POWER Processors, SMT and the True Origins of AI

Crow writes "IBM developerWorks has posted an interview with John McCalpin, one of the guys who works on the POWER line of processors. He discusses work on POWER5 (and how the design process works at IBM -- he's also involved in work on the POWER7) and defends the decision that IBM doesn't hand-tune their ICs (as has often been criticized on Ars Technica. Also covers some of the features in the POWER processors, like SMT, the Hypervisor and virtualization -- even addresses the question of whether AIX was designed by space aliens or not. The POWER5 just broke the 3 million transactions per minute barrier on the TPC-C benchmark."

Comments

AI

[jbloggs]

Where's the bit on the true origins of AI?

Re:AI

[shadow303]

I think that was supposed to be AIX instead of AI.

Comp Eng at IBM

[KingOfTheNerds]

I worked recently as a contractor for IBM where I helped contribute to the design decisions for the POWER7. This article is great because it describes how the design decisions happen at IBM and it is right on the money. It's interesting because the way IBM does it is different from most other companies.

Vector supercomputers vs clusters

[CaptKilljoy]

A good quote to break out when people claim that cluster supercomputers are "better" than vector supercomputers (when they're really two different types of systems with differing strengths that can't be directly compared):

>McCalpin: The majority of high-end systems are used for throughput workloads of one kind or another. The vector machines, both from NEC and Cray(TM), are very well-liked by the end users because vectorization is a relatively easy thing to understand, how to write code that will vectorize. And the machines -- with relatively little effort -- give you a good utilization, you'll get a good fraction of the peak theoretical performance without a whole lot of work. And customers find that comforting. You put the code on there, you get 35% of the theoretical peak performance and you say, "Well, that's pretty good and I don't need to mess with it anymore."

On the machines that IBM sells and that HP sells and AMD(TM) and all of the others, the costs are much lower, but it's harder to get very high utilization on those machines, in part because they don't have so much expensive memory bandwidth. So there's an interesting discrepancy between the end users who love vector machines because they're easy to use and then the purchasing manager who doesn't like vector machines because they cost too much.

Re:POWER and desktop computing

[zpok]

Suppose a 90% Windows marketshare - and be very nice and political and give Linux and the Mac each 5%.

So what you mean is the whole Linux 5% would switch over to PPC beige boxes?

Considering the fact they already can today (well, beige is a bit out of the question, but Linux on PPC is a definite - and pretty cool - option today, and there's always Darwin, BSD and related options) don't you think that even a 5% increase of PPC's is far-fetched in this scenario?

Or is Windows supposed to support IBM chips suddenly?

Seriously, am I missing something?