Updated Hard Drive Benchmarks

[Valimus]

Why are the numbers so inflated from the last time I looked at them maybe a week ago?

[David (PassMark)]

In conjunction with the new CPU Charts and the release of PerformanceTest V8 we have updated the HDD benchmark charts.

The main changes from the old charts are,

• The duration of the standard disk tests have been extended. It was observed that some drives are slow to start up. Our guess is that these drives were sitting in a low power state when the initial I/O request arrives. Or maybe it takes a while for the I/O queue to build up to an optimal level. Whatever the reason the first 1 or 2 seconds of disk activity is way slower than the top speed. This was especially noticed with solid state drives. So on my Intel SSD for example the fist 2 seconds run at ~100MB/sec. Then the drive ramps up to 200MB/sec. So in a short test this first two seconds was pulling down the average. In PerformanceTest V8, the tests now run for a longer period producing a higher average result and be more consistent.
• Larger block sizes have been used. Many drives have very bad performance if you perform I/O operations that contain a small amount of data. In PT8 the block size is larger. Again producing higher results.
• The data written to the disk has a controlled level of compress-ability. In PT7 the data values written were undefined (as it depended on the existing disk content). In PT8 the data written is predetermined. This should produce more consistent results.

So the results in PT8 will be higher than PT7 for some drives. They should also come closer to matching the benchmark numbers claimed by the HDD vendors (who unsurprisingly publish the biggest number they can find, no matter how unrealistic the scenario might be).

The reality however is the hard drives perform very differently in different test scenarios and one number is never going to reflect the drives performance under all circumstances.

[Valimus]

Oh, that's great to know!

Were all the drives retested with the new PT8 or will it be an ongoing process for the next few months until everything normalizes again or will only new SSDs that come out from now on be tested on the new PT8 for the sake of avoiding such a tedious task of such a thing

Thanks!

[David (PassMark)]

We have combined the V7 and V8 results in the same way as we did for the new CPU Charts.

So as of today there is a mix of results in the chart. Over the next few weeks PT8 results will replace the PT7 results as new results come in.

At the moment there are 4300 baseline results for PT8. Should be over 10,000 sets of results by the end of the month.