Testbed and Methods
The following testing utilities were used:
- IOMeter 2003.02.15
- Intel NAS Performance Toolkit 1.7
- FC-Test 1.0
- PCMark Vantage
- Windows 7 Disk Defragmenter
- WinRAR 3.91
- ASUS P5WDG2 WS Pro mainboard
- Intel Pentium 4 620 processor
- IBM DTLA-307015 system disk, 15 GB
- Radeon X600 graphics card
- 1 GB DDR2-800 SDRAM
- Microsoft Windows 7 Ultimate
HDDs are tested with generic OS drivers. We format them as one NTFS partition with the default cluster size (for FC-Test we create 32GB partitions), connect them to a mainboard port and enable AHCI. I would like to remind you that we are now using new HDD testing methodology.
Performance in Intel IOMeter
Sequential Read and Write Patterns
IOMeter is sending a stream of read and write requests with a request queue depth of 4. The size of the requested data block is changed each minute, so that we could see the dependence of a disk’s sequential read/write speed on the size of the data block. This test is indicative of the maximum speed a hard disk can achieve.
Rather surprisingly, the Barracuda LP ST1000DL002 with its 5900 RPM comes out on top in this test. This is due to its higher recording density thanks to 4KB sectors.
Then, we can note that the HDDs split up into two groups when processing small-size data blocks: the Seagate ones are twice as fast as their WD opponents at data chunks up to 2 KB. The WD drives catch up with the Seagate team at 8KB data blocks and reach their top speeds at 16B data blocks only. The RE4 1003FBYX is the only exception as it delivers its maximum speed as soon as 8KB blocks.
The rather low performance of the Barracuda LP ST31000520AS and Caviar Blue WD10EALX should also be noted here.
The sequential write results do not make up such a nice-looking diagram as we've just seen in the test of sequential reading. The Barracuda 7200.12 ST31000528AS slows down while the Barracuda 7200.12 ST31000524AS is good and the Barracuda LP ST1000DL002, just excellent. The WD drives are slow until 32KB data blocks, the Raid Edition 4 model again accelerating to its top speed sooner than the others.
Disk Response Time
For 10 minutes IOMeter is sending a stream of requests to read and write 512-byte data blocks with a request queue of 1. The total of requests processed by each HDD is much larger than its cache, so we get a sustained response time that doesn’t depend on the HDD’s buffer size.
The HDDs are all close to each other in terms of read response, even the 5900RPM models. The WD drives are somewhat faster overall, though. When it comes to writing, the RE4 WD1003FBYX and the Caviar Black WD1002FAEX are the most responsive whereas the Barracuda 7200.12 ST31000524AS falls behind the main group. The Barracuda LP ST1000DL002 is downright slow due to its 4KB sectors.
What can we learn from these results? First, a lower response time is going to translate into higher performance at random-address operations. The Caviar Black WD1002FAEX and RE4 WD1003FBYX should be singled out in this respect: besides moving their heads around very quickly (which is indicated by their low read response time), they feature very efficient request reordering algorithms (indicated by the difference between the read and write response).
Second, the low result of the Barracuda LP ST1000DL002 is indicative of 4KB sectors. This HDD has to perform three operations to write a single 512-byte data block (read it, recalculate the sector’s checksum, and actually write it) whereas HDDs with 512-byte sectors can write 512-byte data blocks in one operation since the sector’s checksum doesn’t depend on the rest of the disk contents.
Third, a low response time has the downside of higher noise and heat. It means that the Seagate HDDs are likely to be quieter than the WD ones, although this can vary depending on the particular sample.
Random Read & Write Patterns
The results match those of the response time test. The Barracuda LP ST1000DL002 falls back to last place after being first at sequential reading. It is considerably slower than the next worst drive, which is the Barracuda LP ST31000520AS. This is due to the spindle rotation speed of 5900 RPM, of course. The main group is headed by the RE4 WD1003FBYX but the gaps are small there.
It’s more interesting at random writing: having failed the previous test, the ST1000DL002 also fails when writing random-address data blocks smaller than 4 KB (to do such a write, it has to read the data block, wait for the platter to make a full turn, and then perform the write operation proper). However, it is better than the rest of the Seagate HDDs with larger data blocks.
The rest of the results are roughly equivalent to those of the response time test. The Caviar Black WD1002FAEX and RE4 WD1003FBYX are ahead, despite the slump at 128KB data blocks. The other WD drives follow the leaders in a tight group and are joined by the Barracuda LP ST1000DL002. The rest of the Seagate drives go next, the Barracuda 7200.12 ST31000524AS having some problems with deferred writing as its speed isn’t much higher than at random reading.
The Barracuda LP ST1000DL002 has a peak of performance at 1KB data blocks, which must be due to the combination of Smart Align technology with a cache line size of 1 KB.