The cell is like the laser
Some time ago I read a hack on el Reg or the Inq say that IBM hoodwinked Sony and Toshiba into paying for the development of a high power HPC chip that IBM could then sell at a massive markup in serious compute machines.
I can't imagine why I would care what processor is inside a TV. Toshiba made noise some months ago about putting 4 SPU cores and no PPU core inside a DTR laptop, very expensive. But I can run CUDA code on the GPU in some laptops for much less money... and is a laptop really the proper platform for high power computation?
Now Toshiba heralds the Cell TV. Who cares?
Toshiba should do three things:
1) Commoditize Cell tech (like the ARM is a commodity processor) in which case the end user does not care that it is Cell tech.
2) Offer Cell based HPC compute machines to the market. IBM charges way too much for Cell HPC boxes and Toshiba can make money in that market even while it drives the price down.
3) Offer a standard issue PCIe/SATA personal computer running Linux plus a C SPU SDK using a full blown Cell CPU (not an x86 + a Cell, just a Cell) to the market. The average punter will not buy it, but I bet there are a few hundred thousand scientists who write simulation and analysis code who would step up.
WRT to (2) and (3) I point out that there are many scientists who have used collections of PS3's running Linux to do serious scientific computation. They paid about $400CDN per PS3. There is a demand for this processor at a reasonably affordable price. If PS3 sells at about a $100 loss, then Toshiba could just charge a couple hundred more for a machine that runs Linux native, not on an emulation layer like PS3. I think it would sell well.
Outside PS3 and IBM's expensive boxes the Cell is currently a solution looking for a problem.
Biting the hand that feeds IT
