Testing real-life desktop hard drive performance isn't rocket science. You simply copy, move and delete files while timing them by some means. I did this for years with a stopwatch and the results invariably wound up reaching the same conclusion as professional benchmarks. But the stopwatch method is tedious at best, and generally a gigantic pain. Thankfully, No Design software has thought to automate the process, or at least part of it, with their DiskBench utility.
You don't really install DiskBench. You simply extract it from the zip file to the location you wish to run it from. This would make it lightweight and portable, except that the program requires Microsoft's .NET. If .NET doesn't exist on the PC you're testing, you must download anywhere from 187MB for the required version 2 to about 500MB for the latest version 3.5. For a program that only requires 158KB of disk space on its own, that's a lot. Many users will have .Net already installed; however I'd still love to see the company switch to something such as Pure Basic to create a less disk-space-hungry redistributable.
Once unzipped, DiskBench works well within its limits. There's a Copy File test which as you might guess, copies a file; a Create File test which creates two files simultaneously using separate threads; and a Read File test which loads two files into memory at the same time. One more test, Create Batch, creates multiple files, one file at a time. Results for all the tests are displayed in the info window at the bottom of the program as both time and throughput (e.g., 32 seconds, 10MB/s). Oddly, while time and throughput are displayed for each file in the Create Batch test, no aggregate score is computed. This is a big minus if you're trying to create a large number of small files which return a useless 0 milliseconds, infinite MB/s timed individually.
Tip: If you don't want to go through the hassle of locating files of different sizes to compare, the Create File test is also handy for creating test files. Simply deselect the "Kill destination file" option and create them in a suitable location. You can even change block size. What would be handier would be an option to automate this.
There are two omissions that prevent DiskBench from being as useful as it could be. One is a small file batch copy test. While the Copy File test is good for assessing sustained throughput, it doesn't test random access/seek speed (how fast the read head can get to the start of a file) as copying a large number of smaller files would. The Create Batch test could conceivably be used to perform this function, but lacking an aggregate score it tells you nothing. The other thing missing is a log file. If you want to compile a database of results you'll have to enter them all by hand or copy them from the program's info window to paste into your own log or a spreadsheet.
Hopefully, No Design software will address DiskBench's shortcomings. Even as is, if you don't mind or have already lost space to .NET, it's a useful way to get a handle on your hard drive's relative performance.
Note: This link takes you to the vendor's site, where you can download the latest version of the software.
--Jon L. Jacobi
Ad Choices