Most drive performance benchmarks either test real file operations (like USBDeview) or some kind of raw performance (like HD Speed). In either case it is often omitted that performance can vary greatly depending on size of data chunks involved.
Roadkil Disk Speed performs linear and random test for array of varied size blocks.
What it does
Disk Speed is quite straightforward to use – you choose volume to test (can be logical or physical) and launch.
Interface fills with results for blocks of various sizes and at the end overall score is calculated. No formula for score given, but with few parameters tested it is suitable for easy comparison between drives.
While benchmarks commonly answer how fast, this one answer how slow. Because of how data organized on hard drives they can be much slower when operating small data blocks and users are often unaware of this limitation.
App notably lacks any kind of write test. Those are usually destructive so this makes it safer to use and recommend, but more boring as well.
There is an option to upload result to developer’s site for comparison. That one gives few details and seems greatly skimmed by someone’s max result of over 5GB/s read (probably had serious hardware to play with).
Clearly not comprehensive, but shows performance at different sizes and is quite unique at that. Single-executable and natively portable.