What? Weitek all over again?
Providing the interface for the massively
parallel super-computer guys to do something is one thing; bringing it
mainstream is probably impossible IMO. This is the same old chicken/egg
situation: software guys don't want to do it because there are no
customers; customers don't want to do it because there's no software... and
the hardware is even more expensive now.
Is AMD starting to howing signs of hubris?
Correct me if I'm wrong, but AMD is basically inviting folks like
Weitek v2 to come and join the party. I don't think AMD wants to
produce coprocessors, but that they want to figure out a way to make
coprocessors work with K8 and Hypertransport. Eventually, successful
coprocessor teams would be purchased (or AMD would provide the same
functionality on-die), and the unsuccesful ones...well, at least AMD
didn't waste money : )
It's a smart move...the only question is whether VCs are dumb enough to
fund the same idea again. You'd hope that they learned from Weitek...
Also, I don't think coprocessors are really attractive for the desktop
market really. I could see it for servers (think crypto cards). I
could almost see it for laptops, but it would be a really hard sell.
Ultimately, I'm not sure why the hell you'd want an ASIC though. Is it
faster than an FPGA? Sure. But with uncertain volumes, you're almost
certainly better off with an FPGA...
DK