AMD talks up graphics Fusion

AMD STARTED TALKING tech about Fusion today, and while there was nothing directly stated, several things could be read between the lines. With that against a backdrop of the industry news, you can figure out a lot of things.

First, the stand-alone GPU, and presumably the stand-alone ATI GPU is far from dead. At first the integrated GPU will be aimed at mobile parts for power savings, it appears to be connected across the internal crossbar, a vastly higher speed interconnect than PCIe. It is also a vastly lower power interconnect than PCIe, which is why mobile is the focus.

On top of that, mobile is far less demanding than a pimped-out high-end GPU. This is the main way that DAAMIT will address the memory bandwidth disparity by lopping huge amounts of bandwidth use out of the equation, low end embedded parts need low end embedded bandwidth. These parts are, duh, better to embed.

The diagrams shown had the GPU with a separate cache, but may not make it into the real world. Intel internally seems to think this is where they will get a big bang out of the integration of graphics in Gesher, cache, cache and more cache.

Also added to the mix is AMD's process technology. ATI can go from a process made for a generalist outsourced fab to a more tightly tuned product. This can add up to big gains for the graphics side, basically a free process node gain.

So, all people seem to think a CPUGPU (pronounced 'ralph' with an eastern Mongolian accent) is a good idea. AMD seems to think this will happen in the late 2008 or early 2009 time frame, so it will be up against Nehalem on the Intel side, and then have a bit of breathing room before Gesher.

The oddest thing is that they kept repeating that this is just one version of the technology, pulling a distinct GPU onto a distinct CPU, and keeping the parts separate until the crossbar. There is at least one implementation that takes a different approach and will be coming out in the same time frame, so what AMD is talking about is far from the only way to do things.

In any case, it looks like late 2008 is going to be a very interesting time for tech, it will be the next big inflection point. Three competing implementations of the same concept from two companies, may the best chip win. µ