CPU/Mobo IBM Japan to sell Cell procesors to PCs ?

IMO jus another way for IBM to milk some more money..

come on what are they thinking..cell architecture is very diff from reg intel,AMD x86 arch..infact its based on PPC arch ..which is not been used any more (MAC's use Intel procys now a days)

So i guess Cell will be restricted to ppl doing research and who have time to sit and write code for the cell.

although cell will be able to do 250 gigaflops (theoretical peak ) it will be dwarfed by the GPU's which are able to hit a teraflop.

also now a days traditional micro processors using FPGA (field programmable gate array ) are able to achieve nearly a petaflop , using less than a 100 nodes.

so jus another market gimmick.
 
i wouldnt say cell is all that bad.when it comes to multimedia and graphics processing it still is the best:hap2: and it does come with XDR memory which runs at 3.2ghz:cool2: which means zero latency .

look at the case of graphics in PS3 they have barely begun to scratch the surface of what cell can do .
even heard cell can actually emulate a graphics card in software :S
 
vijaysurendran said:
i wouldnt say cell is all that bad.when it comes to multimedia and graphics processing it still is the best:hap2: and it does come with XDR memory which runs at 3.2ghz:cool2: which means zero latency .
look at the case of graphics in PS3 they have barely begun to scratch the surface of what cell can do .
even heard cell can actually emulate a graphics card in software :S

no one is saying that the cell is bad ..but its jus wont succeed from the value for money perspective(for normal users) unless they sell it dirt cheap.
Aas it wont be compatible with windows..which means loosing 90% of the market..cell only supports linux.

second abt the latency ..XDR chips are uber fast and have huge bandwidths...but the cost will be similar..out of reach for most. :(
wht abt the availability?

also cell needs some thing like a XDR to run as it have very small L1 and L2 cache..which are 8kb and 256kb respectively..which means a lot of swapping..so if the rams not fast it will slow down the procy.

also been a inorder procy has its disadvantages in terms of developing software for the devs.

also all the SPE's of the cell can do only single precision floating points where as the GPU's and the CPU's do double precision floating point.

So doing a direct flop comparison is not fare.
 
@arnold, i don't think they intend to sell it for "normal" users. Just because its on sale does not mean you have to buy it.

It will be useful for people who are into rendering farms, HPC etc. And for such specialized applications, they will have to write their own optimized code, which will definitely work far better than any other generic application. A good move by IBM IMHO
 
Checksum said:
@arnold, i don't think they intend to sell it for "normal" users. Just because its on sale does not mean you have to buy it.

It will be useful for people who are into rendering farms, HPC etc. And for such specialized applications, they will have to write their own optimized code, which will definitely work far better than any other generic application. A good move by IBM IMHO

that is exactly what i said in my first post. :)
 
i never said Fixstar(the PC manufacturer ) is gonna bring the pc to the masses Yes its for financial products' risk assessments, weather forecasts, disease-imaging diagnostics and other applications.one more question if IBM is so good at helping companies like sony toshiba and amd realise their chip dreams why wont they bring in their own brand of microchips .a lil more competition for amd and intel wud not hurt!!
 
BTW jus for stats..in yr 2000 only 5 supercomputers in the top 500 were powered by intel chips and the rest were all IBM.

but as of march 2008 ..75% of the top 500 supercomputers are Intel based , although the top 5 is still dominated by IBM.

cant find the source though right now..read the article on [H] abt a month ago. [:)]
 
Back
Top