Index Creation is slow after server change

I am setting up a server to replace an existing one.  I am benchmarking
a batch import process that needs to be run nightly, which does some
bulk "COPY" operations, and creates new tables from new data (INSERT
INTO table (fields) SELECT...), and recreates indexes.  Currently, I can
do the import I am benchmarking on the "old" machine in 20-30 minutes
depending on load from other processes.  On the new machine, which is
dedicated to the benchmarking right now, the same task takes 50 minutes,
took me two takes of disk and RAID tweaking to get down from 60 minutes.

Based on my seat of the pants timing watching the two, index creation is
much slower on the Xeon machine, which does not make sense to me.  I can
do some proper benchmarks if nobody has any ideas, so I know exactly
where I stand.

Any ideas why this would take almost 3 times as long on what, other than
166Mhz per CPU, is a much more heavy duty server?  I have more memory
available for shared buffers than postgres is using on the import (1.75g
allowed, its using about 1.1g max).  I only allow postgres on the old
machine 808mb, and it uses it all.  I currently have 4000 log buffers,
with 8 log files created.

I have been fighting this for the last 5 days...something just isn't right.

  The following are the configurations:

Old Server:
Dell 2450, dual 667 PIII
1 gig RAM
2 - 18.x gig system disks in RAID1 configuration (18.x gig usable), ext2
formatted
8 - 18.x gig data drives in RAID-5 configuration, ext2 formatted
PERC2 RAID card, with probably 64MB of RAM
Postgres 7.2.1
RedHat 6.2 as OS

New Server
Dell 6300, dual 500 Xeon w/512k cache
2 gig RAM
3 18.x gig system disks in RAID0 configuration, ext3 formatted
3 36.x gig database drives in RAID0 configuration, ext formatted
Adaptac 2100S 32MB RAID card (older PERC2 with 128MB cache gave similar
results)
Postgres 7.2.1
RedHat 7.2 as OS

Thanks for any help,

-Pete