For a LAPP setup what is better: 1 fast or 2 slower machines

Hello,

I'm running a CentOS 5.6 / 64 bit Linux
with PostgreSQL 8.4.8 and Drupal 7.4
on a 4GB quad-CPU machine.

In Autumn I can change my hoster
and for EUR 100 can either take

1) 2 machines with i7-920 Quad-Core
8 GB RAM, 2 x 750 GB SATA-II HDD
(Software-RAID 1)

or

2) 1 machine i7-980X Hexa-Core
24 GB RAM, 1 x 1,5 TB SATA II

What would you prefer?

I know, the usual answer is
"you don't provide enough details",
but would it really help anyone if I provide
all the details and print here all my scripts?

My current setup - postgresql.conf:
max_connections = 50
shared_buffers = 1024MB
and unix pipe only (for 2 machines I will
have to switch to TCP at 100 Mb/s switch).

I use pgbouncer with

pool_mode = session
server_reset_query = DISCARD ALL;
server_check_delay = 10
max_client_conn = 200
default_pool_size = 16

and at peak times it reports:

70 req/s, in 11616 b/s, out 784748 b/s,query 89556 us
34 req/s, in 5484 b/s, out 176943 b/s,query 60871 us
45 req/s, in 17976 b/s, out 169849 b/s,query 66496 us
53 req/s, in 10961 b/s, out 648585 b/s,query 40546 us

My httpd.conf:

<IfModule prefork.c>
StartServers 8
MinSpareServers 5
MaxSpareServers 20
ServerLimit 120
MaxClients 120
MaxRequestsPerChild 4000
</IfModule>

Thank you
Alex