by Aleksey Meyev
12/05/2007 | 03:04 PM
You can’t have too much of storage space. Every PC or notebook user is going to tell you that. And while in a desktop PC you can increase the amount of storage by replacing your HDD with one that has a larger capacity or by adding one more HDD into your system, you can only go the former way with your notebook or small form-factor PC. Of course, you can use external HDDs with USB, FireWire or eSATA interfaces but they usually prove to be slower than HDDs connected directly to the mainboard via the already widespread SATA interface. Mobile HDDs may also prove unhandy as they mean one more device with additional weight and volume you have to deal with.
That’s why it was 2.5” HDDs that became the first drives to use perpendicular recording technology. You can refer to one of our earlier reviews for reasons why the manufacturers had to change the direction of recording (and also the material of the platters, the heads, and the electronics), see our article called Seagate Momentus 5400.3 160GB Hard Disk Drive with Perpendicular Recording Technology. The new technology provided the best effect possible for HDDs of this form-factor and all the makers eventually transferred their 2.5” HDD series to perpendicular recording. Increasing the amount of platters (like in 3.5” HDDs that can have up to five platters per drive) is not an answer for a 2.5” drive because you can’t fit a third platter into it without making it thicker. And a thicker drive won’t fit into the standard bay. Although such thick models, mostly from Fujitsu, do emerge on the market, they enjoy but a very limited demand.
So, the choice is made in favor of perpendicular recording. Having overcome the unavoidable technical problems and made contracts with the distributors, the developers all introduced new HDD series reaching a recording density of 80GB per platter in the first generation. By the way, the recently released second-generation models feature a recording density of 250GB per platter but that’s going to be the subject of our upcoming review. Today I’ll discuss six 2.5 HDDs with a capacity of 160GB, a spindle rotation speed of 5400rpm and a SATA interface.
Let’s take a look at the HDDs that are about to be tested. I won’t publish a table with their specifications since they are identical for all the drives: a spindle rotation speed of 5400rpm, a fluid dynamic bearing, 2 platters, 4 heads, an 8MB buffer, and a SATA interface with support for NCQ. Four out of the six HDDs have an identical average seek time of 12 milliseconds. Hitachi and Seagate declare a seek time of 11 milliseconds for their HDDs. I’ll measure the real average seek time of each drive later on. Right now I’ll tell you a few words about each of them.
The Fujitsu MHW2160BH is the most secretive model of this review. You can only learn its brief technical characteristics. Take note that most models from the MHW2xxxBH series use longitudinal recording (even the 120GB model), and only this HDD and some of MHW2080BH drives employ perpendicular recording.
Like most other modern HDDs from this firm, Hitachi’s HTS541616J9SA00 features such exclusive technologies as iridium-manganese-chromium heads, thermal fly-height control to minimize errors, parking on a ramp for protection during reboots, and the power-saving features HiVERT and ABLE.
The Samsung HM160JI belongs to the M80S series. It features Silent Seek technology to minimize noise by using a performance-optimized seek trajectory. Samsung claims that its version of such technology has a smaller impact on performance than the traditional Advanced Acoustic Management.
The Momentus 5400.3 series the Seagate ST9160821AS model belongs to was the first series of serially produced HDDs with perpendicular recording. We’ll see shortly if this series was much of a success.
Like the drive from Fujitsu, the Toshiba MK1637GSX doesn’t tell much about itself. I could only learn its basic specs. So, let’s better judge it by its performance in tests.
The Western Digital WD1600BEVS belongs to the ML80 branch of the Scorpio family. As opposed to most other HDDs in this review, it is not the largest-capacity model in its series. There exists the WD2500BEVS that has a 50% higher capacity (and a 50% higher recording density), but the 160GB models are still produced with two platters and four surfaces, and the WD1600BEVS has the same recording density as the other HDDs in this review. As usual, the HDD features a number of exclusive technologies: WhisperDrive technology uses SoftSeek algorithms to minimize noise when seeking; ShockGuard protects the drive’s mechanics and platters from damage; DuraStep Ramp parks the heads off the data zone for additional protection.
To wind up this introductory section, I want to show you a table with the power consumption parameters of the HDDs. We don’t yet have a means to measure it by ourselves, so these are the specified values:
You can see that the HDDs are not so similar after all. The Western Digital turns to be the most voracious device (unfortunately, its manufacturer doesn’t specify the spin-up consumption). The Seagate and Toshiba are the most economical at reading and writing while the Fujitsu and Toshiba require less power than the others in sleep mode.
The following testing utilities were used:
I installed the generic OS drivers for the drives and formatted them in FAT32 and NTFS as one partition with the default cluster size. For some tests 32GB partitions were created and formatted in FAT32 and NTFS with the default cluster size. The HDDs were attached to a Promise SATA300 controller that supported NCQ.
The HDD is receiving 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 you could see how the drive’s sequential read/write speed depends on the size of the data block. This test will show you the maximum speed the HDD can achieve.
The HDDs from Fujitsu and Hitachi boast the highest top speed, but the results of the drives from Toshiba and Seagate are even more exciting: they have a lower top speed than the leaders but achieve it on 4KB data blocks, being about 50% faster than their opponents then. These HDDs are also two times as fast as the others on 2KB data blocks. The Samsung performs poorly irrespective of the data block size.
It’s different at writing: the Fujitsu is now the sole leader in terms of top speed. The Toshiba couldn’t keep up the tempo of the Seagate on small data blocks and joined the group of HDDs with average performance. The Samsung has the lowest top speed while the Fujitsu, somewhat surprisingly, is the worst one on small data blocks.
In the Database pattern the HDD is processing a stream of requests to read and write 8KB random-address data blocks. The ratio of read to write requests is changing from 0% to 100% throughout the test.
The following diagrams show the dependence of speed on the ratio of reads to writes.
None of the HDDs benefits from NCQ technology. The HDDs from Western Digital and Toshiba provide the highest performance growth relative to load growth whereas the Hitachi and Fujitsu slow down when the shares of write and read requests are equal. The HDDs from Seagate and Samsung show themselves as very conservative with weak deferred writing. The Samsung is almost always below the level of 100 operations per second (except for a small range of loads at a requests queue depth of 256), showing non-aggressive look-ahead reading and a very strange behavior at 90% and 100% write requests. In the latter modes the number of operations processed by the HM160JI per second is not increasing as is the case with most HDDs (due to the deferred writing mechanisms) but diminishing. This drive seems to have non-standard firmware algorithms and you’ll see shortly if this only shows up in synthetic benchmarks such as IOMeter or in real-life tests as well.
Besides the firmware algorithms, the results of the Database test are affected by the disk access time. Let’s see how it differs between the HDDs. In this test IOMeter is sending a stream of requests to read and write 512-byte data blocks with a request queue of 1 for 10 minutes. The total number of requests processed by the HDD is over 60 thousand, so we get a sustained response time that doesn’t depend on the HDD’s buffer size. The results are sorted by read response time.
The Seagate is the best one at reading, enjoying a 1.21ms lead over the closest pursuer. That’s quite a difference. It is followed by three drives with similar results, namely the HDDs from Western Digital, Samsung (this partially explains the performance of this HDD in the previous test), and Fujitsu.
It’s all very different at writing: the Seagate turns a loser while the Toshiba and Western Digital go ahead.
It’s interesting to compare the ratio of response times at writing and reading which indicates how aggressive the HDD’s deferred writing algorithms are. The smaller the ratio, the more requests the HDD can store in its memory. This ratio is within 0.47 to 0.54 for every HDD except the Seagate whereas the ST9160821AS has a write/read response ratio of 0.65. This parameter equals 0.50 for the HM160JI which doesn’t quite agree with its Database diagram. This must be due to some peculiarity of this HDD’s firmware algorithms (the Database pattern operates with 8KB data blocks).
The HDD is being bombarded with requests to read and write random-address data blocks of varying size.
There’s nothing interesting at reading: the drives all have the same shape of the graph while the places are distributed according to the read response time they have shown in the previous test. The Seagate enjoys a large lead while the others go in a dense group.
It’s more interesting at writing. The HDDs perform differently now and take places according to the results they had in the write response test. Note the results of the Samsung – it slows down suddenly with small data blocks, becoming slower than the ST9160821AS on 512-byte blocks. By the way, it is not the Seagate, but the Hitachi that is the slowest drive with large data blocks.
And now we can proceed from synthetic to more real-life tests.
The Workstation pattern simulates a typical workstation load at request queue depths up to 32. This test is performed on the full capacity of the HDDs as well as on a 32GB partition created on them.
To calculate the overall performance rating of a HDD in this test, we use the following formula:
Performance = Total I/O (queue=1)/1 + Total I/O (queue=2)/2 + Total I/O (queue=4)/4 + Total I/O (queue=8)/8 + Total I/O (queue=16)/16 + Total I/O (queue=32)/32.
You may have noted that I don’t use the Web-Server and File-Server patterns in this review. These server-oriented patterns have little practical worth for 2.5” HDDs because when a HDD of this form-factor is installed into a server, performance is usually not a priority.
There are two leaders here: the HDDs from Toshiba and Seagate. The Hitachi finds itself in the last place – this must be the consequence of the slump in the Database graph. The Samsung doesn’t show the oddities it demonstrated in the Database and Random Write patterns and takes a place right behind the leaders.
Nothing changed when I reduced the address space to 32GB: the Toshiba is still the winner and the Hitachi is the loser.
The multi-threaded tests simulate a situation when there are one to four clients accessing the hard disk at the same time. The depth of the outgoing request queue is varied from 1 to 8. I’ll discuss diagrams for a request queue of 1 because the speeds do not depend much on the number of clients at request queue depths of 2 and more.
You can see the operation (or non-operation) of firmware here. The HDDs go neck and neck at one read thread, with the Fujitsu in the lead and the Hitachi in the rear. Seagate HDDs don’t like to process multiple threads, and the ST9160821AS slows down even when there are two threads to be processed. The Hitachi loses its speed as well, differing from the Seagate at two threads only.
The Fujitsu wins this test. The Samsung has a good result, too.
The HDDs don’t differ much in the multi-threaded writing test, forming three groups. The first group includes the Fujitsu and Western Digital that are the fastest at processing one thread but worse than the Samsung and Seagate at processing multiple threads. The Hitachi and Toshiba are obviously losers in every case.
First, you can have a look at the data-transfer diagrams of the drives:
The data-transfer speeds at the beginning and end of the logical disk are listed in the following diagram, sorted by the speed at the beginning.
You can’t expect HDDs built on almost identical platters (with the same rotation speed and areal density) to be too different, yet the drives are not absolutely identical. The WD1600BEVS is a leader of this test, having maximum results both at the beginning and the end of the disk. The Seagate is the slowest one at the beginning.
The next diagram shows the results of WinBench 99 tests on a 32GB partition sorted by the High-End Disk Winmark results. FAT32 comes first.
Detailed WinBench 99 results are available in tables below (click to enlarge):
You can see the practical consequences of the “oddities” about the Samsung drive you could observe in IOMeter – this drive is unrivalled here. The second and third places go to the drives from Western Digital and Fujitsu while the other three are far behind the leaders. There are the same three losers in the Business Disk Winmark test while the WD1600BEVS and HM160JI change their places.
The Samsung is in the lead in NTFS, too. The HDD from Western Digital has changed places with the Fujitsu and now occupies second place in High-End Disk Winmark and third place in Business Disk Winmark. Note the improved performance of the Seagate – it is only second to the leader in Business Disk Winmark.
Now we will check performance of the hard disk drives with the FC-Test program. Two 32GB partitions are created on the drives and formatted in NTFS and then in FAT32. After that a file-set is created of the hard disk. It is then read from the disk, copied within the same partition and then copied into another partition. The time taken to perform these operations is measured and the speed of the drive is calculated.
To remind you, the Windows and Programs file-sets consist of a large number of small files whereas the other three patterns (ISO, MP3, and Install) include a few large files each.
I’ll discuss the FAT32 results first.
It’s easy to pinpoint the loser, which is the Hitachi drive. The Toshiba is also poor with every file-set except Windows (which contains very small files). It is somewhat more difficult to see the leader. The Samsung delivers high performance with small files (quite forgetting that it didn’t do well in the synthetic tests) while the HDD from Western Digital is superior on large files.
The losers and winners change their places here. The Fujitsu has the best read speed irrespective of the file-set while the Seagate is the slowest of all. On the other hand, the HDDs all deliver similar speeds with little difference from each other.
The speed of copying within one partition depends on what exactly file-set is being copied. The Samsung puts its excellent writing capability to good use with small files and leaves no chance to its opponents. As the average file size increases in the MP3 and Install patterns, the Fujitsu takes the lead. And the Seagate is the best with the ISO pattern. I want to mark the HDDs from Samsung and Fujitsu as having a consistently good speed across all of the file-sets.
There are almost no changes when the distance of copying is increased: the Samsung is better with small files, the Fujitsu with large files, and the Samsung with the ISO file-set.
Now let’s see what we have in the same tests but when the HDDs are formatted in NTFS.
The results of writing in NTFS are almost the same as in FAT32: the Samsung is the best with small files while the Western Digital, with large files. The HDDs from Toshiba and Hitachi have poor results again.
The Fujitsu is good at reading but is now closely followed by the drive from Western Digital. The HTS541616J9SA00 is the slowest drive in three out of the five patterns. The Seagate is surprisingly slow with MP3 files while the Samsung is the loser in the ISO pattern.
The Samsung is in the lead when copying small files. The Fujitsu wins with large files and the Seagate with the ISO pattern. It’s all just as it was in FAT32.
PCMark04 benchmarks the HDD in four different modes: Windows XP Startup is the typical disk subsystem load at system startup; Application Loading is the disk activity at sequential starting-up and closing of six popular applications; File Copying measures the HDD performance when copying a set of files; the Hard Disk Drive Usage parameter reflects the disk activity in a number of popular applications. These four parameters are used to calculate the overall performance rating according to the following formula:
HDD Score= (XP Startup Trace x 120) + (Application Load Trace x 180) + (File Copy Trace x 28) + (General Usage x 265).
The Toshiba wins the XP Startup trace quite to my surprise as this HDD hasn’t done anything exceptional in the previous tests. It is followed by the Samsung and Fujitsu, which have been the leaders at copying in FC-Test. The HDD from Western Digital shows the lowest speed.
The Toshiba is still in the lead in this Application Loading test. The Samsung is second, followed by the drives from Fujitsu and Western Digital. The Seagate and Hitachi are bringing up the rear.
The File Copying test agrees with the FC-Test results: the Samsung, Fujitsu and Seagate take first, second and third places, respectively. The Hitachi has the worst result of all. The Toshiba, which has been successful in the other PCMark tests, is only fifth.
The Toshiba is on top again, yet is not far from the others. The Samsung is second while the Seagate takes last place. The others have similar results.
The Samsung boasts the highest overall score, over 4000 points. The Fujitsu is second due to its excellent result in the copying test and good results in the other tests. The Toshiba was let down by the copying test and is only third with its overall score. The Hitachi and Seagate are the worst drives according to PCMark.
So, we’ve got six hard disk drives with different behavior. I’ll try to give you a brief description of each of them.
The Fujitsu MHW2160BH is a universal device. Featuring excellent sequential read and write speeds, it is the leader at multi-threaded reading. Coupled with good firmware algorithms, this makes it a stable leader at reading or copying large files. It has also been in the leading group through most of the tests.
The Hitachi HTS541616J9SA00 had the worst results in most of my tests. Firmware algorithms are very important for a HDD but this model was one of the earliest 2.5” HDDs with perpendicular recording technology and the developer had little time to polish its firmware off.
The Samsung HM160JI is a leader of this test session largely due to its high speed of processing small files – it has no rivals in this respect. By the way, this model shows that synthetic benchmarks can often be wrong about real capabilities of a HDD and it’s better to check it out in real-life tests as well.
The Seagate ST9160821AS is the first serially produced HDD of 2.5” form-factor with perpendicular recording technology. It is good at copying very large files (the ISO pattern in FC-Test) but otherwise slower than the other HDDs. It has a very low performance in the multi-threaded reading test, which is obviously inherited in its firmware from the 3.5” models.
The Toshiba MK1637GSX was somewhere in the last places throughout the tests but suddenly went ahead in PCMark and IOMeter Workstation.
The Western Digital WD1600BEVS is an average product that has average results in every test except at writing large files where it knows no rivals.
So, each HDD has its highs and lows, and you should choose one basing on the intended applications. For general purposes, the Samsung HM160JI and Fujitsu MHW2160BH should be a good choice.