We had a really cool presentation on GPGPU from
Mike Houston, one of the scientists at Stanford, when we did the X1800 launch event...
...Rys was enraptured
(I beileve that is the phrase !)
One point Mike made was that this kind of movement
(CPU to GPU) is getting easier over time and that this has opened up the awesome parallel processing power of modern GPUs to a whole new audience
You sound like an expert
(in a field where it is really easy for someone like me to get in over their heads)... so I am sure your views are spot on
The point I was highlighting
(which you have confirmed with your dual Xeon/Opteron example) is that the level of power contained in a 'super computer' from a given generation, is rapidly reduced to 'desktop level' by virtue of the industry that we are in...
...and this 'shrinking' seems to take less and less time with each new wave of technologies
It is an interesting exercise to try and imagine just how powerful things will be in 10 years time
Maybe this question could form the basis of a competition
