Pentium 4 has us flummoxed


— 1:10 AM on November 29, 2000

Folks just don't know what to make of the Pentium 4. It's so radically new, so good (or potentially good) in some ways, and so bad in others. This new NetBurst microarchitecture is more of a departure from the status quo than anything since the Pentium, I'd wager, and its performance balance is just plain funky. Memory access is stellar; the front-side bus is blistering; and the clock rates are higher than a pre-bust Robert Downey, Jr. But the basic integer performance isn't much to write home about; the really fast stuff (SSE2) requires specially optimized code, and the x87 floating-point unit is a dog.

Had AMD introduced this microarchitecture a year ago, they'd be YAVS (Yet Another Via Subsidiary) by now.

But this particular square peg comes from Intel, who has the oomph to stuff the thing into a round hole if it darn well pleases. The industry knows it: chips will be sold, programs will be recompiled, plans will be adjusted, and all will be well by the time the P4 ramps up and ascends to the x86 throne. AMD didn't have this luxury with the Athlon, so they produced a chip that runs Pentium II/III-oriented code like Mario Andretti on smack.

None of this matters to PC buyers immediately. Intel is extracting profits from early adopters, and the P4 is far from being a good value right now. PC buyers would do better with an Athlon for the Christmas season.

The question is: how will the P4 and Athlon match up in six months? A year? Were Intel's choices and compromises in the P4 core design justified, or not?

Some of the discussion going on around the web is indicative of the difficulty reviewers are having trying to anticipate an answer to that second set of questions. Tom Pabst has yet another update to his Pentium 4 review, as he tries to sift through variously compiled binaries with optimizations for the P4, Athlon, SSE2, and the like. He's testing the FlasK MPEG4 video compression proggy, with some help from Intel. The terrifyingly feisty AMD Zone waded into the FlasK, as well, testing with (imagine this) Athlon-optimized binaries. All very interesting stuff, if you want to get a sense of how the P4 will stack up once software starts supporting it. (Or how the Athlon will stack up, once software starts supporting it. Ahem.)

Then there's Dean Kent, hovering above the fray and offering his own commentary on the difficulty of processor benchmarking and evaluation. He offers some sharp criticisms and expresses pessimism about the prospects for processor reviews—which is all true, except that such evaluations must be done. Having them done by reasonably competent folks whose requirements ought to be, conveniently, quite similar to those of their readers is no bad thing, in my view. (But then I have a bit of a bias here.) Still, Mr. Kent's analysis is generally clear-headed and very much worth reading.

In the end, I tend to think the public ambivalence we've seen expressed over the Pentium 4 is, in a funny way, precisely the appropriate response. Intel's taking us all on a very wild ride, and we're not sure we like it yet. But I think it will be fun to watch folks trying to make up their minds.

 
   
Register
Tip: You can use the A/Z keys to walk threads.
View options

This discussion is now closed.