Above graph shows results from Jim's P5 CPU test survey. To make it easier to see trends, I've chopped off all results over 650 secs (there were 4). Also AMD 1800+ is treated as AMD 1.8GHz etc. I've also ignored amount of RAM and OS as I couldn't think of an easy way of factoring these into the plot. -- Mark

Thanks Mark! A super idea, you can see a band that covers almost all of the results, I took the liberty of painting it over your graph. Results much above the bar need work, the two below the bar are very fast for their CPU speed.

Took the analysis one step further... Actually 3 of the 4 way above the bar don't need work per se, they are predictable. The slowest 2 are 256MB memory, the only 2 so configured, the next slowest is the OSX one (clearly an apples-vs-oranges comparison). The only one "off" is then the next which is a P3 933, which as the only P3 left in the set (the other was dropped as being > 650 secs) can't really be characterized. Of the 2 fast outliers, I can find no details behind the AMD 1.4 @ 234 secs (it appears first in Jim's "Results So Far.." post so perhaps it's been deleted/withdrawn?). On the AMD 2.1 at 192 are we seeing the AMD vs P4 clock speed issue? It looks to me as if most current AMD's have been tabulated at the speed indicated by the name - e.g. 2600+ as 2.6. This would make sense as the number is how AMD equates to Intel's current P4 somewhat-inflated GHz numbers (which assume that the big P4 pipeline is fully occupied - rare even with newest software, certainly not an old Mac app converted as a WIn 3.1 application ;-). So if this particular point is converted from actual GHz to P4-equivalent GHz that makes it a 2600, putting it as a close outlier to the grey band. Probably close enough for Jim to have included it when he created the band. Final interesting nugget, the P4 2.8 @ 253 outlier on the slow side is the only system at 512MB. So 512MB is the realistic minimum (with an NT-based OS) for this scene. Having less than 1GB does hurt a little but not much. Probably means the sweetspot for this .pz3 was something like 640MB? So overall a very tight spread, considering all the variablities involved. A very useful experiment, but unfortunately one that ultimately breaks down to nothing more than the old racers' motto: "There's no substitute for horsepower"!

Good points! Gee, I remember it as "There's no substitute for cubic inches" (read GHZ & memory for computers!) ;-)

Only in the good ol' USA ;-) Cubic inches to litres was always a tough mental math job!