by Aleksey Meyev Nikita Nikolaichev
10/23/2009 | 01:09 PM
Somewhat shockingly for ourselves, this is going to be our first comparative review of hard disk drives with Serial Attached SCSI interface. As far back as 2004 we published an article on SAS technology and have kept silent on the issue since then. Our recent series of SAS RAID controller reviews does not count in because we compared arrays and controllers rather than individual HDDs. As a matter of fact, we indeed thought that our readers were more interested in ordinary HDDs for desktop computers and focused on them in our reviews.
But the computing world has changed. Applications and operating systems have become heavier and put forth higher disk subsystem requirements. At the same time, SAS drives have come to cost reasonable money and SAS controllers are now integrated into nearly-mainstream mainboards. On the other hand, a fundamentally new type of storage media has entered the market in the way of solid state drives. Having a fantastically high speed of reading, SSDs are less good at writing. Moreover, their service life is limited by the number of flash memory rewrite cycles (magnetic storage media are not subject to this problem). SSDs also have a very high cost of storage as yet.
Can SAS drives serve as a reasonable compromise in the three-coordinate system (speed, reliability and price)? This sounds like an interesting topic for an upcoming review. Today, we will just get together SAS drives that have been produced in the last four years to provide you with a complete picture of this market sector.
Fujitsu’s drives come first in this review. Unfortunately, we will never see new HDDs under this brand. Following its February press release, Fujitsu sold its HDD development division (both 2.5-inch and SAS models) to Toshiba. Today, all the Fujitsu drives in this review can be seen in Toshiba’s product catalogue. Hopefully, the change of the owner has not affected the production and these HDDs will develop further. We don’t think that Toshiba, which had not produced SAS drives before, bought Fujitsu’s division just to let it die. Anyway, these products have come to our labs with the Fujitsu logo and they can be found in shops still, so we will present them to you under their original names.
The Fujitsu MAX3 RC series were Fujitsu’s first 3.5-inch drives with a spindle rotation speed of 15,000rpm which were equipped with SAS rather than SCSI interface. Their capacity is not large by today’s standards: the senior model can store 147 gigabytes of data on its four platters while the single-platter junior model can only store 36 gigabytes. The latter is not included into this review, but you will see the test results of 73GB and 147GB models. All HDDs of this series come with 16 megabytes of cache memory.
Yes, the specs are far from impressive, but that’s not a reason to exclude them from the comparison. First of all, these HDDs have a very low response time typical of 15,000rpm products. And second, their owners might be interested in learning what performance benefits they can achieve by transitioning to newer models. A server infrastructure has its own laws. A larger capacity and a somewhat higher speed may be unimportant while the cost of a disk subsystem upgrade is crucial for it. Users now often choose cheaper HDDs to store server data and do not hope to increase data security by purchasing newer models. Therefore the market of server HDDs is conservative and buyers do not run stumbling after new models in it.
Some time after the previous series Fujitsu introduced one more 3.5-inch HDD series with SAS interface and a spindle rotation speed of 15,000rpm. The MBA3 RC series is a logical development of the previous one, the main difference being the double storage capacity thanks to increased recording density. Most of the electronics is the same including the main chips LSI L7A1989 and Marvell 88C7500. The senior model of the series reached 300GB by transitioning to perpendicular recording technology. You will see the series in its entirety in this review: 73, 147 and 300GB models. This seems to have been the last SAS HDD series from Fujitsu. The company did not release newer products.
You may be familiar with this series from our reviews already because we use such HDDs for our SAS RAID controller tests.
The Ultrastar 15K147 series was Hitachi’s first series of 15,000rpm drives with SAS interface. It is in fact similar to the above-discussed MAX3 RC and both have the same maximum storage capacity. There is one important difference, though. While the opponents (no only Fujitsu, but also Seagate) reached this storage capacity with four platters, Hitachi needed as many as five. There are differences between smaller-capacity models as well: the other makers produced 73GB drives with one platter and two heads whereas this Hitachi (we’ve got only a 73GB model from this series) has two platters and three heads, i.e. three rather than two operating surfaces. It is also unclear if Hitachi used lower-density platters because it could not achieve higher recording density or wanted to ensure more reliability. Or, if the recording density is all right, Hitachi may have used the 5-platter design to reduce the platter’s operating diameter in order to minimize its response time. We’ll check this out shortly.
We can also add about this series that it comes with 16 megabytes of cache (that was quite a lot then because SCSI drives of the previous generation and many contemporaries had only 8MB). The electronic section was based on Hitachi’s own chip together with Infineon’s UAB-M9611 produced specially for Hitachi.
Hitachi gave up the 5-platter design in the next generation: the maximum capacity of 300GB is achieved in the 15K300 series by means of four platters just like in the opponents’ products. It is the 300GB model that will represent this series in our review. The buffer is still 16MB large while the Infineon chip, which works together with Hitachi’s own chip, is upgraded to version UAB-M9710.
Hitachi’s latest (even though it is not particularly new) series is 15K450. Its maximum storage capacity is 450GB. The series is somewhat queer as it includes only two models (350 and 450GB), both with four platters and eight heads. There are no junior models at all. It looks like Hitachi released these models for users whose priority is storage capacity. Other users are supposed to be content with the previous-generation HDDs with lower recording density (small SAS drives are often united into a RAID to achieve a low response time, so Hitachi may be quite right on this point). There are no fundamental differences: the HDDs have the same design but the Infineon chip is now version UAB-M9612.
In those times when HDDs with the Maxtor logo were not a copy of Seagate’s inexpensive HDD series but products of an independent company, server-oriented HDDs of this brand were considered among the best. Therefore we’ve decided to pay our homage to the memory of the company and include the SAS version of the Atlas 15KII drive into this review. This HDD has the same recording density per platter as its contemporaries: 37GB. Its cache is 8MB large. Its electronics are based on the Marvell 88C7500 chip we already know by the Fujitsu drives together with an Agere Buffy-C1-MP. Curiously enough, Agere does not exist anymore, either. Like Infineon, it merged into LSI some time ago.
It is in the Cheetah 15K.4 series that Seagate’s 15,000rpm drives first acquired the SAS interface. Before that, Seagate had provided HDDs with either SCSI or Fiber Channel. From today’s standpoint, all drives of that generation look roughly alike. Like its opponents, the 15K.4 series has a maximum capacity of 146KB achieved by means of four platters. The cache is 8MB large, like in the Maxtor. As for control electronics, we see a Marvell 88C7500 again and two chips from LSI (L2E2425 and L2A2661) on the bottom of the HDD.
The next generation of the Cheetahs, 15K.5, has an increased maximum capacity (300GB, thanks to perpendicular recording) and a larger cache (16MB). The two LSI chips are replaced with one LSI L2A3075.
As time went by, 300GB ceased to look like a big storage capacity and Seagate introduced another series called 15K.6. Thanks to the rapid development of perpendicular recording technology, the recording density had been quickly boosted by 50%, giving birth to SAS drives with 450GB capacity. These HDDs look different in comparison with their predecessors. The electronic chips are now on the interior side of the PCB. This ensures more protection for them, but on the other hand, the chips get hotter at work.
Finally, in the summer of 2009, Seagate rolled out the next update of its 15,000 HDD series. The top Cheetah 15K.7 model (with four platters and eight heads) offers an unprecedented capacity of 600GB. Besides, the interface has been updated: the new series supports both SAS 1.1 and the recently ratified SAS 2.0. We will soon devote a review to this new interface where we will discuss new SAS 2.0 controllers. We’ve got a simple reason for not paying much attention to SAS 2.0 now: as you will see shortly, the first version’s 300MBps bandwidth is more than enough for existing HDDs. Today, the transition to the newer version can only speed up operations with the HDD’s cache memory.
The 15K7 series comes with 16 megabytes of cache. Maximum reliability is a key factor for server storage and keeping a lot of data in the cache is not desirable.
Winding up this introduction, there is a model with a spindle rotation speed of only 10,000 rather than 15,000rpm. But what a drive it is! The Seagate Cheetah NS.2 has a storage capacity of 600GB like the above-described 15K7. It is going to be interesting to compare old 15,000rpm drives with this model which has a much higher recording density but a lower spindle rotation speed.
The following testing utilities were used:
We formatted the drives in FAT32 and NTFS as one partition with the default cluster size. For some tests 32GB partitions were created on the drives and formatted in FAT32 and NTFS with the default cluster size, too. The drives were connected to an LSI SAS3041E-R controller.
It was not easy to choose the controller, by the way, as we could only use a simple controller without cache memory to compare the pure performance of the HDDs with little influence on the controller’s part.
We chose the LSI SAS3041E-R because it delivered maximum bandwidth and had almost no effect on the performance of the HDDs. However, SAS drives perform very slowly when writing in FC Test on this controller. We should account for that.
There are too many results, so we build three diagrams for each test: one for 73GB drives, one for 146GB models and one for the rest of the drives.
We used the good old WinBench 99 for our low-level tests. Unfortunately, the Hitachi 15K450 does not perform here. It would not pass the test despite all our efforts.
The next diagram compares the HDDs in terms of speed at the beginning and end of the full-capacity partitions created on them.
Everything is neat and tidy. You can easily tell between HDDs of different generations even if they belong to different brands. Each new generation of platters delivers a characteristic and higher speed. Still, there are three things that should be pointed out.
First of all, let’s take a look at the oldest generation of HDDs that was the first to be equipped with the SAS interface. Two models stand out in it: the Seagate 15K.4 has the biggest difference between the speed at the beginning and end of the partition. It means that this HDD most likely has platters with the biggest diameter and, consequently, we cannot expect it to have a good response time. The Maxtor Atlas 15KII is just the opposite. It has a much smaller difference in speed than usual, meaning that its platters are the smallest. The Hitachi, which differs from the other HDDs with the number of its operating surfaces, does not show anything exceptional.
The second interesting thing is the performance of the new Seagate 15K.7. It seems to have very small platters, but take a look at the left part of its data-transfer graph:
There is a small section – about 20GB – at the beginning of the disk where the read speed is higher than 200,000KBps. Then goes a very long (up to half the disk’s full capacity) flat stretch at 184,000KBps. Is it a peculiarity of the HDD’s formatting (in which case this super-fast stretch just calls for being made into a separate partition) or an error of the benchmarking program? We’ll check this out later on.
And finally, the 600GB NS.2 with a spindle rotation speed of 10,000rpm delivers good performance. It is no rival to the 15K.7 but almost as good as the previous-generation 15,000rpm products (represented by the 15K.6) in terms of sequential speed. And it is obviously better than the older models.
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 the drive’s sequential read/write speed on the size of the data block. This test is indicative of the maximum speed the drive can achieve.
The numeric data can be viewed in tables. We will be discussing graphs and diagrams.
It is all very simple with sequential speeds: HDDs do not differ much within the same generation whereas a transition to higher-density platters, when a new generation arrives, leads to a considerable performance growth. Take note of the performance of the Seagate 15K7. It is indeed as high as 200MBps! Seagate has progressed in terms of firmware as well. The company’s older series are obviously slower than their opponents on small data blocks, but the 15K.6 and newer series go ahead.
The 10,000rpm NS.2 model is still as good as the senior models of the previous generation.
The standings do not change much when the HDDs do sequential writing. Seagate’s HDDs improve even more on small data blocks (8KB and smaller): the first two generations are obviously slower than their opponents whereas the newest two generations go ahead of their rivals.
Hitachi’s team is disappointing: even the latest 15K450 series is slow processing small data blocks. The 300GB Fujitsu MBA3 RC has some inexplicable problems with writing. Its speed is very low although its smaller-capacity series mates behave much better.
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 the HDD is much larger than its cache, so we get a sustained response time that doesn’t depend on the HDD’s buffer size.
Save for two models the HDDs all fit into a range of 0.5 milliseconds that lies a little below 6 milliseconds. The NS.2 features the highest response time as its lower spindle rotation speed does not allow to compete with the other HDDs in this test. The venerable Maxtor Atlas 15KII turns out the winner. Small-diameter platters and high-speed heads ensure first place for this very old product. As for the others, you can see somewhat better results from the Fujitsu MAX3 RC (funnily, the next series, MBA, is worse) and Seagate’s newest 15K.6 and 15K.7 series. The youngest Cheetahs claim to be the fastest indeed.
When it comes to writing, the HDDs from Seagate’s two newest generations stand out again. The 146GB 15K.6 model has odd problems that result in an indecently high response time. The 73GB Seagate 15K.5 and the Hitachi 15K147 have very poor results, too (their write response is comparable to their read response, which is not good).
Now we’ll see the dependence between the drives’ performance in random read and write modes on the size of the data block.
We will discuss the results in two ways. For small-size data chunks we will draw graphs showing the dependence of the amount of operations per second on the data chunk size. For large chunks we will compare performance depending on data-transfer rate in megabytes per second. This approach helps us evaluate the disk subsystem’s performance in two typical scenarios: working with small data chunks is typical for databases. The amount of operations per second is more important than sheer speed then. Working with large data blocks is nearly the same as working with small files, and the traditional measurement of speed in megabytes per second becomes more relevant.
Let’s start with reading.
When reading small-size data blocks, the HDDs are ranked up according to their response time we have measured in the previous test. Thus, the old Maxtor Atlas 15KII is first while the NS.2 with slow platters is last.
When the HDDs are reading large data blocks, their sequential read speeds become the crucial factor and the standings depend on recording density. The newer the HDD, the higher its result is. A good response time can only make a difference here if the other factors are equal.
The graphs of writing small random-address data blocks makes us recall the response time test again. The HDDs that had a high response time are bad in this test, too. The 146GB Seagate 15K.6, 73GB Seagate 15K.5 and Hitachi 15K147 draw almost horizontal graphs whereas the other HDDs accelerate as the data block is getting smaller. This must be due to some flaws in the firmware of the specific HDD models.
It is the speed of sequential writing, which depends on recording density, that is the decisive factor when the HDDs are writing large data blocks. You can see that starting from 512KB blocks.
As for the Seagate NS.2, it is just a little slower than the previous-generation 15,000rpm products at three out of four loads. However, it is no match to its older but devilishly fast opponents at reading small random-address data.
In the Database pattern the drive 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% with a step of 10% throughout the test while the request queue depth varies from 1 to 256.
You can click these links to view the tabled results for the IOMeter: Database pattern:
We will build diagrams for request queue depths of 1, 16 and 256.
The Maxtor Atlas looks much better than the others at the minimum queue depth. It can do more operations per second than its opponents thanks to lower response time. Its advantage at mixed loads is especially impressive. We guess the only weak spot of this HDD is its small cache buffer. It is inferior to the HDDs with 16MB cache in terms of deferred writing efficiency. Seagate’s two latest generations look better than the others, too. And we can recall that their response time is good, too.
The secret of the low writing performance of the 73GB Seagate 15K.5 and the Hitachi 15K147 is revealed: they do not have deferred writing at all.
When the queue depth is increased to 16 requests, each HDD begins to display its unique character, showing the peculiarities of its firmware. The Maxtor is very good among the smaller-capacity models but its sluggishness at writing is now obvious. Both drives from Fujitsu outperform it easily at high percentages of writes. The 73GB Seagate 15K.5 and the Hitachi 15K147 are still poor: their graphs are nearly horizontal lines that lower steadily towards higher percentages of writes.
The Seagate 15K.6 is the best of 146GB drives although its unexpected slump at 80% writes spoils the picture somewhat. This diagram is actually very informative as to the progress of the HDDs: the products from Fujitsu and Seagate improve with each new generation even under server loads which are highly sensitive to firmware and almost indifferent to recording density.
When it comes to larger and newer HDDs, Seagate’s two latest generations prove to be faster than the others. The rest of the HDDs go close to each other, the 10,000rpm Seagate NS.2 keeping up with them, too. The 300GB Fujitsu MBA3 RC is a disappointment. Its firmware seems to ponder for a long time over each request, being only more or less good at writing.
At a very long queue depth (it is indeed very long, and if you see such a queue in an operating server, you should think about upgrading its disk subsystem) the Hitachi 15K147 and Fujitsu 146GB MBA3 RC go ahead whereas the old Seagate 15K.4, 15K.5 and the Maxtor turn to be slow. The Seagate NS.2 is uncompetitive to the HDDs with higher spindle speed under such a high load.
Winding up this section of our tests we will show you a diagram with each drive’s performance at five different queue depths.
Aren’t these graphs odd? Yes, we’ve got used to see quite a different picture with 7200rpm HDDs. The reason is that SAS drives do not have aggressive deferred writing algorithms (being reliability-optimized, SAS drives prefer not to store too much data in the deferred writing cache) but have a high reading performance due to low response time.
While the two junior models of the Fujitsu MBA3 RC series don’t differ much from their predecessors, the 300GB model shows a completely different behavior. The HDD does not change performance when the queue depth is increased from 4 to 16 requests. The further performance growth is very modest. So, this HDD obviously has different firmware.
Hitachi’s second and third generations of SAS drives behave in the same way and differ strikingly from the 15K147. Frankly speaking, we like the newer version more. Notwithstanding its less predictable performance and lower speed at very long queue depths, it is overall superior to its predecessor as it delivers higher performance at 50% and higher writes when the queue is shorter.
The Maxtor behaves in a peculiar manner. It is similar to the other HDDs under low loads, being rather slow at writing (but very fast at reading). As a result, it delivers almost the same performance at high percentages of writes irrespective of whether the queue depth is 64 or 16 requests long. A further increase of the queue does not bring any performance benefits as the HDD has already reached its top performance (it cannot maintain a longer queue).
Seagate’s first SAS disk series behaves much alike to the company’s main opponent then, Maxtor, having a low efficiency of writing and no performance growth at queue depths longer than 64 requests. The graphs are all somewhat lower than in the opponent’s diagram: Seagate wouldn’t have a chance in comparison.
As you know, Maxtor was bought by Seagate and their struggle ended due to the default of one of the combatants. Seagate took to improving the firmware then. As a result, the senior disks of the 15K.5 series are almost as good as the Maxtor Atlas 15KII although still somewhat worse than the latter under mixed loads.
The junior model is radically different and resembles the Hitachi 15K147, drawing nearly horizontal graphs, too. However, the Seagate shows a higher performance growth at longer queue depths than the Hitachi. As the result, the Hitachi speeds up at very long queue depths whereas the Seagate reaches its maximum performance at a queue depth of 64 requests and does not accelerate thereafter.
The next generation of Seagate drives is faster yet, as is especially conspicuous with the 15K.6 series at reading and at mixed loads. They also begin to display a very odd peculiarity of their behavior: they are much faster at a queue depth of 64 requests than at 256 requests. The HDD seems to be not meant for such loads and gets overwhelmed by the flood of requests.
This seems to be the stop to the development of Seagate’s HDD firmware. The HDDs of the newest generation, 15K.7 and NS.2, behave alike to their predecessors, although do not get stifled at a queue depth of 256 requests. They have no performance growth at that queue depth either, though.
The drives are tested under loads typical of servers and workstations.
The names of the patterns are self-explanatory. The Workstation pattern is used with the full capacity of the drive as well as with a 32GB partition. The request queue is limited to 32 requests in the Workstation pattern.
The results are presented as performance ratings. For the File-Server and Web-Server patterns the performance rating is the average speed of the drive under every load. For the Workstation pattern we use the following formula:
Rating (Workstation) = 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.
This pattern consists of read requests only and we can see different leaders here. At short queue depths there are no rivals to the two newest generations of Seagate drives. The Maxtor is the only other drive that can get close to them. When it comes to maximum speed achievable, there are a lot of models that claim the leadership: Fujitsu’s MBA3 RC series (except for the 300GB model), Hitachi’s two newest series, and the Seagate 15K.7.
You should also note at what queue depth each HDD reaches its top speed. We’ve got two roughly equal groups here: one group accelerates up to a queue depth of 64 requests while the other goes on speeding up to a queue depth of 128 requests. The Seagate 15K.6 series shows a queer behavior: its performance indeed lowers after the queue grows longer than 64 requests.
The standings might have been predicted: first place goes to the newest Seagate 15K.7. The Seagate 15K.6 drives are second and third thanks to their excellent speed at short queue depths. The Maxtor is fourth.
The 10,000rpm model is no competitor to the grownups here. It can only beat the 300GB Fujitsu which has obvious problems with reading.
The overall picture does not change much when there are some write requests in the pattern. We’ve got the same winners at short queue depths and the same highest-speed models. The Seagate 15K.6 series slow down at queue depths longer than 64 requests, again. We want to note one thing here: the Maxtor and the Seagate 15K.4 stop accelerating at a queue depth of only 32 requests. It looks like this limitation is due to their having only 8 megabytes of cache.
We see the same HDDs at the top of the diagram, the only difference being the 146GB Seagate 15K.6 which has sunk to the middle of the list.
This pattern produces a variegated load typical of workstations. And the HDDs behave so differently at short queue depths that their graphs intertwine like lianas in a tropical forest. Let’s try to make out the most interesting things.
The most impressive fact is that the old Maxtor leaves no chance to any other HDD. Its superiority is obvious. The HDDs from Fujitsu and Hitachi feel good, too. Seagate’s team are not that successful. The old 15K.4 is the only Seagate to be competitive in this test while the later models, including the newest 15K.7, are rather unconfident. They like the server loads much more.
The performance ratings highlight the Maxtor’s huge advantage and the poor performance of the new generations of Seagate drives.
The situation is dramatically different when the test zone is limited to a 32GB partition. The Maxtor is very good again but only under low loads. As soon as there is a queue of about 7 requests, larger HDDs enjoy a considerable performance boost (a 32GB partition takes a much narrower zone on them due to the use of multiple platters and higher recording density).
The outcome of this test is highly illustrative: HDDs with high recording density and multiple platters get all the top places. First place goes to the super-dense 4-platter Seagate 15K.7. Take note that these test conditions help the Seagate NS.2 leave its last place, outperform all the 73GB models (excepting the sprightly Maxtor) and get very close to the 146GB models. It cannot match the 300GB, let alone 450GB, models, though.
The multithreaded tests simulate a situation when there are one to four clients accessing the virtual disk at the same time – the clients’ address zones do not overlap. We will discuss diagrams for a request queue of 1 as the most illustrative ones. When the queue is 2 or more requests long, the speed doesn’t depend much on the number of applications. You can also click the following links for the full results:
It is all simple when the HDDs are reading just one data thread. The results are actually the same as in the sequential reading test above. We can only note that the Seagate 15K.5 series is somewhat faster than its contemporaries.
The number of data threads is increased to two, and we can’t but applaud Seagate’s HDDs. Starting from the 15K.5 generation these HDDs are indifferent to a second read thread, refusing to slow down. Fujitsu’s drives and Hitachi’s two latest generations lost about one third of their speed. The addition of a second read thread is too heavy for the first generation of SAS drives, except for the Fujitsu. These oldies slow down more than fourfold!
There are no big changes in the standings when we add more threads. All the Seagate drives, except for the 15K.4, deliver excellent performance whereas the other HDDs are far from perfect.
Writing one thread is not interesting. We’ve seen these results already, including the terribly low performance of the 300GB Fujitsu MBA3 RC. There is only one unclear fact: the 73GB Seagate 15K.5 is very slow writing even one data thread.
Most of the HDDs survive the addition of a second write thread with minimum speed loss. They seem to have not noticed the harder load. Hitachi’s 15K147 and 15K300 series drives even speed up a little. Both series from Fujitsu and the Maxtor slow down more, but it is the Seagate 15K.4 that suffers a real performance hit. That’s another reason for us to praise Seagate for having quickly optimized its server HDDs for such characteristic multithreaded loads.
Again, the overall picture does not change when we add a third and fourth thread.
For this test two 32GB partitions are created on the SSD and formatted in NTFS and then in FAT32. A file-set is then created, read from the SSD, copied within the same partition and copied into another partition. The time taken to perform these operations is measured and the speed of the SSD is calculated. 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.
We’d like to note that the copying test is indicative of the drive’s behavior under complex load. In fact, the SSD is working with two threads (one for reading and one for writing) when copying files.
You should be aware that the copying test not only indicates the speed of copying within the same HDD but is also indicative of the latter’s behavior under complex load. In fact, the HDD is processing two data threads then, one for reading and another for writing.
This test produces too much data, so we will only discuss the results achieved with the Install, ISO and Programs file-sets in NTFS. You can use the links below to view the full results:
As we’ve written already, this test is far from representative. The HDDs all have awfully low speeds due to the specific combination of the load and the controller’s driver.
Still, we can see a few HDDs that are slower than the main group. These are the 147GB Fujitsu MBA3 RC, Seagate’s NS.2, 146GB 15K.4 and 73GB 15K.5, and the Hitachi 15K147. We cannot see any regularity here, but both HDDs with no deferred writing are in this slower group. But why are the other three HDDs so slow? It is also unclear why the 300GB Fujitsu MBA3 RC has such a huge advantage and why the Maxtor is so good, too.
It is all simple at reading: a higher recording density leads to a higher speed of sequential reading, so the HDDs from Seagate’s two latest generations, the Seagate NS.2 and the Hitachi 15K450 are in the lead. This rule is only really true for the ISO file-set, though. It’s different with smaller files where the Hitachi 15K300 joins the group of leaders. Interestingly, the Maxtor is last notwithstanding its very low response time. Its firmware is optimized for server loads rather than for reading files.
We’ve got the same trio of leaders when copying within the same partition: the Seagate 15K.7 is in the lead, followed by two disks from the previous generation. The Hitachi 15K450 and Seagate NS.2 are fighting for fourth place in the Install and ISO patterns but give way to the 73GB Fujitsu MBA3 RC in the Programs pattern.
The 73GB Seagate 15K.5 and Hitachi 15K147 lack deferred writing and lose this test completely. The old Seagate 15K.4 is on the losing side, too.
We see a similar picture when copying from one partition to another: the top and bottom tiers contain the same models. The only difference is the surprisingly good performance of the 300GB Seagate 15K.5 in the Programs pattern.
PCMark 2005 has the same tests as the 2004 version (not only in names, but also in results as we have seen a lot of times), so we only discuss one test from PCMark 2004 which is not available in the 2005 version. It is called File Copying and measures the speed of copying some set of files. The other results can be learned from the table. The PCMark 2005 tests are:
The final result of the average of ten runs of each test.
Of course, these benchmarks, and the subsequent PCMark Vantage, are not so crucial for server disks as such HDDs are meant for different applications and loads. Anyway, let’s compare the HDDs in such tests, too.
Some HDD models will not be tested in this and next tests for technical reasons.
Click here to see the complete PCMark 2004 results table.
The copying results are almost the same as in FC Test: the two newest generations of Seagate drives are in the lead (the Cheetah 15K.7 being the fastest of all) and the two HDDs with no deferred writing are at the bottom of the diagram. Take note that as soon as the server tests have been passed (they are highly sensitive to response time), the Seagate NS.2 takes a stable position somewhere in the middle of the standings table.
The Fujitsu MBA3 RC series cope well with booting Windows XP up. The Seagate 15K.7 is the only disk to squeeze in among them. The pair of HDDs with no deferred writing are the slowest again. Take note that this fact has a bigger effect on the results than the twice lower recording density of some of their opponents.
Fujitsu’s latest generation are the best at loading applications although it is the Maxtor that gets first place. The 73GB Seagate 15K.5 and the Hitachi 15K147 are still the slowest, and the Seagate NS.2 is slow, too.
There are some changes in the standings in the General Usage test: the Maxtor is first, the Seagate 15K.7 is second, followed by the same Fujitsu drives. And the losers are the same as in the previous test, too.
The test of scanning for viruses is highly sensitive to firmware algorithms. All of Fujitsu’s drives, including the old MAX3 RC, cope well with it. They are followed by the drives from Hitachi and Maxtor whereas Seagate’s team are poor in this test.
This test has no conflict with the controller’s driver, but its results are not interesting at all. The standings are the same as in the sequential writing test with sharply defined product generations (the 73GB Seagate 15K.5 is an exception).
Fujitsu’s HDDs enjoy the top three HDD scores. The Seagate 15K.7 is fourth and the Maxtor is fifth. The two losers have been poor throughout this benchmark.
To make this part of our test session complete, we are going to run the latest version of PCMark called Vantage. Compared with the previous versions, the benchmark has become more up-to-date and advanced in its selection of subtests as well as Windows Vista orientation. Each subtest is run ten times and the results of the ten runs are averaged.
Here is a brief description of each subtest:
Basing on these subtests, the drive’s overall performance rating is calculated.
Fujitsu is in the lead, again. They are contesting with the Hitachi 15K450 under this load. Take note of the last-but-one place of the Seagate NS.2. This is one more test this drive dislikes.
There are on changes under the gaming load: Fujitsu’s HDDs seem to be set on claiming the title of best home-oriented disks.
Photographers are going to be delighted at Fujitsu’s HDDs! Will any other drive offer some competition?
Well, Seagate’s two latest series mix up with Fujitsu’s HDDs in the top of the diagram when booting Windows Vista.
Fujitsu does not want to give up its top positions but the Seagate 15K.7 and the Maxtor have pushed them down a couple of places in the Movie Maker test.
Here is one more test that depends heavily on how efficiently the HDD’s firmware can use cache memory. Funnily enough, this is one of the few tests where 15,000rpm SAS drives are inferior to their desktop counterparts. Anyway, it is Seagate’s HDDs of the latest generations that take the podium here.
The leaders do not differ much from the others here. Still, we can name them: Seagate 15K.7, 300GB Fujitsu MBA3 RC and Maxtor.
There are no changes in the Application Loading test: the Maxtor is first, followed by Fujitsu’s drives.
These results might have been expected. Fujitsu’s MBA3 RC series drives have two positions on the podium, leaving third place to the Seagate 15K.7. Take note that the Maxtor is not among the leaders anymore and the Seagate NS.2 cannot compete with the 15,000rpm products.
Next goes our homemade test of defragmentation speed. We created a very defragmented file system on a 32GB partition of a disk by loading it with music, video, games and applications. Then we saved a per-sector copy of the disk and now copy it to the disk we want to test. Next we run a script that evokes the console version of the Perfect Disk 8.0 defragmenter and marks the time of the beginning and end of the defragmentation process. For more information about it, you can refer to this article.
So, here is one more test where the SAS drives are inferior to desktop ones. The best of desktop HDDs pass it in less than 20 minutes while the Hitachi 15K147, the best HDD in this test, spends 50% more time that that. We can note that Fujitsu’s 147GB and 300GB drives and the maximum-capacity disks in the Seagate 15K.7 and 15K.6 series are better than the others.
There are also two HDDs that are very poor here: the 73GB Seagate 15K.5 and the Hitachi 15K300. It is rather too much to spend one hour to defragment 24 gigabytes of data.
Now we are going to show you one more interesting test in which we use WinRAR version 3.8 to compress and then uncompress a 1.13GB folder with 8118 files in 671 subfolders. The files are documents and images in various formats. These operations are done on the tested drive. This test depends heavily on CPU performance, but the storage device affects its speed, too.
The HDDs differ but little when compressing data, but we can note that the Maxtor and Fujitsu’s MBA3 RC series are in the lead while the 73GB Seagate 15K.5 and Hitachi 15K147 are on the losing side. The Seagate NS.2 is slow, too.
The difference is bigger when the HDDs are unzipping the archive: the leader is almost 50% better than the worst model. The Seagate 15K.7 is first, followed by the 300GB Seagate 15K.6 and 300GB Fujitsu MBA3 RC. The three slowest drives are the same as in the previous test: 73GB Seagate 15K.5, Hitachi 15K147 and Seagate NS.2.
You can refer to our article called Hard Disk Drive Power Consumption Measurements: X-bit’s Methodology In depth for details on this test. We will just list the specific modes we measure the power consumption in:
Let’s check out each mode one by one.
Fujitsu’s HDDs are the most moderate ones in terms of startup current. Even the 4-platter Fujitsu is highly economical. We might commend the Hitachi 15K147 as well, but its low 12V consumption is combined with a rather high power draw on the 5V line. Seagate’s HDDs of the two latest series and the Seagate NS.2 are equipped with the most economical electronics but each of them needs nearly 3 amperes on the +12V line, which is quite a lot. Hopefully, this current is only necessary to spin the platters up as quickly as possible.
The Maxtor behaves interestingly in idle mode: its voracious electronics consume more than its mechanics! The same three series from Seagate are the most economical in terms of 5V consumption. Now we can see the strong point of the Seagate NS.2. With its 4-platter design is needs less power than any other product in this test. It is cooler as the consequence, which is important for data centers and server racks with multiple disks inside. Fujitsu’s single- and dual-platter models are overall more economical and cooler in comparison with their opponents.
When doing random reading, all the HDDs need more power from the 12V line because they have to move their heads around a lot. As a result, nearly every model, except for the highly economical 10,000rpm Seagate NS.2 and the modest single-platter Fujitsu MBA3 RC, goes beyond 10W, the 4-platter and many 2-platter models even exceeding 15W. None of the 15,000rpm drives can be viewed as economical, so maximum-density models seem to be preferable as you can use them to build a disk subsystem with lowest heat dissipation and power consumption (by using models with fewer platters) or by reducing the number of disks.
We have to name the most voracious 2-platter drives that match 4-platter models in temperature and power consumption. These are the Maxtor (its old electronics and powerful heads actuator consume a lot) and the Seagate 15K.5 (its electronics is voracious, too). The 4-platter Hitachi 15K300 and Seagate 15K.5 have the highest power consumption of all, requiring over 17 watts of power!
Do compare the 5V consumption of the HDDs at random reading and in idle mode. You will see that many drives consume more power from the 5V line when idle than at work! What are they doing then? You can answer this question by monitoring the drives’ power consumption: they are doing patrol reading. That is, they are automatically reading from platters in idle mode, searching for potentially bad sectors. Enterprise requirements to reliability are so strict that even HDDs have to be constantly keeping themselves in good shape.
Fujitsu’s HDDs are again somewhat better than their opponents in terms of power consumption, the single-platter MBA3 RC even outperforming the 10,000rpm model. There are no serious changes in the standings, though. It is the second-generation drives that are the most voracious, again.
The overall picture doesn’t change much when we switch to sequential operations. The Seagate NS.2 is the most economical drive, being better than Fujitsu’s products which are in their turn are somewhat more economical than the others. The Maxtor and the Hitachi 15K147 have high power consumption on the 5V line again. Each next generation is a little more economical than the previous one if you compare models with the same number of platters and much more economical if you compare same-size models.
This test session has produced expected results. Newer HDDs are faster than older ones, consume less power and offer a better price/capacity ratio. Thus, it is the right decision to choose newer HDDs for your data storage system (if you consider them reliable, of course).
When choosing a SAS drive for a workstation, it is the acoustic characteristics that may be the crucial factor, but we have not tested them and cannot thrust our subjective opinion about the appeal of particular drives on you. But generally speaking, SAS drives of latest generations can be used in desktop PCs with enough comfort.
The Seagate Cheetah 15K.7 is the indisputable leader of this test session. Its fantastic mix of high performance and large capacity leaves a strong impression. The Cheetah NS.2 must be noted, too. Its lower spindle rotation speed is made up for by higher recording density. So, if you are looking for an HDD for your enterprise data storage system, but do not trust SATA disks, you may want to consider the Cheetah NS.2.
The intriguing performance of the Maxtor drive which despite its respectable age was sometimes competitive to newest models will be discussed in more detail in one of our upcoming reports.
We would like to award the winner of our today’s roundup - Seagate Cheetah 15K.7 - with our Editor’s Choice title as the fastest SAS hard disk drive: