The Cell Chip - how will MS and Intel face the music?
Part II: Place your bets
Analysis A number of readers consider Intel and Microsoft the two dumbest companies ever to file a 10Q, and rejoice at the prospect of an upstart - almost anyone will do - dethroning them. But be careful for what you wish for: it might come true.
Such a dismissive and negative view of these two giants isn't fair, of course. They're very different animals and lumping the two together overlooks all kinds of interesting internal tensions and contradictions. Besides, anyone who can remember that fine purveyor of ladies' lingerie by mail - "gazelle.com" - or any of the other bubble companies that sprung up in this parish during the dotcom bubble, will really know what a dumb company looks like. But for the sake of argument, we shall let the proposition stand.
Since the two giants established their hegemony around 15 years ago, we've seen many candidates threaten to unseat the PC duopoly. RISC, Unix, and Internet appliances (with or without Java) were all taken seriously as competitive threats in Santa Clara and Redmond. The Cell, from Sony, IBM and Toshiba is the latest; it will be unveiled next week in San Francisco and will ship later this year in the PlayStation3 console and in enterprise infrastructure from IBM. But what will a world with Cell supreme look like?
There are two striking aspects to this wonder-chip, if you read our digest of Microprocessor Report's analysis of the Cell patent this week. If you haven't, speed read it, here they are again.
The first is that Cell is designed to be a component in a massively distributed, global computing infrastructure. It's hardware specifically designed for "grid computing". A world full of Cell chips allows an entirely different infrastructure to take the place of today's transaction-based data centers. Software processes will scavenge the resources of the local Cell instantiation first, but if they find more execution resources over a local area network, or even on the other side of the world, they'll go and find them, and execute there.
All previous generations of computers have been based on the idea that it's only efficient to execute the task required inside the box itself, or on one nearby. If this doesn't pan out, the machine is designed to refuse the offer gracefully and give up - and then offer to call the supplier on your behalf for an upgrade. Similarly, none of today's operating systems (we'll come on to Sun in a moment) can migrate workloads across the planet. And we all know what they do when they're overtaxed.
So the Cell architecture has the potential to make computing global. And so a model where you rent computer cycles as a utility, and don't really care where they come from, becomes possible. This is a bit like the old days of the time-shared computing bureau, only this time round, you'll have a choice of utility providers. At least, that's the idea.
The second striking aspect is that this black box finally seals the era of the mythic, have-a-go-hero hacker. You know - the one who's forever saving the world from evil - like Jeff Goldblum in Independence Day, or a blogger at an O'Reilly conference, with his RSS feed.
The Cell is designed to make sure media, or third party programs, stay exactly where the owner of the media or program thinks they should stay. While most microprocessor designers agonize about how to make memory accesses as fast as possible, the Cell designers have erected several (four, we count) barriers to ensure memory accesses are as slow and cumbersome as possible - if need be.
Neither idea is new, but the wild card here is Sony. Not one of the previous threats to the PC hegemony - such as 'CISC vs RISC' - has involved putting one of these devices into the home. But that's where it will go. Cell is designed to scale from handhelds to vast data centers, but the market will begin not with a few enterprising early-adopters, like the micro or the TiVO did, but with a mass market. The Cell will soon be used in quantities that even the giddiest marketing chiefs of Intel at Home, or Windows Media Center would not dare type into a spreadsheet forecast. That's enough reason to take notice.
Now, we have to ask - what chances does it have of succeeding?
There are technical reasons that weigh heavily on whether each of these propositions will succeed. For example, Intel's Itanium depends on compilers parallelizing the code, and has foundered because this is difficult, and cheaper, dumber does a better job. Cell hardware will need really great compilers to work. But in the end, technical arguments like these won't be the decisive factors. We have to step right back and look at how and why people depend on computer technology, and exactly who in the world stands to benefit from each - and there are many - of the possible "victory" scenarios.
Next page: My business cycle, your aura