Pentium 4 Re-evaluated, Again (Again)

An unnamed correspondent writes: "It looks like Tom's Hardware Guide has been busy with the P4. This time a re-compiled version of the MPEG encoder (the same one they benchmarked with in the last article) shows the P4 doing really well. Also interesting is the performance boost that even the PIII and Athlon procs get from the Intel compiler. Take a look at the article here." Seems that as usual, benchmarks are what you make of them. The P4 apparently can perform much better than initial tests have shown. Tom Pabst makes some good (if fawning) points about the complexity and fairness of benchmarking in general, too.

Comments

Re:The Pentium 4 Question

[Chris Johnson]

This is a very good point. _Too_ much of a disparity should be considered a red flag, a warning sign. Apparently the P4's performance is really brittle- if you pick just the right task to test it on it can scream, but hit it with something it doesn't like (something that beats on the excessively deep pipeline or doesn't logically break down into simple SSE stuff) and it bogs, it's an absolute pig.

Logically, then, we must all avoid attempting any tasks that the P4 doesn't like ;P

...ok, maybe not. How about some benches on tasks like POV-Ray scenes, SQL operations, kernel compiles? Some of these map more closely to the sort of tasks you'd get if (for instance) you were an EFX house and wrote your own software to do some heinous particle system rendering task. Those are the people who need _real_ CPU speed, and it's really questionable whether P4 is going to be any use to them at all. When you're doing that sort of stuff you have no time to fuss around with wizzy CISC operations- you're writing hopefully bugfree code and running it against a deadline to produce footage that is needed right away if not sooner, and you frankly don't care how Flask does since you aren't using it. You're running fairly general purpose code- a LOT of it- and live or die by your ability to (a) write it as needed and (b) run it. There's no room for "it's 70% as fast as an Alpha but if we spend another month optimising it for SSE we might be able to run it three times as fast, or it might go splat again if we double the number of scene elements or change a line!" no no no...