[rescue] pursuing a VAX

Francisco Javier Mesa-Martinez lefa at ucsc.edu
Mon Dec 9 16:08:19 CST 2002


> I work in MCAD (Mechanical Desktop) and my work system -
> a lowly Pentium III 866MHz - "feels" at least as fast
> as any of my colleagues' Pentium 4 1.7-2.0 GHz systems.
> These are all Dell boxes, configured similarly, BTW. My
> system was purchased about 2 years ago, and some of the
> other systems as recent as last month. So, Moore's law holds
> true - clock speed has doubled - but performance has not
> improved commensurately. Go figure.

Pentium 4's are really weak when it comes to FPU, I assume that you CAD
package is mostly FP-centric... However they have some nice INT numbers,
which is what people use most of the day (UI, WordProcessing, etc).

MHz per MHz the Pentium III is better than the 4, however the later was
designed to scale better when it comes to speed, hence the ridiculous
number of stages in its pipeline. The P4 also has some terrible penalties
for things like missed branches (heck you have a lot of crap to flush on
these deep pipelines).

Intel realized that their strength is not their microarchitecture, but
rather their production technologies.. which so far almost nobody can
beat. So that is what they are using, in the end it produces waht they
wanted, some of the fastest CPU (at least in paper) on the market. Granted
it is absolutely not an elegant way to do it. But like war, it is not
about who has the prettiest uniforms, but rather about who gets to win it.

Again, we come to another full circle in the computer industry. A decade
ago there was the same debate: Speed Demons (MIPS, Alpha, HP-PA) vs.
brainiacs (POWER, 88K).....



More information about the rescue mailing list