Typical Apple benchmarketing.
Apple's own compile of BLAST is optimized for G4 and OSX (won't run on
anything else). Others run non-optimized NCBI-BLAST that'll run on
anything. Even their own hype lets on that NCBI-BLAST isn't optimized for
anything but word length 11.
"only at a word length of 11, for which NCBI BLAST has been extensively
optimized, do the IBM and Sun systems perform comparably to Xserve"
What they leave out is that Apple's recompile uses THEIR altivec compiler.
The IBM and Sun versions don't get to take advantage of those chip's SIMD
hardware. It's like racing comparable cars, all equiped with nitrous
injection and only one allowed to use it.
As for the Web Bench marketing, Windows Server 2003 beats them all.
For this test an HP ProLiant DL380 G2 server configured with two 1.4GHz
Pentium III processors (note that those were using 1.4GHz P3 processors not
server class Xeons) and Windows served 17,000 requests per second vs.
Windows 2003 is over three times faster than Xserve when serving static
content even when run on previous generation hardware.
HP ProLiant DL380 G2 server configured with two 1.4GHz Pentium III