Article on Cell and its growing role at IBM..

[Titanio]

A Fortune article, on cnnmoney.com:

IBM's Quasar: Is it the future of computing?
http://money.cnn.com/2006/03/02/technology/fastforward_fortune/

'Quasar' is the name for IBM's Cell programme. Here's some quotes:

I said then that the new chip, which is especially good for displaying graphics and video and was originally developed by IBM with Sony (Research) and Toshiba for game machines and consumer electronics applications, could become a foundation for IBM's future systems.

Well, now it's happening.

A critical IBM initiative based on Cell called Quasar (which last summer was merely an R&D effort) has "evolved into IBM's systems strategy -- period." That's what a well-informed IBM executive told me this week. Quasar is IBM's main project for next-generation computing. In other words, IBM believes Cell and related technologies will become part of every computer it builds within just a few years. The just-announced blades are the first commercial evidence of Quasar.

To accompany this revolution in technology, IBM is shaking up the company's structure. On March 13, according to the same IBM executives, the company will announce the creation of a new Technology Collaboration Solutions group, headed by Adalio Sanchez, a favorite of CEO Sam Palmisano.

The IBM Microelectronics chip-making business is the centerpiece of the new group. Until recently a problem child within IBM, the chip business has now been profitable for three quarters, say executives, and grew 48 percent last year

Last summer, I spoke with Bijan Divari, who heads the Quasar project. "Five to 10 years from now," he told me, "practically all transactions will be image-based, not text-based. Text is the most painful way of doing transactions." Divari compares Quasar to the legendary development project inside IBM which led to the System 360 mainframe -- and set the stage for the company to completely dominate several decades of mainframe computing.

With his stock still roughly where it was in late 2002, Palmisano has little choice but to shake things up. Cell, Quasar, and the new structure are all part of an ongoing effort to make IBM exciting again -- for both customers and investors.

Interesting, no? If this is to be believed, Cell's role in IBM's general strategy may be growing significantly..

 

[Gholbine]

I believe the PS4 will have a future variant of Cell in it. It seems like a pretty foward-thinking architecture, and it would be good for backwards compatibility with PS3 games.

Anybody else agree?

 

[Guden Oden]

I believe it will be cold outside tomorrow.

Don't everybody just love making self-evident 'predictions'?

 

[ihamoitc2005]

Future

Thank you for this information my friend.

I find this is the most interesting infomation. Maybe this is good compressed data processing.

"Five to 10 years from now," he told me, "practically all transactions will be image-based, not text-based. Text is the most painful way of doing transactions."

 

[Gholbine]

I believe it will be cold outside tomorrow.

Don't everybody just love making self-evident 'predictions'?

Are you talking to me?

 

[Asher]

Quasar exists, but I can state, unequivocally, it is not the "main" project for next-generation computing. To take it a step further and say "IBM believes Cell and related technologies will become part of every computer it builds within just a few years" is a downright lie. I don't know where they came up with this.

IBM has many on-going projects for next-generation computing. Cell, contrary to popular belief on gaming forums, is not a major architecture IBM is looking at. It is one of their architectures, but its focus is narrow and it isn't getting much resources at all compared to the big guns like POWER itself, and some other upcoming radical PowerPC designs.

I'm starting to get increasingly frustrated behind the hype of Cell and its "application everywhere". Cell and its architecture is for use in the PS3, HDTVs, and highly-parallel low-precision computations. It is a niche market, and it will not be part of "everything", it won't even make a considerable portion of IBM's revenues (and even less of its profit). There are similar architectures to Cell in development, with far less of the drawbacks and limitations...but they are not Cell.

 

[Asher]

You also need to have a perspective on how IBM works as an organizational entity. Bijan Divari, who heads Quasar, is tooting his horn to the press about how important Cell/Quasar are because the more important he makes his project look and the more attention it gets, the more money he's likely to get.

There are basically many businesses operating as one name in IBM, and they do compete with eachother. That's part of why the company works as well as it does -- being internally competitive, they call it -- but to take quotes from one of the project/department/divisions and apply that to the big picture of IBM is always misleading.

 

[one]

It is one of their architectures, but its focus is narrow and it isn't getting much resources at all compared to the big guns like POWER itself, and some other upcoming radical PowerPC designs.
(snip)
There are similar architectures to Cell in development, with far less of the drawbacks and limitations...but they are not Cell.

That sounds interesting. Can you give us a bit more insight into what they are like and why IBM invests in those other radical designs along with Cell while use cases can overlap?

 

[Asher]

Because Cell isn't only IBM. Cell has the baggage of Sony and Toshiba.

Cell is not the only solution to highly-parallel processors. Look at Sun's Niagra for another approach. Looking at the trend to such processors, it's not surprising to think that IBM is also developing similar chips on their own, is it?

Why would IBM keep investing in solutions where there are other people sharing in the IP? It's more profitable to go your own way.

Cell was a contractual cooperation project, and some good experience for IBM, but it isn't IBM's "future solution" by any stretch of the imagination.

 

[one]

Why would IBM keep investing in solutions where there are other people sharing in the IP? It's more profitable to go your own way.

Well, without a customer how can you achieve profitability? Without Playstation CELL is a truly niche chip in today's situation. Besides, the more you build software assets for CELL the less likely the probability of the success of other similar attempts get.

As for highly-parallel processors, CELL is not only that, but a CPU for a realtime computer.

 

[Asher]

Well, without a customer how can you achieve profitability? Without Playstation CELL is a truly niche chip in today's situation. Besides, the more you build software assets for CELL the less likely the probability of the success of other similar attempts get.

As for highly-parallel processors, CELL is not only that, but a CPU for a realtime computer.

If you look carefully at what IBM has been doing, its work to supporting Cell and building software assets can actually be largely generalized to getting software to work better on highly parallel systems. That's one reason that Sony/GCC are largely the ones working on the compilation technologically for Cell specifically.

 

[zidane1strife]

Because Cell isn't only IBM. Cell has the baggage of Sony and Toshiba.

Cell is not the only solution to highly-parallel processors. Look at Sun's Niagra for another approach. Looking at the trend to such processors, it's not surprising to think that IBM is also developing similar chips on their own, is it?

We saw a glimpse of their intended direction in xenon, and in their original rumored suggestions for cell, sony/toshiba showed them the light

 

[Megadrive1988]

I believe the PS4 will have a future variant of Cell in it. It seems like a pretty foward-thinking architecture, and it would be good for backwards compatibility with PS3 games.

Anybody else agree?

yes, I agree. I think PS4 will have a massively multi-core next-gen variant of Cell.

looking at graphics chips for a moment, it's been said in the past that PS2's Graphics Synthesizer is like an improved 16-way parallel PS1 graphics chip. whether or not that is true, that's how i see PS4 CPU evolving from the present day Cell.

 

[superguy]

Do they have any actual customers?

 

[Nicked]

Do they have any actual customers?

Yep. Military companies.

 

[Asher]

The biggest non-PS3 customer are medical hardware companies so far. And that's not a very large market...

 

[Crossbar]

Medical hardware companies may not be a big market, but the margins can be pretty good for these kind of devices. I guess the same could be said about the defence industry.

Anyway they are early adopters that will provide some proof of concept for the use primary in image processing devices. It seems like these guys believe it's a rapidly growing market. I don't know.

There have been more than one hint from people at IBM that they are working on a new version of Cell with SPEs having better double precision flotating point performance. Don't you think those will likely be used for number crunching in super-computers and workstations?

 

[one]

IBM at CeBIT 2006: Innovation that matters
http://www-5.ibm.com/de/pressroom/cebit/pressreleases/060303_1en.html

The cell chip: outstanding performance and new design

Outstanding computing performance, large memory bandwidth and unusual design enable the cell chip to set new standards in the development of microprocessors. The cell processor's technology allows much faster processing of data- and computing-intensive applications.

IBM and the Fraunhofer Institute of Technical and Business Mathematics (ITWM) are exhibiting the PV-4D visualization software on a cluster of Cell Blade prototypes. This software, which has won the Fraunhofer research prize, enables the interactive visualization of large quantities of data from scientific calculations and imaging processes in the fields of medical technology and oil production.

Visitors can also view the volume rendering of a human heart in real time. Wearing special glasses, they can see the heart in 3D, turn it interactively and view any cross-sections they like. This solution enables doctors to diagnose illnesses more quickly and accurately and to plan operations better. Cell chips are also used in consumer electronics and multimedia as well as in data- and computing-intensive applications in brain research, bioinformatics, aviation and the processing of seismic data.

The cell chip is the result of a five-year collaboration between Sony, Toshiba and IBM. The Sony Playstation 3, which is to be marketed later this year, will be the first mass product to contain a cell processor.

 

[SPM]

I think you are all missing the main application field for Cell - embedded devices. Supercomputer clusters, high end imaging applications, and military applications are the more specialise uses. However IBM (and Sony and Toshiba) see the cell chip (perhaps with one SPE and a simple embedded CPU rather than the PPE) as an embedded processor for mobile phones, TV and media encoders/decoders/(de)compressors etc. The embedded devices may even find their way into PCs or graphics cards as hardware MPEG decoders for graphics cards with TV decoders built in.

IBM has also mooted the Cell chip with it's network security model as a Java substitute. In other words, instead of using the Java VM for security and applet download capability on embedded devices, the patented Cell migration of processes across the network would be used, except instead of a VM emulated in software, the Cell concepts are directly implemented in hardware, resulting in very high speeds. This could revolutionise downloadable media processing in devices for which Java is far too slow.

 

[dodo3]

Off-topic for a second, since Cell is created by IBM, Sony, and Toshiba, when military or medical companies buy a Cell-based computer, do Sony, IBM, and Toshiba all get royalties or does just IBM get royalties?