I build SAGE and all components from scratch on fermat and modular.
The output of time are as follows.

FERMAT (Dual G5 OS X):
user    8m41.610s   (but see below!)

MODULAR (Dual Opteron 248):
user    3m56.356s

THINKPAD T42p (Pentium-M 1.8Ghz):
user    5m30.516s

Given the huge difference in time, and that I just typed "make" in
both cases on two-processor machines, I tried "make -j5" on both
machines.  The modular timing was the same, but the fermat timing was
much better.

FERMAT (G5): 
user 4m35.170s

Thus make on modular probably by default uses all available
processors, but make on fermat does not.


I also have a program that tests all examples in the reference manual.
The timings on Fermat and Modular are as follows:

FERMAT (G5):
user    0m5.520s

MODULAR (Opteron 248):
user    0m3.154s

THINKPAD T42p (Pentium-M 1.8Ghz):
user    0m3.393s 

Parallel make is irrelevant for this, since the makefile just runs a
single Python script that does all the tests. I ran these timings
several times and the timings were always about as indicated, so I
don't think disk is very relevant.