What Effect Does Short Stroking Have On Performance?

1. Short stroking a hard disk drive will not make it "run faster". Physically, there are no changes - the drive still spins at the same spindle speed, with the same number of platters and read/write heads.

2. However, short stroking a hard disk drive will significantly improve its random seek time, by restricting the head movements to a smaller number of tracks.

3. Short stroking a hard disk drive will also significantly improve large sequential reads and writes, but only because it restricts the drive heads to the outer (fastest) tracks of the hard disk drive.

4. On the other hand, short stroking a hard disk drive will not improve its sequential seek time, as the distance between each bit on the platter remains the same.

Now, we are going to prove all that by actually testing it out. However, we must first remind you that hard disk drive benchmarks have certain limitations - they cannot actually simulate a real-world situation.

For example, when we test random accesses, the benchmark would test the entire platter even though in a real world situation, the drive heads would have a much more restricted range of movement. Benchmarks also cannot simulate the effect of defragmentation (and the moving of frequently-used data to the outermost tracks), so we cannot test the performance effect of drive defragmentation.

To demonstrate the perception skewing effect of the LBA mod versus other methods on actual hard disk drive performance, we chose to short stroke the 2 TB Western Digital Caviar Black and compare it to the 600 GB Western Digital VelociRaptor. The reason? Both drive start with almost the same transfer rate on the outermost tracks - the VelociRaptor at 153 MB/s and the Caviar Black at 146 MB/s.

The Testbed

Testing Methodology

We tested in both Windows 7 and Windows Vista, with the latest updates. We chose to use IO Meter as well as our "old faithful", WinBench 99 2.0, with the following tests :

• Platter Data Transfer Profile
• Business Disk WinMark 99
• High-End Disk WinMark 99
• Disk Transfer Rate (Beginning)
• Disk Transfer Rate (End)

Business Disk WinMark 99 is a real-world simulation based on three office application suites - Microsoft Office 97, Lotus SmartSuite and Corel WordPerfect Suite 8, as well as a web browser, Netscape Navigator. They are quite dated, but should still reflect the usage patterns of users in an office environment using such applications. The test runs through a script that keeps multiple applications open, while it performs tasks that switches between those applications and Netscape Navigator. The result is the average transfer rate during the script run.

High-End Disk WinMark 99 is a real-world simulation based on AVS/Express 3.4, FrontPage 98, MicroStation SE, Photoshop 4.0, Premiere 4.2, Sound Forge 4.0 and Visual C++ 5.0. However, it differs by running the applications serially, instead of simultaneously. There are individual results for each application but in this comparison, we will be looking only at the weighted average score, which is the average transfer rate during the tests.

Unfortunately, WinBench 99 seemed to have some issues with Windows Vista, or the current Vista driver for the SATA controller. It would register a SetFilePointer error in the Disk Access Time test. So, we were not able to obtain any Disk Access Time results.

In addition, it would keep crashing if the hard drive was tested with a single partition. This is likely due to a limit on the size of the partition that is supported by WinBench 99. We came up with a workaround by dividing the hard drive into 3 partitions of similar sizes. We tested each partition individually and averaged the results.

Support Tech ARP!

Help support our work by visiting our sponsors, participate in the Tech ARP Forums, or even donate to our fund. Any help you can render is greatly appreciated!