RC Cola
02-12-2005, 07:04 PM
Lighting Up the Cell
By Ed Sperling -- Electronic News, 2/11/2005
IBM’s collaboration with Sony and Toshiba to create the Cell processor -- being billed as a supercomputer on a chip -- has resulted in an interesting technology. It even gets the gee-whiz award for the year.
Nine cores at 4.5GHz or more? Heck, that makes Intel’s forthcoming dual-core chip look like a turbocharged Yugo. Assuming that the doors would stay on in the first place, this theoretically would blow them off. But after years of looking at interesting next-generation technologies, it’s clear that interesting and gee-whiz don’t necessarily translate into successful products.
From an outsider’s perspective, this new chip appears to be fit somewhere between IBM’s old microkernel concept and a massively parallel computing model. Despite plenty of time to prove themselves, not to mention enormous marketing muscle, neither gained significant market share.
Going back into the mid-90s, IBM and others began touting a microkernel that could run multiple operating systems on a single chip -- namely a PowerPC. While IBM touted it privately as a way to minimize Microsoft -- a classic example of IBM’s surround and conquer strategy that would limit Microsoft to competing as one of many operating systems instead of the operating system -- to the best of anyone’s knowledge it never even saw the light of day.
Incidentally, the PowerPC itself was supposed to crush Intel’s grip on the PC market. Power -- initially an acronym that stood for Performance Optimized With Enhanced RISC, something most IBMers can’t even remember -- has had a relatively successful career inside IBM and Apple computers. But it has done little to stem the growth of Intel, which remains the 800-pound gorilla in the processor world. And it hasn’t displaced Microsoft, which continues to dominate the desktop and increasingly the embedded market.
Parallel computing, meanwhile, has proved to be a wonderful technology for very specific purposes. It has proven to be a tremendous time saver for identical tasks, such as search, which is what has made it popular for large databases and processor farms in the bioinformatics world. It has proven totally unworkable, however, in mainstream computing.
Enter the cell processor. The initial version comes with nine cells, each of which can support multiple operating systems. And while each of these cells may be able to perform multiple tasks, the real limit isn’t in the hardware. It’s in the software that will be needed to get the most out of this technology.
So far, no one has developed an operating system that can take advantage of massively parallel computing, which has been around for at least a couple of decades. Even dual-processor systems are rare because of the complexity of getting both chips to work in sync. And multiple cores all doing independent tasks are something we can only marvel at, because in the short term there’s probably not much chance of effectively using them.
The fact that the three developers are aiming this technology at the home entertainment market is intriguing. And together, IBM, Sony and Toshiba have deep pockets for R&D, not to mention worldwide brand recognition among consumers. But getting this to work across a broad suite of applications will remain a monumental challenge.
While we may wish for the fastest transport -- and we may get it in the gaming market where similar tasks are part of an artificial world’s program -- the rest of us mere mortals may be stuck driving Yugos for a long time to come.
Source (http://www.reed-electronics.com/electronicnews/article/CA503388.html)
Pic of Cell:
http://images.cxotoday.com/cxoimages/storyimages/matter2761.jpg
I'm curious to see what everyone here thinks of Cell. I've pretty much have given my thoughts about it already, thinking it should do well in the PS3 (and PS4+ if Sony sticks with it, which I assume they will), other multimedia devices, workstations and servers. However, unless some serious work is done, I doubt it will make much of a dent in the Intel/AMD PC world, at least not for a while.
It should also be able to take the #1 spot in supercomputers if they hook up enough of them. Not sure what they'd use it for though, except for saying that they have the most powerful network of processors in the world.
For a more in-depth and technical analysis of the Cell presentation earlier this week, you can read a report here (http://realworldtech.com/page.cfm?ArticleID=RWT021005084318&p=1).
By Ed Sperling -- Electronic News, 2/11/2005
IBM’s collaboration with Sony and Toshiba to create the Cell processor -- being billed as a supercomputer on a chip -- has resulted in an interesting technology. It even gets the gee-whiz award for the year.
Nine cores at 4.5GHz or more? Heck, that makes Intel’s forthcoming dual-core chip look like a turbocharged Yugo. Assuming that the doors would stay on in the first place, this theoretically would blow them off. But after years of looking at interesting next-generation technologies, it’s clear that interesting and gee-whiz don’t necessarily translate into successful products.
From an outsider’s perspective, this new chip appears to be fit somewhere between IBM’s old microkernel concept and a massively parallel computing model. Despite plenty of time to prove themselves, not to mention enormous marketing muscle, neither gained significant market share.
Going back into the mid-90s, IBM and others began touting a microkernel that could run multiple operating systems on a single chip -- namely a PowerPC. While IBM touted it privately as a way to minimize Microsoft -- a classic example of IBM’s surround and conquer strategy that would limit Microsoft to competing as one of many operating systems instead of the operating system -- to the best of anyone’s knowledge it never even saw the light of day.
Incidentally, the PowerPC itself was supposed to crush Intel’s grip on the PC market. Power -- initially an acronym that stood for Performance Optimized With Enhanced RISC, something most IBMers can’t even remember -- has had a relatively successful career inside IBM and Apple computers. But it has done little to stem the growth of Intel, which remains the 800-pound gorilla in the processor world. And it hasn’t displaced Microsoft, which continues to dominate the desktop and increasingly the embedded market.
Parallel computing, meanwhile, has proved to be a wonderful technology for very specific purposes. It has proven to be a tremendous time saver for identical tasks, such as search, which is what has made it popular for large databases and processor farms in the bioinformatics world. It has proven totally unworkable, however, in mainstream computing.
Enter the cell processor. The initial version comes with nine cells, each of which can support multiple operating systems. And while each of these cells may be able to perform multiple tasks, the real limit isn’t in the hardware. It’s in the software that will be needed to get the most out of this technology.
So far, no one has developed an operating system that can take advantage of massively parallel computing, which has been around for at least a couple of decades. Even dual-processor systems are rare because of the complexity of getting both chips to work in sync. And multiple cores all doing independent tasks are something we can only marvel at, because in the short term there’s probably not much chance of effectively using them.
The fact that the three developers are aiming this technology at the home entertainment market is intriguing. And together, IBM, Sony and Toshiba have deep pockets for R&D, not to mention worldwide brand recognition among consumers. But getting this to work across a broad suite of applications will remain a monumental challenge.
While we may wish for the fastest transport -- and we may get it in the gaming market where similar tasks are part of an artificial world’s program -- the rest of us mere mortals may be stuck driving Yugos for a long time to come.
Source (http://www.reed-electronics.com/electronicnews/article/CA503388.html)
Pic of Cell:
http://images.cxotoday.com/cxoimages/storyimages/matter2761.jpg
I'm curious to see what everyone here thinks of Cell. I've pretty much have given my thoughts about it already, thinking it should do well in the PS3 (and PS4+ if Sony sticks with it, which I assume they will), other multimedia devices, workstations and servers. However, unless some serious work is done, I doubt it will make much of a dent in the Intel/AMD PC world, at least not for a while.
It should also be able to take the #1 spot in supercomputers if they hook up enough of them. Not sure what they'd use it for though, except for saying that they have the most powerful network of processors in the world.
For a more in-depth and technical analysis of the Cell presentation earlier this week, you can read a report here (http://realworldtech.com/page.cfm?ArticleID=RWT021005084318&p=1).