Perspective: Cell chip: Hit or hype?

Will the Cell processor be the new Itanium?

At the International Solid-State Circuits Conference on Monday, the joint developers of the long-awaited processor--Sony, Toshiba and IBM--unveiled a number of the details about it amid a surge of dramatic speculation. The New York Times said the chip could create "a new digital computing ecosystem that includes Hollywood, the living room and high-performance scientific and engineering markets."

Others speculated that the chip could drive everything from cell phones to servers, tying them into a grand computing grid.

"We believe a 10x performance over the PC, at the same power envelope, can be achieved," said IBM's Dac Pham, one of the designers of Cell. "It will usher in a new era of media-centered computing."

This sort of excitement and speculation about chips is driven by the "Battlestar Galactica" principle.

Intel's limping Itanium debuted with a similar level of fanfare. In 1994, the Microprocessor Report, examining the investment Intel planned to put behind the chip, predicted that it would become commonplace in desktops by 2004. It didn't happen.

Similarly, feelings ran high about the Emotion Engine, the microprocessor inside the original PlayStation 2 game console. Analysts said it could undercut chips from Intel and Advanced Micro Devices in PCs, and become the nerve center for DVD players and other home electronics. Toshiba even created a company, Artile, to license the Emotion.

But the Emotion Engine never migrated outside the PlayStation, and Toshiba snuffed out Artile in 2003. The PlayStation 2, meanwhile, didn't live up to the suggestion that it would serve as a conduit for movies, TV, e-mail and the Internet.

This sort of excitement and speculation about chips is driven by what I call the "Battlestar Galactica" principle. It goes as follows: If the domination of the universe isn't contested on a weekly basis, ratings will go down. Analysts, reporters, consumers and even executives need a gladiatorial contest to keep the job interesting.

The high-public profile of Sun Microsystems can partly be attributed to its role as the William Shatner of computing--donning a new uniform every three seasons to battle a new nemesis.

Put in that perspective, the Cell story starts to look different.

Cell will be a victory if it doesn't lead to layoffs.

Going by papers presented at ISSCC, Cell looks like a tremendous achievement. However, this is the chip industry: Only a handful of companies--Samsung, Intel, Texas Instruments and Taiwan Semiconductor Manufacturing Co.--consistently produce profits. Most everyone else is seemingly always two steps away from the trailer park. Over the past few years, IBM Microelectronics has often reported quarterly losses. Cell will be a victory if it doesn't lead to layoffs.

In all likelihood, Cell will sell in far greater numbers than the just-as-trumpeted Itanium. Sony will put it into the PlayStation 3 video console. Unless gamers lose interest in stock cars, ninja stars and wiping out space aliens between now and 2006, that thing will sell. IBM and Toshiba will put it in products, too.

Still, whether the chip will be able to enter different markets is another question that hinges on factors such as:

Size: Cell contains 234 million transistors and takes up 221 square millimeters in the 90-nanometer production process. That's about double the size of the 90-nanometer 3.6GHz Pentium 4, with 112 square millimeters and 125 million transistors.

Why invite your rival to your top-secret design meetings?

Big chips cost more to produce, can hide more bugs and can be tough to cram into portable devices. Cell will get cheaper when it goes to 65-nanometer production, but so will the alternatives.

Cost: Remember liquid crystal on silicon (LCOS)? The chip that would bring down the price of big-screen TVs? Intel and Brilliant Technologies failed at it. JVC and Sony succeeded. However, the latter two companies sell their LCOS chips to their own television units. The cost of the chip gets absorbed into the TV set.

Sony, Toshiba and IBM don't have to worry about the cost of Cell because they will sell it to themselves. It becomes part of a product that is tagged at a slightly higher price. An expensive Cell, however, will be a tough sell to any other manufacturers.

Alliances: Consumer electronics companies won't want to buy a processor from Sony and Toshiba. Similarly, not a lot of server manufacturers will line up to buy a Cell server chip from IBM. Why invite your rival to your top-secret design meetings?

Power: Cell will have to be air-cooled, IBM said. In other words, fans will probably be required. Ever talk on a cell phone with a fan?

While IBM didn't disclose the exact heat statistics, some at ISSCC said it could run as hot as 130 watts, more than most desktop and notebook chips. If Cell is in this range, kids will really be huddled around the PlayStation 3 at Christmas--for warmth.

On the cool engineering side, however, the chip will come with 10 digital heat sensors to warn of problems and another sensor to regulate temperature.

Memory: Cell comes with an integrated memory controller for high-performance XDR memory from Rambus--which means that the current design works exclusively with this pricey stuff. Sony used an earlier version of Rambus memory in the PlayStation, but it's been a tough sell outside of consumer electronics.

Cell is an outstanding achievement. But we have to wait and see whether it can get a job from someone other than its parents.

Biography
Michael Kanellos is editor at large at CNET News.com, where he covers hardware, research and development, start-ups and the tech industry overseas. He has worked as an attorney, travel writer and sidewalk hawker for a time share resort, among other occupations.

More Perspectives

[alex_H]

I think cell processors are totally the future

I think, like, that processors have been too small for too long, man, and that its time we had a revolution towards larger sizes. Too often have I lost mine down somewhere stupid like the toilet, or its fallen out of a hole in my pocket. The move to new fronteirs of size is just fabulous. Also, I like the fact that its faster.

RAMBUS???

While I think you are mostly right, cell technology will be part of the future. I think IBM/Sony's use of Rambus technology will be their implementations downfall. Rambus has been laying low trying to repair it reputation, but it is still going to try and wring every cent it can from this. My guess is that they have already submitted some more patents to try and lock this technology down (if IBM did not do it first) and profit from it, all without producing a single chip themselves.
You will not see most of the mainstream chipset makers(Intel especially) use Rambus again after getting burned the first time by their technology.

Of course this is my opinion, not to be confused fact...

Rambus is not the devil

Rambus-bashing is so last century! Despite their lack of tact in the past, which has burned them severely in the desktop space, they do have very interesting and very fast memory and interconnect technology. (Their mistake with the PC was trying to sell expensive high-speed memory at a time when memory throughput was not the major bottleneck. As it became more of a bottleneck with multi-GHz P4's, Rambus memory demonstrated it strengths, but, by then, it was too late.) XDR will not be the downfall of Cell. Hype and ensuing unmet expectations are more likely hazards.

The Cell - Tough Market

Great article Mr. Kanellos. Cut right through the pomp. To get onto a PC a chip has be a lot more than good as AMD found out. AMD, you will recollect, never ran well on a board designed for Intel processors. It took years and gallons of red ink to grow the volume to the point where it became feasible for motherboard manufacturers to produce motherboards for AMD processors. Until that happened, AMD processors always had to make compromises to run on existing hardware designed for Intel processors. And this was a processor that set out from the get go to run on existing hardware without infringing on Intel intelectual rights. The Cell makes no such attempt. The cell is designed for proprietary game boards and not PCs. They are going to have to manufacture the cell and the motherboard it runs on at the same time and sell it at a competitive price in a market in which volumes for existing products have already been developed.

Tougher yet, they have to get Microsoft onboard with an OS that supports the cell or design the cell to support Microsoft OSs without infinging on AMD and Intel instructions sets. Without Microsoft they would have to also develop an OS that runs all current software. They should talk to Linus Torvaldi about that act.

Speaking of Linus, there is always Linux of course, but until Linux gives up on GNU and Open Source it is never going anywhere. Ever. It will remain a great OS for dedicated applications but will never reach the consumer desktop. That, however, is another story.

[raylopez]

I disagree--why do they need MSFT?

I'm not an expert in this field but I would think that with a standard API of the kind TI regularly supplies hardware designers, that CELL's owners could do the same thing: provide such an API (free of charge hopefully) to hardware designers that want to build applications using CELL's architecture. Why do they need MSFT to build an OS? Of course that would be ideal, since every high school programmer could use the best in the world MSFT MFC API, but I don't think experienced hardware designers need such hand holding. I could be wrong however.

[scdecade]

Intel needs to wake up.

As a Sun shareholder I think Intel needs to wake the heck up and buy SUNW. If Intel follows their own processor roadmap, they'll be left further behind. IBM is looking very strong.

parallel tasks

The 10x performance pressbyte is derived from the Cell's parallelism, but they don't want to mention what will NOT benefit from this. Rewriting parts of an OS that apply graphics algorithms etc. are targets of course, but the Cell won't make Office run faster. And reasonably priced graphics engines that already naturally focus on parallelism have been around on gamer cards and increasingly in board chipsets, and won't be just blown away by Cell. So even with more advanced dataflow in the parallel area, the Cell's biggest difference may be that it packs everything onto one big chip. There is hype per Mr. Kanellos' Battlestar theory.

[Llib Setag]

Dual Processor parallel 64 bit chips + OSX

Mac OSX UNIX/Open Source/FreeBSD/Dawin based OS that will
NOW operate on dual 64 but IBM POWER G5 chips, plus will by
June be updated to Tiger 64Bit OS version.

OSX has many Linux & Unix developers that use X11 program to
modify their programs to run on the Mac OSX Darwin flavor of
UNIX.

If IBM is developing the new CELL parallel chips & is currently
supplying 64Bit G5 chips to Apple, then it is not such a strech to
think that PowerMacs & Xserve Apple servers will soon use the
Cell Chips. Mac OSX Tiger64 GOLD Public version will be issued
in just a few months to run on dual 64Bit chips.

Major developers such as Microsoft/Adobe/Macromedia/Alias
Maya,etc already run on the OSX/G5 chips, so with X11 & Apple
developrs, it should not be a great task to tweek future
programs to run on an IBM Cell G5/G6 64Bit chip with Mac OSX
Tiger 64Bit OS.

[mortis9]

it is a stretch

it's a stretch in time: the current G5 is based on a single core power4 processor. who knows when they will transition to a power5 architecture or dual core, for apple. That said, ibm doesn't plan on using cell to replace the current power line. that said, i doubt apple will ever see cell machine with their name on it.

It is not CELL. It is PS3, Intel and MS fears.

The CELL chip is just the start of the end of home-pc era (It was called IBM-PC). Intel's fear is the saturation market of Home-PC. Once the PS3 is in the market, almost every home will buy it. Adults and kids will love it. Then they will find out that PS3 can do everything in addition to the gaming, that current PC is able to, e.g., web-surfing, e-mail, home entertainment, music and movie streaming, word-processing, etc. If SONY will have the LINUX as OS, then Microsoft also should be worried. SONY, Toshiba, and IBM are making new home PCs with CELL. CELL has a PowerPC core. We have to remember that Apple and other systems use the PowerPC architectire. PS3 is not a simple gaming machine. It is a new-PC to come!

[hberenson]

That's what everyone said about the PS/2

Before the PS/2 shipped everyone was saying exactly the same thing. Even one of Microsoft's senior architects was running around the company making this point to every architect and VP who would talk to him. The PS/2 was a bust as anything other than a gaming console. And it's likely that's where the PS/3 will be stranded as well.

All the panic about the PS/2 did was spark Microsoft to create the Xbox. So even if the PS/3 does gain traction outside of gaming, Microsoft already has a response ready with its next generation Xbox.

[Unknown2You]

Crazy

That's just crazy! Why would someone want to do word processing on their TV? People are always going to use a PC. I can just imagine corporate offices using the PS3 to create documents.

PS3 is a console and Cell is not that revolutionary

The PS3 lack the support of a strong OS. You propably could do everything a pc does with a PS3 but that's not going to happen. The PS3 would easily break 2000$ if it had a competent OS. Cell on the other hand is built with the concept of parralelism in mind. Each of the core will have it's specific instruction given by the application. For gaming a developer could use a core to process A.I while use another to process physic and so on. Multithreading is the future of computing and that's what Cell is designed to do. Personaly i dont think a major player like microsoft is just going to vanish because a new cpu it's the market. As long they can do C++ on it or any other language there's nothing restricting them to make a Cell compatible OS. Seriously do you think a 11 years old would use linux with a gamepad!!! oh and by the way as long as IBM is in the project pc have nothing to worry about.

[Stomfi]

A P2P chip

The Cell is P2P at the hardware level. The long awaited P2P ISP network, secure, no spam, no illegal use, can be done with home connection boxes based on this chip.
And I wont have to have another stupid old fashioned PC, I can spend all my money on the Human Interface parts.

This is proof

that the audience can also be afflicted with Battlestar syndrome.

[raylopez]

LOL a funny tech article <g>

The article was well done, pointing out some apparent defects in CELL and injecting humour at the same time. LOL, funniest article I've seen in a while. Well done; that travelogue experience from Cannellos came in handy.

[henry wattson]

Some things to consider

STI are spending Billions on plants to build cells.
There will be many version of the cell. Version in cell phone will
not need a fan. It will have fewer cores and run at less than 4Ghz

Graphic card for the PS3 will not be built by Sony but will use a
cell processor. Not the same cell processor in the PS3

A majority of cost are in R&#38;D. IBM has project cost of $70 to
$80 a chip.
Xbox is Not using a dual core G-5. Its Using a chip based on
power and has special features developed jointly with IBM. We
do NOT know details yet.
The G-5 is a modified 970.
The cell DOES NOT use a g-5.
It uses a power chip, Likely a 980, that has been modified.
IBM has been working on the 980 for 2 years.
Apple will come out with a G-6 computer in 2 to 15 months. We
DO NOT know what the G-6 will be.
People ASSUME the 980 is based on the Power5.
A Modified 980 COULD be one of the cell processors.
The Cell was not designed for desktop PC's.
The cell will find it way to PC's IF IBM can sell it at the right price.
Most PC users do not need more power.
Cell will Run Lenox.
MS has to provide server software for the cell or see it market
share in the server market greatly reduced.
The Cell COULD be very successful without hurting MS
Intel needs MS MUCH more than MS needs Intel (not always the
case)
The Cell should dominate the workstation market.
Intel Failed with the Itanic(sic) because it was delayed and by the
time it came out IBM had comparable processors with an
installed base running software designed for it.
Supercomputers Will be built using the Cell.
Workstations based on the cell MIGHT run photoshop and video
editing programs so fast that people will Switch from Apple.

The cell will be used in many markets not mentioned
It is LIKELY that more cell processors will be used in
automobiles than in computers.
The cell COULD result in greatly reduced prices in HDTV.
Other HDTV cannot afford not to use the cell.
We WILL see the Cell demoed with 90 days
IBM has 32% of server market, has gain market share for 14
straight Quarters.

[zamene]

Boon for Open Source/Linux

This is a very exciting product for the open source community. Since Cell is capable of running with Linux, this may be the killer combination for the demise of the Microsoft operating systems and the continuing rise of Linux.

Of course this depends on a lot of factors, e.g. price/performance of hardware/motherboards,how quickly IBM can enter the market with servers using Cell processors, and if some minor problems may be corrected with the GUI of Linux, etc.; Microsoft either needs to develop into an open source OS or quickly develop an OS that will run on the Cell. I am sure that they are seriously considering/doing the latter at this very instance. If they are not you may want to unload your MS stock, before it is to late.

[Llib Setag]

Boon for Linux + X11 = Boon for OSX UNIX

Go to X11 Web Site for Linux/Unix program conversion to OSX
Unix/Darwin:

<a class="jive-link-external" href="http://www.apple.com/macosx/features/x11/" target="_newWindow">http://www.apple.com/macosx/features/x11/</a>

OSX Tiger will be released in a couple of months as a GOLD
PUBLIC 64 Bit Mac OSX Unix OS that many Linux &#38; Unix
developera are already developing / converting software
programs via X11 for.

Who knows... IBM POWER chips are inside of G5 single &#38; dual
processor Apple computers. Are Dual Core IBM POWER CHIPS OR
IBM CELL chips in the near future for Mac OSX PowerMac Towers
or Xserve Servers?

Don't hold your breath on MS developing an IBM CELL POWER
CHIP OS version of Shorthorn. It's taking them forever to develop
&#38; market their upcoming cure-all-savour OS (Longhorn). MS stil
shaved out major upgrade components that would make it
desirable to upgrade to Shorthorn, now the only thing going for
it is that it will be a desktop computer 64 Bit OS. Only Beta 1.0
test version by June.

Mac OSX Tiger 64 Bit Unix OS GOLD Public Version by June.

With X11 &#38; Terminal Window "pop the hood Unix custom
coding" feature of OSX, Linux &#38; Unix developers can customize
OSX to deliver powerful Linux/Unix options on top of Mac OSX
Unix.
Future is bright for Open Source/Linux/Unix/Mac OSX Unix-
Darwin-NeXT-FreeBSD-Apache based OS.

[panacealater]

Apps alutely fabulous

Not a tech historian, I am interested in whether the apps that
will grow out of this shift will prove its success or failure. What is
the likely lead time for non-PS3 apps to show up in broader
markets?

[lawrephord38]

wanted A BLOCK 9 LAYERS OF COOLING 8 LAYERS OF CPU WITH MORE THAN 8MEG CACH

YOU NEED A 3D PAINT EDITOR FOR THIS
---------------------------------------------------
[[http://..........COOLING1...............]|http://..........COOLING1...............]]
[[http://..WIREBUS1.......CPU.............]|http://..WIREBUS1.......CPU.............]]
[[http://..........COOLING2...............]|http://..........COOLING2...............]]
[[http://.................CPU....WIREBUS2.]|http://.................CPU....WIREBUS2.]]
[[..........COOLING3