[citation][nom]webbwbb[/nom]So AMD is essentially trying to replicate the Fermi architecture but with a stronger emphasis on the CPU than the GPU side of things. This is not necessarily a bad thing, it's just good proof that Fermi isn't nearly as bad as most fanboys make it out to be. You can downvote this as you did the other guy who made a similar observation but it is the truth.[/citation]
AMD's GPUs already support OpenCL as well as ATI Stream and can do GPGPU functions just like Fermi. Their downfall is more that Nvidia pushed CUDA very well, and obviously, they can't support CUDA on their cards. While true, an AMD card would not perform in GPGPU functions as well as Fermi, saying they're trying to replicate them is silly considering they already have similar applications, and have had for a while. Nvidia is just ahead in the software game, so here's hoping OpenCL (key word Open) takes off and CUDA dies. Then it won't matter what GPU you have. Fermi just takes GPGPU to a new level (with more hardware to support it), however, the difference between this solution and Fermi is, I believe, that Fusion will *automatically* shunt OpenCL and floating points and whatever is more efficient on the GPU, straight to the GPU, whereas Fermi isn't automatic.