AtomicInternet Archived News

Before the LAN Party, I was under the impression that RAID disk controllers were only useful for redundancy and stability. That changed after the rantings of Zorka regarding his new Serial ATA RAID0 array. RAID0 stripes across multiple drives and actually load balances the work between them, resulting in a performance boost. After running a base benchmark I discovered my single WD1200JB drive was half the speed of his array. Like an annoying mosquito the potential gains I could achieve from a RAID0 setup kept buzzing in my ear. I finally gave in and bought an identical WD1200JB drive, which has an 8MB cache and decent specifications (and is affordable) but nowhere near the speed realized with a faster Serial ATA drive. Still excited, I setup both drives as master on separate IDE controllers. I debated using the same controller with a master/slave configuration but using my experience from single drives decided separate controllers would be optimal (if you know something I don't please email me!). The GA-7N400 Pro2 motherboard in my computer has a built in ITE 8212 GigaRAID controller that was just waiting to be configured, which I did. The setup was simple, and wiped my existing drive clean (yes it was backed up silly!). After a little research I found the "Mass Storage Driver" required for WinXP (why do we still need to load 3rd party drivers from floppies?) and reloaded the computer with the new improved RAID0 array. While the load seemed somewhat faster, nothing spectacular occurred. However, the new benchmark results was definitely an anticlimax. The new setup gave me an extra 20MB/s average read speed (significant) but actually went down in burst and random reads. However the graph does show a more sustained rate for the RAID as compared to the single, which is hopefully an indication of better archive decompression speeds (my primary need for speed). While the RAID0 now means I'm more vulnerable if a drive failure occurs (complete data loss instead of a bad block) the cool factor and marginal improvement are enough to justify the risk. Yes I know what you want now, so here it is: an animated benchmark comparison. I used HD Tach for the benchmarks, send me yours so I can gloat or cry accordingly.