Test Results with Single Disks:
Looking at the WinBench 99 disk transfer rates we can see that the IBM 120GXP and the Western Digital 1000JB are performing close together: Both are starting with an excellent transfer rate at around 45 MB/second in the outer zones. The 120GXP is transferring slightly more at the beginning than the 1000JB but dropping more towards the end of the partition. The D740X is not only starting lower with just over 41 MB/sec but also soon starts dropping transfer rates. The difference between the beginning and the end transfer rate is more than 40%. The 120GXP is only dropping 10% and the 1000JB remains almost flat. The "flatter" the transfer rates the better, because the range of high performing disk space becomes wider. A drop in transfer rates is unavoidable to a certain degree because the outer zones of a disk are rotating faster and providing therefore higher transfer rates than the slower rotating inner zones. All 3 disks returned access times, which are close, but always slightly higher (means worse) than the manufacturers' ratings. The D740X has a slightly lower disk access time than the 120 GXP and the 1000JB. As we will later see, that does not translate into better performance when applications are running.

To demonstrate how much hard disk performance has improved in relatively little time, we show here the transfer rate of one our ATA 33 "museum" disks produced in September 1998. This is just 3 and half years ago. The performance has more than tripled since then. Click images to enlarge:

IBM 120GXP	Western Digital 1000JB
Maxtor D740X	ATA33 Reference

The WinBench 99 "High-end Disk" sub-tests show the IBM 120GXP and the Western Digital 1000JB neck-to-neck with the 120GXP slightly ahead. The application subtests in WinBench 99 have quite a fluctuation between single test runs and differences as small as we can see them here between the IBM and the Western Digital disks are irrelevant. The Maxtor comes in third, clearly better than our ATA 100 reference disk, but again behind the 120GXP and the 1000JB. The SANDRA 2002 "File Benchmark" confirms the picture. Both the 120GXP and the 1000JB are close together delivering scores far ahead of the SANDRA 2002 ATA 100 reference score. The Maxtor again comes in third.

click to enlarge

  
Sandra File System Benchmark - Maxtor, IBM, Westen Digital - Click images to enlarge

Content Creation 2001 shows a slightly different picture. As already mentioned CC 2001 measures overall system performance. In our tests only the hard disks drives are different and all other parameters are identical between the runs. This means we can attribute the performance difference as shown by the CC 2001 scores solely to the performance of the tested hard disk drives. It seems that when running the demanding CC 2001 applications the 8 MB cache puts the 1000JB slightly ahead of the 120GXP. The more often the test applications have to access the disk, the higher the impact of the large 1000JB 8 MB disk cache/buffer, because the need for physical disk access is lower than with a 2 MB cache. Not surprisingly we see the D740X again behind the 1000JB and the 120GXP.