by Aleksey Meyev
02/08/2011 | 03:27 AM
The free space on our computers’ hard disks is getting consumed at a tremendous rate in our never-ending search for higher quality. HD video, 15-megapixel photographs, lossless audio, high-resolution textures of latest games are the reasons why we don’t measure hard disk capacities in just gigabytes anymore. It doesn’t matter if a hard drive has a few gigabytes more or less now that we use the term terabyte to describe it.
After a long period of exponential development hard disk drives have slowed their pace down in the last year and a half. Besides the obvious manufacturing difficulties, the problem is that moving beyond the 2-terabyte barrier is not as easy as it seems.
One aspect of this problem is in the Master Boot Record, a table located at the beginning of a partitioned disk and designed for choosing a partition an OS is going to be booted from. The MBR wouldn’t be necessary if we all had but one partition on each of our hard disks. The computer’s BIOS would initialize the disk and pass execution to code contained at a certain address. But as we do have multiple partitions, we cannot do without the MBR. The BIOS loads the MBR which in its turn loads the OS located in one of the partitions described in the MBR. The problem is that the MBR was invented when HDDs used to have much smaller storage capacities than today (by the way, the alignment requirement for today’s HDDs with 4-kilobyte sectors goes back to that time, too). The MBR could originally address only 224 sectors. One sector being 512 bytes large, this equaled 8.4 gigabytes. And even this method required some tricks with CHS (cylinder-head-sector) addressing which involved presenting an HDD as a monster with as many as 255 heads. Some users may still remember that problem. Later on, the LBA addressing mechanism was introduced to replace CHS and expand the addressable space to 232 sectors. This is about 2.2 terabytes. Thus, the MBR cannot access sectors beyond 2.2 TB. No new partitions can begin there, which means that a single-partition disk can have a maximum storage capacity of 2 terabytes. As HDD makers hold that 2 terabytes means 2 trillion (1012) bytes rather than 241 (which is a bigger number) as operating systems think, today’s 2TB HDDs are free from that problem, but something had to be done about 3TB models.
Like in RAID arrays, a high-capacity disk could be sliced into logical partitions (LUNs), each smaller than 2 terabytes, and presented as individual physical HDDs, each with its own MBR, to the BIOS. The manufacturers prefer another solution, though. They use the GUID Partition Table, a new and advanced disk partitioning method. It is a table of partitions with unique partition IDs and no code to execute. It is the Extensible Firmware Interface, a new standard devised to replace the conventional BIOS, that is responsible for orchestrating the boot-up process. The GPT can address up to 9.4 zettabytes (9.4x1021) which should suffice for a long time. It stores a backup copy of the main table and checksums for partition IDs, supports up to 128 partitions and even has an MBR at the beginning to solve some compatibility issues. There are such issues, actually. The GPT is supported by every 64-bit version of Windows, by Windows 2003 Server and Vista SP1, but old OSes, including such popular ones as 32-bit Windows XP and Windows 2000, won’t work with GPT-enabled HDDs. The replacement of the MBR with the GPT involves one more important nuance: the standard BIOS cannot work with GPT partitions. That is, the OS will see such partitions, but must be installed on an MBR one for the computer to boot it up. It is only computers with the mentioned EFI (instead of BIOS) that can boot up from GPT partitions but EFI mainboards are very few as yet.
Besides, the HDD controller’s drivers must support disk partitions larger than 2 terabytes. That’s quite an important requirement because otherwise the OS would only see about 800 megabytes instead of 3 terabytes. Such drivers for Intel chipsets were only released in December 2010 and generic Windows 7 disk drivers had had to be used until that.
For all these pitfalls, the transition to the new storage capacity could not be delayed much longer and the HDD manufacturers got to work on it. Their websites now offer special sections detailing the subtleties of the move to 3 terabytes, and we’ve got three HDDs of the record-breaking capacity, two of which can already be found in retail shops.
We begin with the acme of Hitachi’s new 7K3000 series. Unlike in the past years when Hitachi used to be lagging behind its opponents in bringing new products to market, this time around there have been almost no delay at all. Hitachi have done everything to deliver a cutting-edge product just in time. Besides the record-breaking capacity, this 7K3000 model is the company’s first to feature a 64MB buffer and SATA 6 Gbps interface. Of course, the spindle rotation speed of this top-of-the-line HDD is 7200 RPM. It looks like Hitachi have finally woken up from the slumbers they have been in for the last few years.
This HDD is not yet presented on the Seagate website. It is the single product of the three that has not yet made it into retail. We mean, not as a standalone product. It comes in the new external FreeAgent GoFlex Desk drive we actually extracted it from. As far as we know, it has five 600GB platters, like the above-described Hitachi. It is suspected to use SmartAlign technology, similar to Advanced Format. That is, it may have 4-kilobyte sectors. We will check this out shortly. Like the rest of this product series, it has a spindle rotation speed of 7200 RPM.
By the way, it is the problems with using 3TB HDDs in computers that must have induced Seagate into releasing this HDD as an external one without offering an internal counterpart. They have also found an elegant way to solve the MBR-related problem. The external HDD emulates the use of 4-kilobyte sectors (it does the emulation even if the disk inside really has 4KB physical sectors because hard disks with 4KB sectors appear to have 512-byte sectors to the outside world). Thus, the number of sectors being the same, the maximum addressable space is increased from 2 to 16 terabytes.
Western Digital have already established a tradition of releasing their next highest-capacity model in the energy-efficient series first, i.e. with a spindle rotation speed of 5400 RPM. Thus, this 3TB drive is just another step in the evolution of the Green series. It has 64 megabytes of cache and 4KB sectors.
Western Digital took the problem of high-capacity HDDs seriously. Like Seagate, they first released external products of this kind in the My Book Essential series. But unlike Seagate, Western Digital also offered such HDDs as individual products to be installed into desktop computers. It is in fact a kit including an HDD and an RR620 controller from HighPoint. This simple dual-port SATA controller is necessary to avoid the problem with drivers not supporting 3TB HDDs. We shall see what it is capable of. As we’ve seen a number of times in our tests, a disk controller can have a big effect on performance.
The WD Green doesn’t look as fast as its opponents in its specs, yet it has one advantage. Advanced Format helped the manufacturer increase the recording density to 750 gigabytes per platter, which means that this HDD has only four rather than five platters. Coupled with the reduced rotation speed, this should make it more economical and, consequently, cooler.
The following table lists the firmware versions of the tested HDDs:
You should keep it in mind that the performance of an HDD may vary depending on its firmware version.
The following testing utilities were used:
HDDs are tested with generic OS drivers. We format them in NTFS as one partition with the default cluster size (for FC-Test we create 32GB partitions), connect them to a mainboard port and enable AHCI. We have transitioned to a new method of testing HDDs, by the way.
The Western Digital drive will be tested in two ways everywhere save for the low-level tests: 1) connected to a mainboard port and 2) connected to the included HPT-RR620 controller.
We use our internal IOMark tool for low-level tests. Let’s begin with sequential reading.
Let’s compare the disks according to the read speed at the beginning and end of the full-capacity partitions created on them.
Well, it looks like it’s time we said goodbye to SATA version 1.0 just as we did to PATA earlier. 7200RPM disks have already reached the theoretical peak bandwidth of that interface, 150 MBps, and are almost as fast as 160 MBps with lucky head/surface pairs. On the other hand, the current transition to SATA 3.0 looks like reaching too far into the future because HDDs still have a long way to go up to 260 MBps which is the practical limit for SATA 2.0.
The WD drive isn’t as fast as the leaders, delivering no more than 130 MBps. Moreover, the data-transfer rate of that HDD fluctuates wildly, indicating that the manufacturer has squeezed everything possible out of the platters.
Now, what about the speed of the cache buffer?
According to the top speed diagram, the Hitachi wins at reading and the Seagate at writing. However, we prefer the smooth and neat graphs that the WD disk has drawn. We can also note that Seagate have managed to reduce the performance hit on large data blocks we observed with the company’s other products. The Hitachi seems to use firmware from the 1000.C series with some improvements. The only thing we don’t like about it is the performance hit when it is reading data blocks 32 to 256 sectors large.
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.
The numeric data can be viewed in tables (click the links below). We will be discussing graphs and diagrams.
The 7200RPM disks deliver the same top speed when reading large data blocks. When processing small blocks, the Seagate is superior whereas the Hitachi even falls behind the 5400RPM disk from WD. The latter shows us a difference between the controllers: the included HPT-RR620 is faster than the chipset’s controller when reading small data chunks.
The 7200RPM HDDs both behave in the same way at writing, the Seagate again enjoying an advantage with small data blocks. The WD performs differently with its two controllers. It is slower on the HPT controller, even in terms of the top speed.
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 Seagate wins this test, its good results being indicative of what you can expect from a modern HDD with multiple read/write heads. The Hitachi doesn’t move its heads around that fast. Its designers must have thought about its noisiness besides other factors. As a result, the 7200RPM Hitachi is inferior to the 5400RPM WD in terms of the read response time.
The WD disk now shows the same result irrespective of the controller. Take note that this HDD is handicapped when writing small data blocks (to write a 512-byte chunk of data, it has to read a 4KB block, modify those 512 bytes in it, and then write it back to the platter) but we don’t see anything like that with the Seagate. So, the Seagate drive either has no 4KB sectors or features some miraculous technological innovations.
Now we will see how the performance of the hard disks in random read and write modes depends on the size of the requested data block.
It’s all rather boring in this random read test. The HDDs do just as we could expect them to, basing on their read response results in the previous test.
There are some interesting facts at random writing. First, the WD confirms it really has 4KB sectors because its performance is low with data blocks smaller than 4 kilobytes. The Seagate is just as positive about having no such technology. Second, the Hitachi has an unusual performance hit with 256KB data blocks. This may be some unlucky coincidence or a defect in its data caching algorithms.
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% with a step of 10% throughout the test while the request queue depth varies from 1 to 256.
You can click this link to view the tabled results.
We will build diagrams for request queue depths of 1, 16 and 256.
The Seagate comes out the winner at the shortest queue depth, just as we might have expected, considering its response time results. The Hitachi uses deferred writing to overtake the leader in the right part of the diagram, though. The WD is not that good at deferred writing although has the same amount of cache (64 megabytes). It may take a different approach to using the cache, though, like creating fewer, but larger cache lines.
As the requests queue gets longer, the Seagate enjoys an even larger advantage at high percentages of reads. The WD also performs well at such loads, competing with the Hitachi. The latter goes ahead at high percentages of writes, though, leaving the WD far behind.
Winding up this part of our tests, we will build diagrams showing the performance of each HDD at five different request queue depths.
The Hitachi has rather aggressive firmware algorithms but enables its deferred writing only at high percentages of writes or at very long queue depths.
The Seagate Barracuda XT makes us recall server-optimized HDDs, also from Seagate themselves. They usually have such smoothly rising graphs. So, Seagate have come up with efficient server-optimized firmware.
The WD disk shows what we can call a special character. Other WD disks behave like that, too. Take note that the HDD slows down at certain percentages of writes when the request queue is long. It seems to allocate more cache memory for reordering read requests at the expense of deferred writing when the queue is long.
The HDDs are tested under loads typical of servers. The names of the patterns are self-explanatory. The results are presented as performance ratings which are calculated as the average speed of an HDD at every load.
The Seagate is an unrivalled leader at both types of load. The Hitachi isn’t very good here because it is equaled by the 5400RPM opponent.
Take note that each graph becomes horizontal from a queue depth of 32 requests. And if you take a look at the Database graphs, you can see that the HDDs all deliver the same performance at the maximum queue depth and at a queue depth of 64 requests. We didn’t benchmark the HDDs at a queue depth of 32 requests in the Database pattern, but we are sure they would deliver the same performance then, too. In other words, all these HDDs seem to support a queue of no longer than 32 requests. This is not a feature of Intel’s new drivers because the WD disk behaves like that on the HPT-RR620 controller, too. So, this must be a specific trait of the whole new generation of HDDs.
The multithreaded tests simulate a situation when there are one to four clients accessing the hard 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 simultaneously running applications. You can also click the following links for the full results:
This test often produces unpredictable results. This time around, the WD performs superbly whereas the Seagate suffers a catastrophic performance hit. The Hitachi is in between them, taking a 50% performance hit at multithreaded reading.
The Seagate is the only drive to have some problems with multithreaded writing. Depending on the number of data threads, it goes ahead or falls behind the others. So, its performance is rather inconsistent here. We can also note that the 7200RPM Hitachi is no better than the 5400RPM WD when writing multiple data threads.
For this test two 32GB partitions are created on the hard disk and formatted in NTFS. A file-set is then created, read from the disk, copied within the same partition and copied into another partition. The time taken to perform these operations is measured and the speed of the HDD 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.
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.
You can use the following link to view the full results.
We don’t see one leader in the Write test. The Seagate is better at writing large files while the Hitachi is good at writing small ones, especially the Windows file-set. The WD drive is unable to compete with the 7200RPM models here. It processes small files at about the same speed irrespective of the controller, but can write large files faster when connected via the ICH.
Quite expectedly, the Hitachi and Seagate go neck and neck at reading. The WD disk is somewhat slower than the leaders and works better when connected via the ICH.
The Seagate and Hitachi are equals when copying large files, but the Hitachi is somewhat faster with small ones. The WD is again behind the leaders and prefers the mainboard-integrated disk controller.
Compared with the previous versions, the Vantage version of PCMark is more up-to-date and advanced in its selection of subtests as well as Windows Vista orientation. Each subtest runs ten times and the results of the ten runs are averaged.
Here is a brief description of each subtest:
Basing on these subtests, the HDD’s overall performance rating is calculated.
The new 7200RPM 3-terabyte HDDs have high and similar results in this benchmark: the Seagate has faster heads but the Hitachi has advanced caching mechanisms. When connected to the mainboard, the WD disk performs just as a regular power-efficient HDD. But it gets worse when connected to the controller included with it. If you take a look at the individual tests, you can see that the included controller is slower in tests that require fast caching or have a lot of write requests.
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 that disk and now copy it to a disk we want to test. We run a script that evokes the integrated defragmenter of Windows 7 and marks the time of the beginning and end of the defragmentation process. For more information about this test, you can refer to this article.
In this and the next test we will benchmark the HDDs with 4KB physical sectors without aligning their partitions. The results are interesting enough even though we don’t usually do such tests.
A look at this diagram is enough to realize that the Seagate Barracuda XT has no 4KB sectors. Otherwise, it wouldn’t be able to be as fast with the unaligned partition as with the aligned one. You can see this with the WD which needs 50% more time to defragment the unaligned partition. Overall, we can see again that the Seagate and the Hitachi are equals, the WD falling behind them both.
Now we are going to show you one more interesting test in which we use WinRAR version 3.91 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 HDD. This test depends heavily on CPU performance, but the storage device affects its speed, too.
Here, we don’t see much difference between the aligned and unaligned partitions. The partition status doesn’t affect the speed of reading. As for writing, data can be written in large blocks in this test, which minimizes the negative effect of the unaligned partition. Switching from the faster to the slower HDD improves the result by 10 seconds only. The controller is more important because the faster controller helps the WD disk pass the test faster by 10%.
The speed of an HDD and the partition status (aligned/unaligned) are both important when unpacking the archive but the controller is not. The Hitachi comes out the winner, the unaligned WD being more than 50% slower than the leader!
You can refer to our 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.
The WD is the most economical HDD when starting up. It expectedly needs less power than the others from the 12V line which powers an HDD’s mechanics. It takes more time to spin up, without having high peaks of power consumption, and has fewer platters with heads. You should note that the peak power draw of these HDDs is up to 12 amperes on the +12V line when they are starting up. This makes it clear why the SATA standard introduced the staggered spin-up feature. If ten such HDDs were starting up simultaneously, they would require up to 20 amperes from the +12V line, which is quite a lot.
The WD makes up for its losses in the earlier tests by being the most economical product in idle mode. It needs only half the power required by the Hitachi and Seagate. Well, it has one platter less and its platters are rotating at 5400 RPM, after all. The 5-platter 7200RPM models are good enough, too. We can also note that the Hitachi needs half a watt less than the Seagate.
According to our methodology, we measure the power consumption 10 minutes after we stop all disk requests. But when we left our HDDs on the testbed for a longer time, we saw very interesting numbers:
Yes, every HDD is consuming from the 5V line only, and a very modest amount of power at that. What does it mean? It means that they have parked their heads and turned off the spindles. They have switched into a deep sleep mode. The numbers you can see in the diagram are the power consumption of the electronics waiting for a command to wake up. Each of the three HDDs falls asleep after 20 minutes of being idle. So, don’t worry if you can’t hear a connected HDD. It may be just sleeping and saving power when there is no work for it to do.
The energy-efficient drive from WD is still in the lead at random-address loads, but its advantage isn’t very large at random reading where an HDD has to move its heads about very quickly. The Hitachi is somewhat better than the Seagate at random reading: this must be the tradeoff for the Seagate’s excellent read response time. The two 7200RPM models need about the same amount of power for random writing where deferred writing algorithms come into play and reduce the load on the HDD’s mechanics.
The WD is the most economical HDD at sequential reading and writing, too. However, it doesn’t have the advantage it enjoyed at random-address operations or in idle mode. The Hitachi is somewhat more economical than the Seagate, probably because its electronics has lower power consumption. We must acknowledge the dramatic progress of Hitachi products, by the way. The company’s 5-platter HDDs of the previous generation consumed about 50% more power!
Each of the three tested 3-terabyte drives is good in its own way. The Seagate Barracuda XT is an all-purpose drive which is especially good at server loads. It is a shame it is not available as a standalone product. The Hitachi Deskstar 7K3000 is already selling in retail, being a well-balanced continuation of the company’s 5-platter traditions. Although not as good at server loads as the Seagate, it is faster with small files and somewhat more economical, which is good for home users. As for Western Digital, their Green series product is good as static storage. And don’t forget that this 5400RPM drive is only slow compared to the 7200RPM models of the new generation. It can deliver the same performance as the previous-generation 7200RPM products and has fast read/write heads to compete with today’s 7200RPM drives in the IOMeter: Database pattern, for example.
We should note that installing 3TB drives may be difficult unless you run Windows 7. You should take a look at the informational materials from the HDD makers to see if your computer supports GPT partitions and whether your disk controller can work with 3TB disks.
And we are now waiting for 3TB drives not covered in this review: a WD Caviar Black and a 3TB HDD from Samsung.