Hard Drive Benchmark Test

The hard drive benchmark test is designed to test the read and write performance of a hard drive. In order to do this, we execute three separate sub-tests: Small File Read, Large File Read, and File Write. To compute the final hard drive throughput score, we then average the results of the three sub-tests.

Results are measured in MB/s (megabytes per second).

It is important to note that we test both read and write performance.
The drive being tested must allow write access and have at least 500 MB of free disk space.

Note: Some PCs have more than one hard drive. You can configure which hard drive is tested by selecting Tools – Options in the main menu.

Multi-Threading

Hard drives these days are rarely accessed by only one application at a time. Usually, hard drives are accessed by multiple threads and often need to queue up requests in order to respond to multiple simultaneous requests.

It is very important to measure performance using multiple threads. Dacris Benchmarks performs two multi-threaded tests: one for reading, and one for writing. These tests are given the most weight in the calculation of the final result.

If a hard drive has NCQ (native command queuing), it should score higher than a drive without NCQ, because NCQ is a technology that helps improve performance in multi-threaded scenarios.

RAID

Dacris Benchmarks also measures the impact of RAID on performance. With RAID-0 (striping), both read and write performance should be increased. With RAID-1 (mirroring), only read performance is increased.

Overall, RAID-0 or RAID-1 should increase the result of the hard drive test.

Buffering and Cache

Every hard drive has a small on-board buffer (usually 8 MB to 16 MB) that acts as a cache, so that the drive is bypassed for the most frequent requests. This buffer has a very minor effect on performance, since Windows already performs caching.

In Dacris Benchmarks, all caches and buffers are bypassed completely. We write directly to the drive, and read directly from the drive, with no intermediate caching. This gives a more accurate result representative of the performance of the actual physical drive.