You know, as a community, we are pretty good about spouting-off in grand chorus about what we don't want. But in thinking about all the comments I've read over the years, I'm not sure I could determine what we reasonably do want (myself included, though admittedly I only spend infrequent brief, but highly focussed, periods thinking about how graphics cards fit into the scheme of things I want). We generally lament that high-end consumer graphics cards cost $500 or $600 (though the guy forking out $2500 for a Quadro using the same GPU isn't going to be too sympathetic), yet we all know that no matter how good a graphics card you can sell for $250, you could always make a little better, even if you have to charge more, and why stop there... . We don't like the immaturity and awkwardness of multi-card solutions, and when two cards are bonded into a single unit to make it somewhat easier to apply, that is somehow cheating.
Frankly, I think a modular approach to graphics/physics/multistream coprocessing power. Naturally there is huge room for improvement in packaging (multiple modules on a single unit), efficiency (work better together in more situations), etc... , but a little more pressure on these types of configurations would probably be a good thing. On the other hand, maybe we should continue to encourage the graphics companies to wring everything they can from designing end executing the most onto single chips and be content that the modular approaches will come out of the CPU-focussed organizations soon enough.
I am more amateurish than average on this board and don't have much more to add than this overly-simplified, high-level view, but if any of you that have invested more into your understanding of graphics card technology and applications would care to talk about what you would like to see and why, I would be interested in following the discussion.
nBd
Frankly, I think a modular approach to graphics/physics/multistream coprocessing power. Naturally there is huge room for improvement in packaging (multiple modules on a single unit), efficiency (work better together in more situations), etc... , but a little more pressure on these types of configurations would probably be a good thing. On the other hand, maybe we should continue to encourage the graphics companies to wring everything they can from designing end executing the most onto single chips and be content that the modular approaches will come out of the CPU-focussed organizations soon enough.
I am more amateurish than average on this board and don't have much more to add than this overly-simplified, high-level view, but if any of you that have invested more into your understanding of graphics card technology and applications would care to talk about what you would like to see and why, I would be interested in following the discussion.
nBd