Actually, the speed difference is significant. For instance, on a 14 disk 3.4 TB XServe RAID(RAID 50) I measured throughput at 365MB/s, whereas with my 10 disk 1.8TB array(RAID 0) I measured throughput at 480MB/s.
You may remark that 365MB/s is way fast enough for HD, but bear in mind that performance decreases markedly after the drive becomes over half full.
However, you’re 100% right: A RAID array without fault tolerance is foolhardy, to say the least. If the OEM option is out of one’s price range, I’d recommend two RAID 5 PCI 8 port cards with ‘hot swap’ and ‘hot spare’ capability. Populate both cards fully, stripe each with RAID 5 including one hot spare on each. Then, either in MacOSX Disk utility, or in the cards’ RAID Admin utility(if it’s supported), stripe across both arrays with RAID 0, effectively resulting in a RAID 50 array. Performance should be more than satisfying.
Bear in mind that, at present, the above is not yet possible in the newest(PCIe) G5’s.