Originally Posted by ZachPruckowski
Is software RAID-0 as fast as hardware (or nearly?) Will I see a performance benefit from 2 RAID-0'd 160GB HDDs versus a 250 GB HDD, assuming all are 8-16MB cache and 7200rpm? Because I read somewhere that the harddrive could be a potential Mac Pro bottleneck (along with RAM latency)
In my G5 DualCore 2.0 GHz, I have a RAID-0 with two (not even identical) 7200rpm SATA drives (8 MB of cache each). Read and write speed for the RAID-0 is about 100 Megabytes per second (measured with Helios LAN Test 3.1 on the local harddrive -- which gives quite accurate results when using 300 MB test files, saturating both the two SATA buses and all drive caches). One single, not RAID-ed drive goes up to about 60 MB/sec, depending on the drive.
Modern SATA-II buses have a theoretical bandwidth of 3.0 Gbit/s (around 300 MB/sec effective speed) and most 7200rpm harddrives today use up only about one fifth of that bandwidth. So harddrives can indeed be some kind of bottleneck. Using 10K or 15K drives is better and faster, but they run louder and a lot hotter. You might get over 80 MB/sec per drive using those.
Hardware RAID is always faster because of the deditated processing units. Maybe in RAID-0 configurations, the speed difference is not so big, but when using RAID-5 where a lot of parity information needs to be computed, the difference is huge! Servers almost always use hardware RAIDs for that reason.
One single 250 GB drive, however, is for sure a lot slower than the two-drive 160 GB RAID-0.