Replies to post #14393 on Intel Corp (INTC)

[tecate]

Hi Keith,
Marketshare in what? Thanks. :)

[chipguy]

From what I have seen disclosed so far about Cell it isn't
apparent to me how it is a material improvement on countless
previous attempts to harness many processing elements on
a chip. For example, the Texas Instruments 320C80 (4 DSPs
+ RISC + shared memory + crossbar switch + sophisticated
DMA/ memory controller) the architecture was *very* difficult
to program and most apps never came close to theoretical
peak performance. The second issue is problems with big
performance requirements have massive datasets and that
means a bottleneck getting data on and off chip. I remain
very sceptical about Cell achieving success outside of video
game consoles and perhaps special purpose computers.

[wbmw]

Re: thoughts on Cell? IBM seems to think they can take 15% of marketshare from INTEL with that one

There are very few PC applications that can take advantage of a processor like Cell. Most apps are single threaded, and it's hard enough to create additional threads for multithreaded or multicore processors. Cell designs are supposed to implement hundreds of processing elements, and even a well trained emulator won't help it to perform well under Windows, or even MacOS if IBM goes that route.

I think IBM's 15% MSS figure is complete crap. They will not get 1%, or even 0.1%, IMO.