Everyone who has been worried about the forthcoming 1.5GHz Pentium 4 and the aggressive Intel roadmap that puts the P4 at 2GHz before Apple plans to ship 800MHz G4es can breathe a big sigh of relief today -- even Wintel partisans PC World have slammed the Pentium 4 for choosing clock rates over performance.And youyou know who you aremanaged to provoke me. I can't stand it. Time for another episode of When Damage Attacks.
Here's my take:
Once wild zealots for the merits of RISC-y microarchitectures, including scalablity in terms of clock frequencies and performance, the Mac-minded are now deeply committed to the wonders of clock-for-clock performance. The considerable merits of the P4's NetBurst architecture are lost on them. No doubt the glut of ne'er-more-than-500MHz Mac cubes has something to do with this. (Listen, guys, if that's what it's all about, I'm sure we can dig up a few K6-3s to sell you. They're absolute beasts at 450MHz, and they double as hoss room heaters during the winter.) Amazing how the Mac-ites' superior knowledge of CPU architectures can't be brought to bear in evaluating new x86-compatible designs.
The Mac processor shock troops also delivered this P4 launch day tirade, apparently intended to blunt the impact of yet another x86 competitor to the G4. Riddled with snobbery, contradiction, and technical silliness, this article will amaze and perplex you. For instance, there's this:
One can easily understand why the smiths hammering away at the red-hot silicon in the IBM and Motorola foundries might be a little miffed at the “not enough megahertz” complaints coming from users. They have not been sitting around on their duffs. They have been working hard to improve their products.Which is all well and good, except that it seems mostly untrue. They have been working hard to make the G4 a better chipto stick into a Ford Contour or a Maytag. The desktop market hasn't been getting proper attention. Nevertheless, our Mac-lover author thinks the PPC trio is doing just fine, thanks.
That is, until we get to this complaint further along in the article:
Sadly, just as all of this is finally taking place, Motorola seems to have elected to sit upon its laurels and look smug, not even troubling itself to issue an tentative upgrade schedule for its variant of the PowerPC, the G4. What makes this particularly annoying and alarming is that Apple has chosen Motorola’s G4 chip for the heart of its most powerful desktops. Now it looks as though we will finally get an operating system fully capable of exploiting the PowerPC, just as Motorola has decided to allow the chip to languish.Don'tcha just love contradiction? I get the sense the first quotation exemplifies the pretty face Mac fanatics like to put on the Motorola G4's inability to ramp past 500MHz, while the second one is more internally directed toward the Mac community as they deliberate the merits of a possible move to OS X on x86which explains the dissonance, but doesn't excuse it.
And you'd think Mac fans, of all people, would understand the need for new compilers and revised code to take advantage of new processor designs. Heck, they're still emulating 680x0 CPUs in common system calls. For the love of all that is good and holy, figure it out, guys! Then again, anybody who can endure the P4 launch hypefull of talk about decoders, micro-ops, and trace cachesand still end up framing things in terms of RISC versus CISC probably isn't accustomed to thinking clearly about these matters, anyway.
For a much more clear-headed analysis of the P4's merits, do not miss this excellent analysis by Paul DeMone. He explains lucidly why clock-for-clock performance comparisons aren't really helpful, not only because older designs don't ramp as well in terms of frequency, but also because they suffer in terms of instructions per clock (IPC) as clock rates rise:
This means that relative to a hypothetical PIII design that could keep up in frequency, the P4 looks better and better, and eventually exceeds it in "IPC" as the clock rate heads towards the 2 GHz that Intel claims is possible in their 0.18 um process and beyond in 0.13 umJust go read it; it's brilliant. DeMone concludes that the Pentium 4 may make life difficult for AMD over the course of the next year. No doubt it'll crush Apple's RISC wonder in the process.