Comparing Apples and Penguins

by Scot Hacker

Moshe Bar at has performed a careful and thorough side-by-side benchmark/comparison of OS X as a pure server (not desktop) environment by installing both OS X and the SuSE PowerPC Linux distribution on an XServe. Short story: OS X lost to Linux on every measurable count, but Moshe is still impressed, since OS X is so new while Linux has had a decade to tune its subsystems. To my knowledge, this is the first such test done with this degree of care.


2002-10-28 16:57:17
Why not Jaguar?
I'd be curious why he used 10.1.5, rather than Jaguar (10.2), and how that might change the results of the test. Informal testing on my part---by no means as extensive as Moshe's---shows a 10-15% increase in performance for web serving on Jaguar (non Server version). That would come close to equalizing the results.

Also, he said he used the CDs that came with it, so I'd assume he used 10.1.5 Server, not the desktop version. There's quite a few tuning differences.

2002-10-29 00:15:28
Also..... about "lack of profiling tools"
Look into /usr/bin, there are a bunch there off the top of my head: sc_usage, fs_usage, heap, leaks, gprof, sample, vmmap etc.
2002-10-29 00:42:04
the mach core to OSX has been around in one form or another for longer than Linux.

There undoubtedly improvements still to be made to OSX systems but there has been almost fifteen years for those to have been done already....

2002-10-30 19:22:07
I'd be really interested to see the results with os X 10.2 and gcc 3.x. From what I understand there is at least 10% performance gain and that get darn close to some of those linux numbers.

Jamie Orchard-Hays