Several older drives are included in the testing for perspective. Note the huge increases in buffered read speeds gained by the addition of larger cache memory to the drives. These increase from 5 MB/s with 128KB of cache to 114 MB/s with 8MB of cache. The Hitachi SATA-150 drive had a higher buffered read speed than the comparable Western Digital ATA-100, perhaps due to the higher bandwidth of SATA.

The FastTrak controller was installed into a 32-bit 33Mhz PCI slot, whose maximum bandwidth is 133 MB/s. For this test, only an inactive modem was sharing the PCI bus. The available bandwidth seems to be sufficient for the all the performance metrics except the 111 MB/s buffered read and 79 MB/s sequential read, which may be hampered by the peak 133 MB/s limit. To determine if the 33Mhz PCI slot is a bottleneck would require a benchmark against a 66Mhz PCI slot, which was unavailable to the reviewer.

The random read performance of the RAID-5 was 16 MB/s versus 6 MB/s for the comparable stand-alone hard drives. The RAID-5 array does better than the stand-alone disks primarily because of the disk striping. The very high random read performance explains why RAID-5 is the preferred configuration for many types of servers. Note that for these test, the Sandra benchmark is using a one gigabyte test file because the test system has one gigabyte of memory.

While the read performance is high, the write speeds of the RAID-5 array are either slower or equal to the stand-alone drives. When compared to a RAID-0 array of the same number of disks, the RAID-5 write performance would be categorically slower. Hence, RAID-0 is better for large image or video file editing.

Too much should not be read into these benchmarks. Performance is the complex interplay of many different factors, and the above tests were not performed in a controlled environment and are not scientific. However, they can be used for comparison with your own benchmarks, and to suggest the relative merits of different drive configurations.