Well, it's faster...

[Chip Morningstar -- "Software Without Moving Parts" <chip@xxxxxxxx>]

Just for giggles, I timed a complete recompile of the Amix backend on
Grand-Central and on the new 4/490, Victoria.  I tried it two ways on Victoria,
once using my files on Grand-Central over NFS and once using a copy of my
Working set on Victoria's local disk.  Here are the stats returned by 'time':

Grand-Central	193.4u 57.1s 6:23 65% 0+584k 467+1854io 389pf+0w
Victoria (NFS)	 85.9u 16.1s 3:29 48% 0+592k 209+1789io 215pf+0w
Victoria (disk)	 85.6u 14.8s 2:31 66% 0+600k 286+1807io 330pf+0w

For those of you unused to reading this stuff, the first number is user CPU
seconds, the second is system CPU sections, the third is elapsed clock time and
the fourth number is the percentage of the CPU that you got while doing this.
The remaining numbers relate to IO activity are are pretty much useless.  The
bottom line from these numbers is that the 4/490 is a little more than twice as
fast as the 4/280, at least for compiles.  I don't know what that translates
into in terms of number of simulatneous Amix users, but historically I've found
that compiles make good general-purpose benchmarks.