The reason I consider this a low-end server rather than a beefy desktop is because of the disk configuration--the disk controller alone cost more than the motherboard/CPU/memory combination! The Areca ARC-1210 is run in JBOD mode, which it performs very well in, and it includes 256MB of write-back cache. This allows the kind of write acceleration needed for high-volume commits even with syncronous writes to the WAL.
Were this a production system, I would have added the matching battery backup for this disk controller, along with a monitored UPS and integrated server shutdown. That combination would make sure writes in cache were properly commited even in the case of unexpected server outage.
The three hard drives in this system are typical consumer-grade SATA models running at 7200 RPM. The disk dedicated to WAL use was partitioned with a 10 GB filesystem, because like most drives the first few logical GB of these drives are considerably faster than the rest of the disk.
The main deficiency of this system is its relatively small amount of RAM. The motherboard is also a very cheap design based on the nForce4 chipset. Unsurprisingly, its memory speed is on the low side compared with AMD's current Opteron processors or Intel's Core 2 Duo designs. But given a disk-bound situation, as you normally expect in database performance testing, the CPU is capable of saturing the hard drives with plenty of capacity to spare. In fact, proving that's the case is actually the goal of an early lesson in this tutorial series.