Visit Your Local PBS Station PBS Home PBS Home Programs A-Z TV Schedules Watch Video Donate Shop PBS Search PBS
I, Cringely - The Survival of the Nerdiest with Robert X. Cringely
Search I,Cringely:

The Pulpit
The Pulpit

<< [ What Carly Will Be Missing ]   |  It's the Software, Stupid!  |   [ Killing Us With Kindness ] >>

Weekly Column

It's the Software, Stupid!: The Cell Processor May (or May Not) Defeat X86, but the Deciding Factor Will Be Software, Not Hardware

Status: [CLOSED]
By Robert X. Cringely
bob@cringely.com

My son Channing turned three last week. A few days before, we were finishing a meal of takeout Chinese food when he asked me which fortune cookie was his? I suggested that the one in his mouth probably qualified. But he said no. That was someone else's fortune cookie he was eating. Which of the remaining cookies was his?

Outwitted, by a two-year-old.

I think a lot of us were outwitted last week when IBM, Sony, and Toshiba unveiled their Cell Processor, which is slated to be the brains of Sony's PlayStation3 and an unnamed workstation from IBM. The chip certainly looks impressive with its modified PowerPC CPU and gaggle of co-processors. On pure bandwidth alone the Cell appears to beat the heck out of just about every other processor -- a true supercomputer on a chip.

Not so fast.

We've heard much of this before. Remember Intel's own i860 and i960 processors? They were aimed at 3D workstation market circa 1988 and combined RISC with a gaggle of co-processors. Sound familiar? Yet i860 and i960 processors found their niche in networking products and video arcade games, not personal computers.

Remember the Emotion Engine from Sony's own PS2? Emotion chips (modified MIPS processors with and two very fast Floating Point Processors) were supposed to end up in personal computers, but didn't because the FPUs were single-precision, and any scientific work requires double-precision.

Any viable challenge to X86 comes down to software, as my friend Nick Pasch explains. Nick, who is now the CTO at Rolltronics, racked-up dozens of chip patents as a senior R&D guy at LSI Logic. He knows what he is talking about.

"No product has even been unsuccessful in the marketplace because the designer chose the wrong (or right) processor. In the times when processors were slow and memory precious, you could want something you couldn't make, but I doubt seriously that anyone suffers from such a problem today. You might not be able to battery power the portable critter, but at least you can design it."

"The X86 architecture didn't succeed because it was good (or bad) but because the Microsoft OS worked. If DOS/Windows had been written for the Motorola 68000, the world might be a very different place. The X86 architecture has survived several challenges from Apple/Moto/IBM, primarily. The key is, where the Microsoft OS is required, the X86 owns. Where the Microsoft OS is not required, and Real-Time applications are high on that list (can you imagine Windows controlling your car's power brakes as an example?!) the X86 slugs it out, often unsuccessfully."

"While at LSI Logic, I watched the MIPS 3500 RISC microprocessor (which had been Sun's processor for several of their workstations only two years before) become the heart of the original Sony PlayStation. Processor, Video Controller, and Bus Controller were all on the same chip. No off-chip timing issues and relatively easy to program. One reason Sony did so well was that their games came out about a month sooner than their competitors, because software developers had an easier job."

"Since processors do not drive applications (in my world), I do not find the announcement of a new processor all that simulating."

Sony embraced Metrowerks Code Warrior as the development environment for the PlayStation2. This was a brilliant move because it was a mature IDE that allowed very quick game development. But what are they doing for the PS3? As far as I can tell that is unclear. IBM says it has built (or is going to build) a compiler for the Cell Processor, but they haven't committed to do an IDE, nor has Sony. Game programmers I hear from salivate over the Cell Processor performance specs, but none of them so far claim to have the slightest idea how to actually write games for the thing.

I'm sure we will see games, though, because without games nobody will buy PS3s. But how many games will we see when the console finally launches? My guess is we'll see half a dozen hand-tuned games that do a great job of showing-off the Cell, but may have taken months to write. And then there is the issue of backward compatibility, since Sony says the PS3 will play older PS games. How are they going to do that? My guess is they'll just paste in a PS2 and support double booting.

Another reader chimes-in. This guy prefers that I not use his name or reveal his company but he, too, knows what he is talking about:

"It seems to me to just be a re-run of all those computers that look amazingly powerful on paper, yet are absolute dogs to program (e.g. all the Connection Machines, Fujitsu CAP, all those vector processors, Transputer, MasPar etc.) As someone who did their PhD dissertation using the MasPar (a lovely machine that provided 16,384 processors and gigaflops in the early 90's), I do know that massively parallel programming is a lot of fun. However, as well as being a lot of fun, there are twice as many things to think about while you're actually doing the coding, not to mention the time taken to acquire the necessary skills. Even MMX/SSE instructions only get used where somebody gets down and hand-codes the assembler. Unless it excels as a simple CPU, I can't see that all the parallelism bells & whistles will do much more than make it too expensive to be used."

I have no inside information and I would love to be wrong about most of this, but I think Sony is in some difficulty. Part of that difficulty may come from the way Japanese companies are run. I have spent a lot of time in Japan so I speak here from some experience dealing with many companies, including Sony.

The typical Japanese executive will apologize for anything EXCEPT what he has actually done wrong. The weather, lunch, your hotel room are all fair game, but if he's missing a deadline, all you'll get is stony silence. This extends to whole companies. So Sony said it was going with the Cell Processor, set some delivery dates (since changed) and explained in certain terms how they were hoping to go about revolutionizing the world of video games. Don't believe it. The dates have changed, the specs have changed, who is doing what has undoubtedly changed, and -- since this is among the most complex technical projects in the history of consumer electronics -- there is a huge amount of risk. But Sony also has very deep resources, so I'm sure they are madly trying to either work or buy their way out of this mess.

One indicator of the state of things is that now Sony is saying it is "betting the company" on the PlayStation Portable (PSP), trying to divert attention from the PS3.

I think they need an extra fortune cookie.

Comments from the Tribe

Status: [CLOSED] read all comments (0)