Much has been written about the next-generation cell processor, a collaborative effort among IBM, Toshiba and Sony. It is an intriguing technology, already being dubbed as a “supercomputer on a chip,” but IT executives should not be making plans for using it in their enterprises any time soon.
The chip landscape has been littered with innovations that did not work out for any number of reasons: lack of demand, lack of tools support and lack of applications. The last processor to receive this level of hype and speculation was Intels Merced, which as the Itanium has achieved decidedly mixed results.
The Cell processor comprises a main PowerPC processor core linked to eight special-purpose “synergistic processing element” cores. If all goes well, the processor will be the building block of powerful distributed computing grids. The design is very different from the multicore processors starting to arrive, which pack more computing power into todays processors without radically changing the computing model.
Based on articles and rumors flying around the Cell processor, the technology will first find its place in consumer markets, from HDTV and gaming systems to high-end supercomputing. Mere speculation on the Cell becoming the next processor for Apple, in a partnership with Sony, was enough for financial analysts to raise their estimates for Apples stock.
If IT managers have learned anything from the Merced/Itanium debacle, it is that they cannot plan for the next generation until promises turn into products. Too often “the next big thing” is superseded by improved old technology that is backward compatible. This is what happened with AMDs Opteron processor, which proved that 64-bit technology can support the 32-bit world, something the 64-bit-only Itanium could not do.
No matter how impressive the Cell processor becomes, all the innovation will be for nothing if developers have a hard time creating applications for it. Cell processor vendors need to be upfront about what the processor can do and offer a realistic road map for when the chip will be ready, avoiding the lofty promises and debilitating delays that the Merced will long be remembered for.
Just as the leap from 32-bit processing to 64 bits has been hard to digest, the Cell architecture will challenge developers. Cell processors will not rule the desktop market in the near future, but the architecture could be used in distributed computing environments.
No matter how much performance the Cell can churn, the key values that will determine whether it is a success will not be transactions-per-second or gigabits-per-second bandwidth; it will be the number of applications written for it and the number of processors sold. Right now all that is sure is that the Cell processor will be powering the PlayStation 3 and Toshiba HDTVs; any prediction of it going beyond the world of fun and games is dangerous.
Were interested in your Opinion. Tell us what you think at [email protected].