personally, I think the question is whether Pimax is planning to do something about that in the near future. I still remember 5700 XT being unsupported for months on end. I am going to get 6800 XT ASAP, so I am interested in the general support and FFR support also.

1 Like

yeah but OP was asking if it would be supported out of the box:

Like 1% chance of that happening. Pimax aren’t that big of a company to get hardware ahead of consumers, plus it does take time to implement and debug.

which basically was my point :wink:

Near future is a not well defined abstract in Pimax space :wink:.

7 Likes

It’s kind of like “Elon Time”

3 Likes

Just to be sure, FFR will work with 3080 out of box?


(xkcd)

… more importantly, Smart Smoothing?

1 Like

Yes.

Still waiting for a link to a source showing AMD claimed any of these VR related features will be working with any of their 6000 series GPUs.

It has been confirmed by AMD that they support variable rate a year ago.
https://www.amd.com/en/technologies/radeon-software-fidelityfx

Pimax hasn’t confirmed that their implementation works with AMD. In fact, they’re straight out ignoring every question in that regard. Most likely because they don’t support it and are not planning to support it, or are only now starting with the implementation.

Smart smoothing has always worked on AMD.

3 Likes

AMD is only claiming to support some sort of variable rate shading, which does NOT necessarily mean the same performance gains, limitations, or compatibility. I wouldn’t count on decent foveated rendering just from that.

If AMD support is based on a different API, then ‘only now starting with the implementation’ is entirely reasonable. I have my doubts anyway.

Radeon VII? I had that card, and I don’t think so.

1 Like

AMD is pretty specific on what they are supporting.
They support DX12 variable rate shading: Variable Rate Shading | DirectX-Specs

From a performance perspective, the performance gains are pretty much given. The fragment shader gets executed only once, per pixel group, instead of per pixel. Fragment shaders are usually by far the most compute intensive stage in the graphics pipeline. Depending on the code in the shader.

Compatibility, etc. are what is unclear, but AMD is most likely willing to answer those question and help with the implementation, when a developer asks them. And if they are not, then Pimax could just state that.

It’s common to prepare software well in advance. When the new consoles launch, there will be many games on release, and not only 2 or 3 years later, as if they only started developing on release day.
There’s a good chance that other HMD companies have already been working with AMD on that for a while.

I’m surprised to hear that smart smoothing is not working on your Radeon VII. I see no reason for that being the case, unless the code deliberately excludes the GPU. It works fine on my RX 580, which is an older GPU.

1 Like

Unless of course the card again does not have enough H.264 motion vector compute units, in which case, perhaps not enough of the pixels could be excluded. Foveated Rendering, due to most of the area of a circle being towards the outside edge, can exclude a majority of the pixels on-screen, and supersampling pushes these pixel counts astronomically high. It is quite likely the latest AMD card, being more dedicated to number crunching than the NVIDIA offering, simply will not have enough dedicated silicon to do this in a single clock cycle.

Logically, no. If it takes multiple GPU core clock cycles with the AMD card, frame latencies/rates could be worse.

No, it is not, and much less so for smaller companies.

The Vega (and IIRC even RX 580) units were known to have more H.264 motion vector compute units, and under some conditions, could provide VR frame interpolation, where the Radeon VII could not.

I remain highly skeptical whether AMD has given these VR features some priority, given their latest card optimizes for raw number crunching.

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.