Testbed and Methods
Our testbed configuration looked as follows:
- ASUS P3B-F mainboard;
- Intel Pentium III (Coppermine) 600MHz CPU;
- 2 x 128MB PC100 ECC SDRAM by Hyundai;
- IBM/Hitachi DPTA 372050 host HDD;
- Matrox Millennium 4MB graphics card;
- Promise Ultra100 TX2 and Promise SATA 150 TX2 controllers;
- Windows 2000 Professional SP4.
We used the following benchmarking software:
All drives that support “quiet seek / normal seek” operation modes were switched to the fast mode by means of Hitachi Feature Tool. For WinBench tests, the drives were formatted in FAT32 and NTFS as a single partition with a default cluster size. We used Paragon Partition Manager for FAT32 formatting. The benchmarks were run seven times each; the maximum result was taken for further analysis. The HDDs didn’t cool down between the tests. The tests in Intel IOMeter were run in SequentialRead, SequentialWrite, DataBase, WorkStation, FileServer and WebServer patterns. If you are looking for the detailed description of these patterns, please, see our previous articles.
Performance: Access Time and Read/Write Speed Benchmarks
Let’s start with the easiest things: the ability to read data sequences. The funny thing is that even this unsophisticated test
Prepared a few surprises for us. Have a look here:
The first thing that catches your eye is the significant lag of Barracuda 7200.7 Serial ATA when it works with small data blocks. Interestingly enough, but Barracuda Serial ATA V didn’t have any problems like that and copes with thie type of read tasks not any worse than its ATA fellows.
The second surprise is the slower performance of 160GB ATA models compared with that of HDDs with lower storage capacity. They run almost evenly fast up to 4KB data blocks and then the elder models with ATA interface slow down (this graph is marked in red on the diagram above) while all the bothers retain the same pace. The only explanation I can think of here is the slower performance in case of 48bit addressing. Actually, Barracuda 7200.7 Serial ATA doesn’t have any problems like that, and the models with different storage capacity perform almost equally fast here.