A "nuclear winter" for CPUs?
Intellibees,
The slowdown of roughly the past year, affecting Intel, AMD, and of course the dot coms and optical providers (and the list goes on...), seems not all that surprising in retrospect. I sure wish I'd seen it coming in time to unload some Intel and not to have bought Red Hat, Akamai, etc. when I did.
Actually, Paul Engel and I used to talk about what happens when a lot of people--home consumers and businesses--are basically happy enough with their XXX MHz PC and don't see the need to replace it with a YYY MHz version. I remember arguing to Paul around 1998 that probably most users would be happy with machines of roughly 400 MHz, able to handle most demanding graphics apps and achieve that magical "full-frame video" sweet spot. Paul thought people would want more. As it turned out, the place they seemed to have "levelled out at" is roughly the 500-800 MHz Pentium III level, courtesy of a surge of buying for Y2K and Napster and video games in 1999-2000. More on this below.
I'm not an analyst, I don't pore over 10Ks and analyst reports, so my views are somewhat anecdotal.
Reasons for the levelling-off:
* A year ago, there was still much excitement about Napster, MP3, ripping CDRs, and multimedia in general. Kids going to college were setting up Napster collections in their dorm rooms. Those who had older machines bought new Pentium IIIs or Athlons. Games, MP3s, everyone wanted to have the latest.
* This has fizzled in the past year. Napster is effectively gone. College kids still have MP3 players, but with the fountainhead of free music turned off, the bloom is off the rose. (There are file-trading, or P2P, alternatives like Morpheus, Gnutella, Freenet, MojoNation, but these are lacking the _twenty terabytes_ (!!) of music I used to routinely find when I conected to Napster.
* The kids have moved on to other things. Having a 600 MHz Pentium III or a 900 MHz Athlon or whatever is good enough. I see no clamoring of most kids to have 1.4 GHz Athlons or Pentium 4s. (Turns out I see a lot of college kids on the Cypherpunks list I co-founded nine years ago. They are a barometer of sorts for what speeds of machines college kids are clamoring for.)
* 1999 was also a year for businesses across America and Europe to upgrade their funky old machines, networks, and software to be fully Y2K-compliant. A lot of old 386s, 486s, and even slow Pentiums were retired. A lot of software was upgraded to the latest versions...Microsoft Office, Oracle's data bases, Siebel, SAP, accounting packages, billing software, etc.
(Intel has acknowledged, IIRC, that 1999-2000 PC sales spiked upward because so many businesses were upgrading entire systems.)
* After a thorough housecleaning, after a series of upgrades of systems and software, it's not surprising that businesses are ready to "coast" for a while. The average desktop user simply is not doing CPU-intensive tasks that require gigaflops.
(I realize the circularity of the argument that people don't need more CPU power when in fact the longterm trend has been upward. The point is that there are "surges" in perceived need, as with Windows '95, multimedia, Y2K, and there are "lulls" in perceived need. A few years from now there may be desktop apps calling for gigaflops, but not now.)
* People are buying non-computer things. DVDs, for example. (I was at a Circuit City yesterday, shopping for a replacement DVD player. Lots of people in the t.v. and DCD areas, almost no one in the PC and Mac areas. While Circuit City may not be the preferred place to look at PCs, it's the main such store here in coastal Santa Cruz. It gives some idea of what local consumers are concentrating on.)
* The failure of the "last mile" efforts: most home users still don't have bandwitdth. DSL providers are going bust for bizarre reasons, ExciteAtHome is on the verge of bankruptcy, and most home users aren't connecting faster than 56K (if that). This has an obvious effect on the downloading of MPEG II movies, DIVX (not the abortive Circuit City thing, ironically), and other high-bandwidth-requiring things which would make faster machines more desirable.
I'll bet a lot of you reading this are doing so on machines a year or two behind the cutting edge. Fact is, a 600 MHz Celeron is still plenty fast enough for all but the most demanding of apps.
My own main machine is a 400 MHz G4 Macintosh (roughly the equivalent in CPU power, neutral reviews show, to a 600-800 MHz Pentium III, because of the architecture--This is not meant to provoke a flame war over PCs vs. Macs, folks!). While I would _like_ to have one of the new dual-800 MHz G4 machines, I just have better things to spend a couple of grand on. (I like the new iBooks, but I just bought a 466-MHz G3-based iBook last October.) And I'm quite happy with the performance of OS X and all of my major apps on this machine...the dual 800 MHz machine would no doubt feel "snappier," but would not allow me to do any more than I'm now doing.
What a lot of people, students, and businesses did was to go on a buying binge in 1999. For several different reasons. Now, sated with their new machines, they seem to be in no hurry to upgrade to a faster CPU that offers no other notable advantages except CPU speed.
(I hear analysts talking about this, so don't think I'm claiming to have discovered this effect! Just talking about it.)
What will stimulate demand? Video editing is CPU-intensive, but even sub-GHz machines are able to handle the load...the graphics chips matter more, proportionately. Apple is getting a good chunk of this business (iMovie, Final Cut Pro, a la the recent t.v. ad showing a couple getting married on a Polynesian beach and then "burning" a DVD for their parents to watch at home.)
I don't see WiFi/802.11b suddenly stimulating demand for CPU power in the main home PC. A lot of the wireless apps will be light on CPU use. (Intel's wireless efforts, and StrongArm/Xscale efforts, are wonderful. But I don't see wireless driving the upgrade cycle for _desktop_ or _laptop_ CPUs.)
So, perhaps a "nuclear winter" for a couple of years. Maybe another year to go...
--Tim May |
