AMD Render Fusion – moving from the meta information paradigm to a pixel information paradigm?

In his blog, Simon Solotko, talks about how Render Fusion could change the way the Cloud and the clients  interact. His vision is very futuristic, and it contains a little “Blade Runner” like thoughts, but I think that we need this kind of vision.  As a summary of his vision I have selected the following part of his text.

“Rendering in the Cloud can solve the bandwidth problem by capping the bandwidth problem. All I need to do is refresh a screen with data at a particular resolution appropriate to the client’s screen and available bandwidth. No additional bandwidth needed between the client and the server, ever – only three variables, screen size, interface visual integrity, and upstream user input. Peak bandwidth requirements become fixed and predictable.”

In my opinion, it looks like AMD wants to move from a meta (hypertext) information paradigm, to a pixel information paradigm. With this paradigm changing they are promising  bandwidth utilization reduction, and they also want to reduce the applications complexity – eliminating supposed complex synchronization algorithms.  Although the idea of transforming the relation between a computer and the Cloud into something like the relation between a monitor and a graphic board could be interesting, I think that with it we will break, in a radical way, with all the work that has been done in the last 20 years of the WWW.

The idea of using network quality parameters together with parameters of the Cloud client, such as  its screen size or visual integrity, could, in fact, reduce same bandwidth utilization. But this is something that modern browsers and service providers are already doing. As an example, when we are connecting to the Gmail, using a mobile device, we are dynamically presented with an interface (hypertext page) that is different from the one obtained when we are connecting using a desktop web browser.  The selection is dynamically done by the service provider in order to obtain the most suitable for the specific client/device. Another example is the way web 2.0 applications are refreshing the screen  in specific selected areas,  using AJAX or other asynchronous transfer mechanism.

Another question finds its importance, and that is the so called  latency problem. I’ve questioned  Simon Soloko about this and he had the kindness to summarize his talk with Jules Urbach about the Render Fusion:

“Two latency issues for gaming (again, probably the most demanding use case), upstream and down. For upstream input flow, the critical issue is the ability of a user to effectively aim and respond to extremely fast paced visual cues. I need to look at my crosshair, see a target within, press a button, and have the server know that I pressed in the instant that the target was in my sights. Today’s online games already suffer upstream latency, and play at game-time rates which are faster than real time life experiences – as in Unreal Tournament 3 and other amazingly detailed, fast paced, aim-critical games.

The difference is that that last steps within a single cycle of the game flow, which is transforming game-world data from the server to produce visual effects and render the scene. These steps happen on the client today, and the proposal is that they could occur on the Render Cloud, be compressed to video, and decompressed by the client. The question is how much lag does rendering in the cloud vs. rendering on the client add? Input latency is an issue but it is not the critical issue – it is the time it takes to deliver the scene. On high bandwidth, low resolution clients (laptops and wifi smartphones) the problem seems much smaller, making those, in my view, lower hanging fruit. Displacing traditional PC gaming will require an experience with established client rendering and very potent client GPU’s and increasingly affordable prices.

Displacing traditional PC gaming will require an experience competitive with established client rendering – very potent client GPU’s at increasingly affordable prices.”

Thanks Simon for this summary and also for your work on the Render Fusion presentation. It’s always important to have this kind of refreshing thoughts. Although I think that a lot of work is still to be done by AMD, the path that they are following could lead to interesting technical solutions.

Share/Save