Cell

I read all the news items about the Cell processor, and had three initial thoughts.

The first was, “This is just a PowerPC with little helpers.” The Cell processor is a PowerPC with eight extra floating point units and extra cache attached.

The second was, “How much is this Playstation 3 going to cost? The CPU takes up 221 sq. mm at 90nm, which is even bigger than 1Mb Hammers. Not exactly what you want in a cheap-by-PC-standards box, and this PS3 is supposed to have three of them?

The third was, “If IBM can’t make a 90nm 3GHz PowerPC for people who’ll pay a lot for it, how are they going to make a 4 or 4.6GHz one for people who won’t? And do it with 30 watts power? Either that 4/4.6 isn’t doing much work per clock cycle, or IBM has come up with some massive breakthroughs in process technology in the last few months. Somehow, I don’t think it’s the latter.

Then I thought a little bit more.

This CPU is built for parallelism from the ground up. That’s going to mean a whole lot more work (including learning) on the programming end. Granted, if you want to make a PS3 game, you’re not going to have a choice in the matter, but what about anything else?

And just what is all this wonderful hardware going to run on? The frontrunner would seem to be Linux, but why does the phrase, “One ring to rule them all and in the darkness bind them” keep going through my head when I think that?

Will IBM finally take over outside of a figleaf or two (they’re hardly going to wait around for people to do the necessary development in their spare time)? I’m sure they’ll let people continue to do free work, but when push comes to shove, you don’t think IBM is going to get its way? It would truly be one of the greatest ironies of all computing time if the Linux hobbits escaped the clutches of infinitely evil Microsoft just to get assimilated into IBM.

Something IBM and Company didn’t seem to want to talk about very much were those tasks that could not be passed off to a floating-point unit. You’d think a 4Ghz+ PowerPC CPU would be something to talk about; that alone ought to do a number on anything AMD or Intel will have in the next couple years, but they didn’t.

Not OMG, But Good Enough?

My gut reaction to all this is that we’re getting hyped on all this, and what we’re going to end up seeing in something like a PS3 isn’t going to make companies like Intel and Microsoft or AMD evaporate on first contact.

However, even if you dismiss 80% of the claims as hype as it applies to the PS3, what remains would still be good enough to give those companies a very rough time.

For instance, even given 65nm architecture, I just don’t see how you can have three of these processors in a cheapish PS3. Indeed, I suspect that those schematics we’ve seen of multi-Cell systems might be for the workstations we’ll see later this year.

Then again, even if there’s only one of them in there, if you’re a parent with a kid or two, a hot gaming box that can also handle typical PC functions OK would be awfully appealing.

The truth is, you don’t need something ten times better than today’s PC to beat Intel or Microsoft. All you need to get a stampede over the next few years is something somewhat better, cheaper, and most important of all, easier to use.

Yes, there will be XBox 2, which will be MS’s answer to all this. Only time will tell how those two will compete against each other.

However, it’s hard to see how the big eventual loser in any such competition won’t be the PS3 or XBox 2 makers, but the makers of the big boxes. If the little people don’t need the big boxes anymore, what happens to the PC industry?

Ed