The previous two pages showed very mixed results. Now comes the ultimate test: using real applications. This is normally not done because it takes a lot of time to setup and perform the benchmarks in a consistent manner. But since the application benchmarks seemed to conflict the other benchmarks i felt required to take this step. I had to know what performance people would get with common tasks such as copying files and extracting archives - both in which the hard drive is a 100% bottleneck.
WinRAR compression
First, we will compress a large amount of small files into an archive using WinRAR. We used the store method which means no actual compression is performed, all files and directories are simply contained in one archive. This does require both reading and writing, causing the drives to seek a lot - therefore being an obvious bottleneck.
Interpretation: both drives performed equally in this task. The difference is negligible.
WinRAR extraction
Next, we extracted the just created archive with over 100,000 files in it to the same drive. A test where the mechanical limits of the hard drive are explored. Intelligent firmware and filesystem optimizations can accelerate this process, though. Again, we tested the time it takes to extract the entire archive 5 times, then took the average time.
Interpretation: kind of to my surprise, the AAE disk leaded this benchmark. Up until now, I suspected the AAK disk to perform better at non-sequential I/O while the AAE disk performed better at sequential I/O such as copying tasks. This benchmark disproves this theory. I re-ran the benchmark up until 8 times just to be certain. No, the AAE disk really is almost half a minute faster, and this is real waiting time!
Explorer recopy
This benchmark is an interesting one. We put a 10GB file on an empty formatted disk and make a copy of the 10GB file. This means explorer will read blocks, then write again. This type of I/O is also called "recopy" or "rewrite" and is pretty I/O intensive. The windows filesystem cache played a negligible role because of the large filesize. This one really comes down to the disk itself. Firmware can play an important role here by using read-ahead and write-buffering intelligently.
Interpretation: again the AAE disk clearly leads this benchmark, gaining a noticeable advantage.
Sequential file copy
In previous pages we did copy tests, but not in Windows and not using a stopwatch. This time we will measure the time it takes to copy a 1950MiB file from a ramdisk to our disk and back again. The ramdisk, ofcourse, is many times faster than the hard drive itself. The results confirm earlier findings:
Interpretation: just like earlier sequential copy-tests the AAE disk gains a clear advantage when reading, though the advantage is smaller when writing, just like before. It is safe to say the AAE disk is superior at any copy test.
WinRAR compression
First, we will compress a large amount of small files into an archive using WinRAR. We used the store method which means no actual compression is performed, all files and directories are simply contained in one archive. This does require both reading and writing, causing the drives to seek a lot - therefore being an obvious bottleneck.
| WinRAR compression (time to completion) | |
|---|---|
| AAE |    4:50 |
| AAK |    4:49 |
| Benchmark interpretation: lower is better | |
Interpretation: both drives performed equally in this task. The difference is negligible.
WinRAR extraction
Next, we extracted the just created archive with over 100,000 files in it to the same drive. A test where the mechanical limits of the hard drive are explored. Intelligent firmware and filesystem optimizations can accelerate this process, though. Again, we tested the time it takes to extract the entire archive 5 times, then took the average time.
| WinRAR extraction (time to completion) | |
|---|---|
| AAE | 3:35 |
| AAK | 4:02 |
| Benchmark interpretation: lower is better | |
Interpretation: kind of to my surprise, the AAE disk leaded this benchmark. Up until now, I suspected the AAK disk to perform better at non-sequential I/O while the AAE disk performed better at sequential I/O such as copying tasks. This benchmark disproves this theory. I re-ran the benchmark up until 8 times just to be certain. No, the AAE disk really is almost half a minute faster, and this is real waiting time!
Explorer recopy
This benchmark is an interesting one. We put a 10GB file on an empty formatted disk and make a copy of the 10GB file. This means explorer will read blocks, then write again. This type of I/O is also called "recopy" or "rewrite" and is pretty I/O intensive. The windows filesystem cache played a negligible role because of the large filesize. This one really comes down to the disk itself. Firmware can play an important role here by using read-ahead and write-buffering intelligently.
| 10GB file recopy (time to completion) | |
|---|---|
| AAE | 5:37 |
| AAK | 6:16 |
| Benchmark interpretation: lower is better | |
Interpretation: again the AAE disk clearly leads this benchmark, gaining a noticeable advantage.
Sequential file copy
In previous pages we did copy tests, but not in Windows and not using a stopwatch. This time we will measure the time it takes to copy a 1950MiB file from a ramdisk to our disk and back again. The ramdisk, ofcourse, is many times faster than the hard drive itself. The results confirm earlier findings:
| Sequential read (seconds to completion) | ||
|---|---|---|
| AAE | Read | 29.8 |
| AAK | Read | 41.7 |
| Sequential write (seconds to completion) | ||
| AAE | Write | 26.2 |
| AAK | Write | 28.2 |
| Benchmark interpretation: lower is better | ||
Interpretation: just like earlier sequential copy-tests the AAE disk gains a clear advantage when reading, though the advantage is smaller when writing, just like before. It is safe to say the AAE disk is superior at any copy test.
execution time: 168 msec, queries: 18
