The last week or so has been an unsettling one, perhaps the most unsettling one in this field in recent times.
What will multicores be like? Are they worth waiting for? Is there really a choice, what do we do in the meantime? How much, if any good will they do?
One glimmering note in this sea of uncertainty is an excellent point made by Charlie Demerjian over at the Inquirer.
He points out that the next generation of game consoles (XBox2, Playstation 3, etc.) are all going to be multicore.
This means that anybody who expects to sell any games for these game consoles is going to have to go in thinking and coding for multithreading from the getgo.
By 2006, you’ll have Microsoft AND Sony AND Nintendo AND Intel AND AMD all singing the same hardware hymm: Two heads are better than one.
What other choice are they going to have? Not like this is getting sprung on them out of the blue. With the consoles, it’s not like game developers can seriously pretend that the user is doing other things with the machine. It’s hard to check email when you’re under attack. 🙂
So for gamers, multicore ought to be a good thing from the start.
Then There’s The Rest of Us
This is good news for gamers, but what about the rest of the planet? There won’t be the same compelling competitive need for most applications to put multicore first.
Yes, even today, those relative few who use multiprocessors on desktops today like “the smoother ride” one gets when the CPU doesn’t have to jump back and forth on two big tasks. No doubt the ride will get even smoother once Longhorn shows up.
But how do you sell that to the other 85%+ of the computing population? Seems like a much tougher sell to me, especially when the computing being done on the box is so light to begin with that the task-switching bumps are barely or not even noticeable.
I mean, really, if you offered the average person a choice between a multicore machine, and a machine that was even slower than what they have today, but could boot up in five seconds, 90% of the Sixpacks would pick the second box.
Especially if it were cheaper.
Look at it this way. If you just spend a few hours a week on the Internet, a broadband connection will do the average person more good in real-time savings than ANY current or future computer.
Nonetheless, broadband has been a pretty tough sell, simply because it costs more. This ought to tell you that most people out there don’t exactly share your love of technology.
If the average person is reluctant to spend more for that, how much more so will they be to pay more for a slightly smoother ride?
The standard tech response to such consumer reluctance, of course, is “give them no choice.” I don’t doubt it will work well enough with the next generation of gaming consoles.
I”m not so sure about the PC industry the next go-round. For many if not most computer users, there’s simply no compelling advantage to multicore, especially for reluctant computer users.
If such people get offered a cheaper, lighter alternative two or three or four years from now; I think a large proportion of current PC users have at least the potential to defect to “thin and light.” Build it into a phone (maybe with a “desktop” docking station to start), and I think they’re going to take it.
People like portability; that’s why notebook sales are increasing much more so than desktop sales.
If the choice becomes one between unappreciated greater power and much appreciated greater portability, I don’t doubt how the vote is going to go longterm.
And that will really shake up the heavy iron PC industry.