I do genetic algorithm research on CUDA, and I'm waiting anxiously for CUDA 3.0 on Fermi. Unlike most, I'm not too bothered (yet) about their massive DP performance increase, rather the simultaneous different kernel executions and increased caching/memory behaviour.
I was seriously looking at Larrabee, and hoping it would resolve many of the headaches suffered by CUDA programming. Unfortunately Intel have pulled it for 2010 at least. Larrabee on 22nm might be magnificent, but what will the gfx manufacturer's have on the same scale? I worked for Intel years ago, and they are very cost-conscious and I wonder if they're not reluctant to invest massively outside of their primary realm? Will they just keep their investment in the on-board graphics market instead?
ATI's Close to the Metal has nothing going for it realistically, so I'm drooling in anticipation of Fermi. Now, this has nothing to do with gaming, so I can understand most people getting disappointed with NV's nonsense - but for anyone in scientific research there's nothing to compare.
My only concern at the moment is Time, waiting for a decent refresh of Fermi, and not jumping on their first release, which by all accounts looks rather untrustworthy. Larrabee was scheduled for H1'10, so I suppose I can wait 6 months - I really wish ATI/AMD could get their GPGPU-act organised.
I used to weary of Charlie' Ds rage against the green-machine, but I've added him to favourites lately - whatever about his personal opinions, he does tend towards semi-accuracy most of the time.