Testbed and Methods
We used the following test programs:
- FC-Test version 1.0
- FC-Test version 2.0
- IOMeter version 2003.02.15
The testbed was configured like follows:
- Albatron PX865PE Pro mainboard
- Intel Pentium 4 2.4GHz
- IBM DTLA-307015 15GB hard disk drive
- Radeon 7000 32MB graphics card
- 256MB DDR SDRAM
- Microsoft Windows XP SP2
This time we’ll only be benchmarking the flash drives with IOMeter and two versions of FC-Test. The difference between the two versions is in a slightly different test algorithm that ensures more stable and accurate results.
We won’t publish the results of the synthetic AIDA benchmark whose main feature was that it produced diagrams ready to be published. Since the test conditions have been changed considerably, we won’t compare the drives with those that we have tested earlier in our labs. However, we retested the OCZ Rally and the A-DATA My Flash RB15 because the former was once a leader in performance and the latter provoked some questions from users who reported that the 256MB version of the A-DATA RB15 that we had tested in our labs differed quite greatly in speed from the 1GB version.
Intel IOMeter: Sequential Read & Write
We’ll start out with the synthetic IOMeter benchmark. The first pattern measures the sequential read and write speed of the drive on data blocks of different size (from 0.5 to 1024KB).
The first diagram shows the sequential read speed of the drives. Some graphs almost merge into one which is indicative of the high accuracy of the benchmark as well as of the fact that different manufacturers may use the same flash memory type and controller in their products. We’ve got two such pairs here: the Kingston Data Traveler II Plus Migo Edition and the Apacer Handy Steno HT203, which are also the fastest in this test, and the Sony Micro Vault USM-EX Turbo with the OCZ Rally drive that perform quite well, too. The A-DATA My Flash RB15 and the Kingston Data Traveler Elite also deliver high performance while the Transcend JetFlash TS1GJF210 FingerPrint is the slowest device here.
The sequential write diagram shows that the Sony Micro Vault USM-EX Turbo is unrivalled in this parameter. The Apacer Handy Steno HT203 takes the second place and is followed by the Kingston Data Traveler II Plus Migo Edition. The Transcend JetFlash V30 has the lowest speed.
By the way, our using different data block sizes provides us with some thinking matter. You can note that in some instances a drive’s performance increases in more than two times when the data block is doubled in size. Perhaps we hit at the request size that is the maximum the controller can process at once? This information can be used to try to maximize a drive’s performance by changing the cluster size at formatting.