by Sergey Romanov
07/12/2007 | 07:57 PM
It is a long time since we last tested products from Hitachi Global Storage Corporation (for details refer to our article called Hitachi Deskstar 7K400 HDD: Godzilla Resurrection). Their Deskstar series has got a lot of modifications (7K400, T7K250 and 7K80, 7K500, T7K500 and 7K160) after the Deskstar 7K250 batch which had taken so long to come out. Looks like Hitachi decided to keep up with Seagate in offering a huge variety of solutions within their product families (for details on Seagate's products see our extensive roundup called 500GB HDD Shootout: Seagate Barracuda 7200.10 and Others!).
However, the innovations have been mostly meant to squeeze the last drop of juice out of the longitudinal recording technology prior to the mass transition to perpendicular recording. The areal density growth has been accompanied with an appropriate increase in the read/write speed. The amount of onboard memory has been increased, too, but have Hitachi’s engineers finally put it to good use?
To remind you, our tests showed that the 7K250 was just a little better than the 180GXP whereas the 7K400 differed from the 7K250 in preferring IDE controllers. Hitachi has developed a second generation of native Serial ATA electronics and doubled the cache buffer size since then, but does it do any real good to the company’s HDDs?
Let’s check it out, especially since improvements have been seen in the junior model of Hitachi’s current HDD series with a naked eye (for details see our article called Battle for 160GB: Hard Disk Drive Roundup Part I).
The Deskstar 7K1000 embodies the best of Hitachi’s technological potential. It features second-generation perpendicular recording technology (previously polished off on the 2.5” Travelstar series), read sensors made of iridium-manganese-chromium alloy on femto sliders, thermal fly-height control (previously used in the Deskstar 7K160 and T7K500). It is also the world’s first hard disk drive with a storage capacity of 1 terabyte and with a cache buffer of 32 megabytes. Is it not enough to provoke our interest? Considering the lack of immediate opponents, we have nothing else to do but to compare the 7K1000 with the senior models from previous Deskstar generations (to remind you, we found considerable differences in performance of senior and junior models within the same series, see our article called Hitachi Deskstar 7K250: Vancouver 3 HDD Review for details). It will be interesting to see with what achievements Hitachi is entering the era of perpendicular recording, what defects they have managed to correct, and if they haven’t got any new problems somehow.
Let’s study the specifications first.
Click to enlarge
The most interesting thing you can note in the table (except for the twofold increase in the cache buffer size) is the reduction of the read error rate. Introducing perpendicular recording technology into its top-end 3.5” HDDs first, Hitachi claims to have improved data storage security considerably. The error correction mechanism has been modified for the first time since the Deskstar 120GXP: the Deskstar 7K500 does not use ECC interleaving while the Deskstar T7K500 has a longer ECC symbol. This has helped reduce the amount of auxiliary information while providing the same user data recovery capabilities. Hitachi also claims that the five-platter design contributes to high reliability as well because it provides the desired disk capacity without using extremely high areal density.
On the other hand, more platters consume more power and, usually, generate more noise. Have they made any improvements in these aspects?
Yes, the engineers have reduced the idle power draw of the Deskstar 7K1000 in comparison with the 7K500 and 7K400 and redistributed the load balance between the +5V and +12V rails at reading/writing. The HDD consumes no more than 2 amperes even when it starts up. Interestingly, the Deskstar 7K250 proves to be the most economical drive when spinning up its platters, and the consumption currents are increasing with every new generation.
Some improvements are promised as concerns the noise factor. The five-platter Deskstar 7K1000 proves to be quieter than three-platter models from the previous generations both when idle and when seeking!
Having moved up, the label gives you the view of the more robust three-screw fastening of the heads drive (there are only two screws in the three-platter models). The electronics of the Deskstar 7K1000 differs but little from the previous five-platter generation except for an updated processor model.
This processor is also employed in the Deskstar T7K500 whereas the 7K500 uses the same processor as the Deskstar T7K250. The memory has a fetch time of 6 nanoseconds which corresponds to a frequency of 166MHz. The memory frequency and, obviously, the CPU frequency too, has increased with the release of the Deskstar T7K250 so we can expect the new models to have a considerable advantage over the 7K250 and 7K400.
The Deskstar 7K1000 is produced in two versions, 750GB (on four platters) and 1000GB (on five platters) but it’s also got two twin brothers. These are the CinemaStar 7K1000 that features support for read/write stream control commands and offers a quiet seek mode and the Ultrastar A7K1000 that features protection against vibrations and a faster seek in comparison with the ordinary 7K1000. The desktop version of the HDD lacks all these features and has a slower seek. The three varieties of the HDD all come with a Serial ATA-300 interface only, but Native Command Queuing is officially supported only by the server-oriented modification. Does it mean the Deskstar 7K1000 lacks NCQ? We’ll try to find it out with our tests because the acceleration sensors (VR1 and VR2) are soldered on our sample of the Deskstar 7K1000 even though the Rotational Vibration Safeguard technology is not declared for it!
We have transitioned to a new testbed sometime ago (see our article called Raptor X HDD from Western Digital: Only the "Naked" Truth... for details). The main feature of this testbed is a second-generation IDE controller - Promise SATAII150 TX2plus that supports Command Queuing protocols which are so important today. Moreover, the support for these protocols can be enabled/disabled easily using a Windows-based configuration utility, which allows examining the influence of NCQ on disk performance. Many reviewers use Silicon Image controllers for that purpose but those controllers disable the command queue altogether along with Native Command Queuing! The new testbed also features a faster CPU and Windows XP SP2 instead of Windows 2000 Professional.
The following programs are used in the tests:
We will compare the new HDD with top models from Hitachi’s earlier Deskstar series, beginning with the 7K250 with a Serial ATA interface, so that you could see all the improvements introduced into the architecture inherited from IBM. The exact model names, firmware versions and other parameters can be found in the next section.
The Promise SATAII150 supports not only Native Command Queuing but also the now-obsolete Tagged Command Queuing and we’ve finally got the opportunity to see how it affects performance. The results are really shocking as you’ll see.
We have employed this program, written in our labs, for our earlier reviews. Now we’ll use it again to have an insight into the internal world of hard disk drives.
The growth of the data buffer in the Deskstar 7K1000 is accompanied with a twofold increase in its segmentation at reading, which is good, but the read ahead distance has not changed. Moreover, the read cache is disabled in all the models with a data buffer larger than 8MB – data cannot be read from the buffer, it has to be extracted from the platters again. This modification seems reasonable considering the sophisticated caching mechanisms of modern operating systems, but the adaptive look-ahead read algorithms introduced in the Deskstar 7K250 drives with an 8MB buffer are forgotten for some reason and data is just being read into the buffer… only to be discarded due to the lack of read cache. This “innovation” coincided with the introduction of Native Command Queuing support into the electronics. Hitachi’s programmers seem to be still exploring IBM’s intellectual heritage and try not to mess anything up. We’ll see in the following tests if this conservative approach is justifiable.
Note that the Deskstar 7K250 is the unrivalled leader in average seek time among the 7200rpm IDE HDDs while the average seek time of the Deskstar 7K1000 is worse even in comparison with not the fastest of its predecessors. The increase in the number of servo marks per track must have been insufficient to make up for the increase of the number of tracks.
The results of the low-level tests allow us to make the following suppositions about performance of the Deskstar 7K1000 in comparison with its predecessors: its IOMeter results will be somewhat worse, but its performance in file-processing tests should be higher due to its higher linear speed and huge buffer that can be used for effective caching of writes. Now let’s perform one more unusual experiment.
Besides the new testbed, we’ve got new tests. Making the hard disk read (or write) data sequentially into multiple address ranges, we check out the ability of its firmware to optimize performance by means of look-ahead reading and deferred writing.
The other HDDs’ diagrams can be viewed here .
Hitachi HDDs’ multi-threaded read speed is still only half their linear speed although every new model brings some improvement into this area. The Deskstar 7K500 is better than the others at processing two streams, but the newest 7K1000 is the overall leader.
It is a special thing with the driver of the Promise SATAII150 controller that the first thread blocks all the subsequent ones. When a large file is being read from the hard disk, the reading of any other file is almost impossible, which makes the system irresponsive to the user’s commands. This situation improves at a request queue depth of over 2 outstanding requests, but do you know many programs that ask for a second data chunk from the system before they receive the first one?
You can note that it is the typical behavior of HDDs without Command Queuing but the T7K500 is supposed to support that technology! So far we cannot see any optimizations of multiple requests in the two latest HDDs from Hitachi (equipped with the same processor version, by the way), but they are much better at reading three and four threads alternately.
So, Hitachi still has a lot of room for improvement. We do not check out the additional command set for controlling data thread, which was first introduced in the Deskstar 7K400, because we don’t know of an IDE controller that can utilize this feature. It means we have to wait for look-ahead reading algorithms to improve, especially as the Smooth Stream cannot ensure high speed on multiple data threads without them, either.
It is hardly better at writing: the first thread blocks the others at a request queue depth of 2 even on HDDs with enabled NCQ. This is surely our controller’s fault as we haven’t seen this thing happen on other controllers. Anyway, let’s see how performance of the Hitachi drives degenerates depending on the number of simultaneous write requests.
The HDDs have progressed. Each new generation copes with the threaded load better and better. The T7K250 and 7K500, the first HDDs from Hitachi to be equipped with Native Command Queuing, behave characteristically when processing 2 simultaneous threads. For the more recent models the performance hit is almost proportional to the number of threads and the Deskstar 7K1000 leaves no chance to its precursors irrespective of the number of simultaneous requests thanks to its highest linear speed.
Now that we’ve begun to talk about writing, we’ll run our traditional average response time test.
Like in IOMeter, the Deskstar 7K1000 is the slowest of all Hitachi drives in terms of average read access but performs much better at writing. The ratio of read access time to write access time is indicative of how efficiently disk requests are accumulated and reordered in deferred writing mode. As opposed to our earlier reviews, the efficiency is calculated in percent rather than as a relative coefficient. It shows how much deferred writing reduces the average sector access time.
It is the newest 7K1000 that boasts the highest efficiency, yet it isn’t far better than the Deskstar 7K250. It means there have been no fundamental improvements in the deferred writing implementation despite the introduction of NCQ.
Talking about IBM’s heritage, Hitachi only disabled Tagged Command Queuing in its T7K250 and later models and this technology worked even in the Deskstar 7K400, which was equipped with the modernized electronics from Infineon. And TCQ worked poorly, much worse than in the Deskstar 7K250 even! There was no optimization at reading. At writing the command buffering on the HDD would conflict with deferred writing, resulting in a performance hit in many tests. Surprisingly, the Deskstar 7K400 with a Serial ATA interface, which is also benchmarked with enabled TCQ, does not suffer much. However, its lower speed in many tests is obviously due to TCQ.
Now let’s check out for any surprises at simple reading/writing of sequentially located data.
You can see improvements here. The record performance of the oldie Deskstar 120GXP is finally surpassed. The speed reduction when processing small-size data blocks, which we first saw in the Deskstar 180GXP and which became much worse in the Deskstar 7K250/7K400, has finally been dealt with. Moreover, the Deskstar T7K250/7K500 HDDs achieve their maximum speed on 2KB chunks (the best result ever shown in our labs) and deliver two times the speed of the 7K250/7K400 models. This acceleration cannot be only due to the increase in the clock rate of the electronics from 133 to 166MHz.
The 7K1000 and T7K500 have inherited all those helpful features and added to them their high speed of processing large data blocks. The 7K1000 achieves an unprecedented 82MB/s and the last generation of longitudinal recording from Hitachi, represented by the Deskstar T7K500, isn’t much slower. Interestingly, these HDDs seem to get a second wind when the data block size gets larger than 8KB.
We’ve got a similar picture at sequential writing, too. The two newest models are somewhat slower than the previous-generation leaders when processing small blocks but deliver a considerably higher maximum write speed.
The outcome of this test holds a promise of high performance of the new HDDs in NTFS where most operations are performed over small data blocks.
Now we return to non-sequential loads. The patterns that emulate the operation of an SQL database at different ratios of writes to reads and at different request queue depths will show us the performance scalability of the HDDs as well as their ability to combine reading with deferred writing.
Let’s first examine the case when there are no write requests at all. Reading only.
The average sector seek time is the most important factor when there is no request queue. The long-out-of-production Deskstar 7K250 is the best HDD in this parameter, as we’ve found above, so it is ahead of the others until a load of 4 outstanding requests when it is overtaken by the two 500GB models with a Serial ATA interface. They seem to make it by means of NCQ. NCQ technology also helps the Deskstar 7K1000 leave last place as the load grows higher. The Deskstar 7K500 shows the best scalability of speed depending on the request queue depth. It had no problems in the threaded tests, either.
When there are write request to be performed, the old Deskstar 7K250 with disabled TCQ is in the lead through most of the load range. The others are roughly equal to each other except for the two 500GB models which are outsiders here. NCQ does not work well for them together with deferred writing whereas the Deskstar 7K1000 seems to have improved in this respect.
When there are no read requests, the damage inflicted on older HDDs by Tagged Command Queuing is obvious. Without this technology every Deskstar fits within ±7% of the average: the T7K500 SATA, 7K500 PATA and 7K400 SATA are the slowest drives here and the newest 7K1000 is close behind the leaders 7K250 and 7K500 SATA.
A few patterns that emulate real-life disk applications are discussed in this section.
A heavily loaded workstation comes first.
The old Deskstar 7K250 wins this pattern with a considerable lead because it has a small access time and efficient deferred writing algorithms which are important for this pattern. The three most recent models with an enlarged buffer, including the Deskstar 7K1000, are lagging behind the dense group in the middle of the list.
The 7K1000 is slowest in the file-server mode, even falling behind the Deskstar 7K400 with enabled TCQ! It has a considerable performance hit at a request queue depth of 64. The Deskstar 7K250 is again in the lead while the others have similar results.
The new HDD doesn’t profit much from not having to perform any writing. Native Command Queuing doesn’t help it, either, while the Deskstar 7K500 Serial ATA has put this technology to good use and almost caught up with the Deskstar 7K250, the leader of our IOMeter tests.
We developed this test to measure the speed of processing several file-sets in two file systems. Five file-sets, representing the folder structure of a real computer, are first created on the disk, then read from the disk into system memory, and finally copied into the same partition and then into another disk partition.
To avoid overloading the article with diagrams we present the FC-Test results as tables and as summary diagrams in which the rating of each HDD is calculated as the average of its performance with each file-set in percentage from the best result.
When creating files we have three groups of results that are determined by the linear speed of each HDD. As one could expect, the Deskstar 7K1000 is on top in each file system, although closely followed by the T7K500. Their speeds are almost 50% higher than those of the oldest HDDs in this review, the Deskstar 7K250 and 7K400.
The 32MB cache buffer doesn’t affect the HDD’s performance much whereas there is a significant difference between 8MB and 16MB of memory thanks to which the 7K500 with a Serial ATA interface is ahead of the T7K250.
As we have supposed in the IOMeter section, the Deskstar T7K250 and the newer models have all become faster in NTFS due to the speed improvements in processing small-size data blocks.
The size of the cache buffer has no effect on the HDD’s performance at reading files. The read speed is determined by the speed of reading from the platters and, partially, by the microprocessor in the HDD electronics.
The Deskstar 7K1000 is ahead of its predecessors by the value of the difference in their specified data-transfer rates. The Deskstar T7K250 with a Parallel ATA interface stands out in its group due to the slightly higher areal density it has due to adaptive formatting.
The comments to the write speed diagram are true here as well. Copying files within the same disk partition doesn’t reveal any advantages from the 32MB buffer, and the Deskstar 7K1000 is also slightly slower than the 7K500 in the ISO file-set.
The two most recent generations of Deskstar HDDs exhibit the old dislike of copying large files, but in a lesser degree.
Copying small files is even more impressive than writing them: the Deskstar T7K250 is two times as fast as its predecessors while the newest 7K1000 is three times as fast!
The newest model has its triumph when copying files into another disk partition: it has at last broken free from the pursuing Deskstar T7K500.
There is no doubt the 32 megabytes of buffer memory help a lot when copying data at long distances. The buffer allows to move the heads less thus increasing their efficiency. The new HDD is especially good with the MP3 file-set.
We create two 32GB partitions for this test, which is less than 10% of the total capacity of the 1000GB disk. If the starting point were farther away from the destination point, the Deskstar 7K1000 would have an even bigger advantage over its opponents.
There have appeared no comprehensive programs for testing hard disk drives since Ziff Davis Media Inc. abandoned their WinBench project. Considering the lack of alternatives, PCMark’04 enjoyed a warm welcome and earned popularity as a simple-to-use and comprehensible benchmarking tool. Notwithstanding a number of objective drawbacks of this test, we can’t but check out how Hitachi HDDs have progressed in it.
The diagrams show median values for 10 runs of the test.
The user will feel the new computer being faster even at the step of booting the OS up. The newest Deskstar 7K1000 is going to leave the most pleasant impression, but the T7K500 Serial ATA and the 7K500 and T7K250 models with a Parallel ATA interface aren’t much slower. It’s interesting that the classic ATA is preferable to Serial ATA in this test although all the HDDs are tested on the same controller with the same driver. The server-oriented Deskstar 7K400 and the T7K250 and 7K500 with a Serial ATA interface, which have distinguished themselves in some of the earlier tests due to their NCQ support, aren’t brilliant here. This gives us another reason to suspect a lack of NCQ in the two newest HDD models from Hitachi.
The Deskstar 7K1000 is far ahead of the others in this application loading test. The similar results of the three 500GB HDDs indicate that this test is not sensitive to the amount of buffer memory or the linear speed but depends on the total storage capacity of the HDD.
Except for the identical results of the Deskstar 7K1000 and T7K500 the overall picture resembles the FC-Test results. The HDDs with an 8MB buffer are on the losing side whereas the 32MB buffer doesn’t seem to bring in much profit.
As opposed to the previous test, this one seems to value the amount of the buffer memory and the new Deskstar 7K1000 is unrivalled here. The T7K500 is again slower than the older 7K500, but both are far ahead of the rest of the HDDs that are equipped with only 8 megabytes of memory.
The overall ratings line the HDDs up in an orderly manner except for the Deskstar T7K250 Serial ATA which falls out of the order. Its low score can be explained by defects of the particular sample (for example, its linear read speed in IOMeter did not correspond to its specification).
The General Usage test has the biggest weight in the overall rating, putting the Deskstar 7K1000 on top with a considerable lead. The four models with same-density platters and an 8MB cache buffer have absolutely identical results. The newer models increase their advantage in PCMark04 slowly and the T7K1000 makes a performance jump only due to its 32MB buffer.
We didn’t have time to test all the HDD models included into this review in the new version of PCMark. Anyway, you’ll be able to see the progress of Hitachi’s HDDs over a substantial period of time.
The results are similar to the same test of the previous version of the benchmark except that the Deskstar 7K1000 enjoys a bigger lead over its predecessors and that the 7K500 Serial ATA is faster than its version with a Parallel ATA interface. The capacity of the HDD seems to determine the result here.
The results are almost 100% identical to those of the previous version of the benchmark, but the T7K250 has fallen farther behind. The HDD capacity must be the decisive factor again.
This benchmark agrees with PCMark04 again: the size of the buffer is the decisive factor. And the dependence on the storage capacity has become stronger.
The file copy test has been replaced in the newer version of the benchmark with two separate tests of write and read speed when scanning files for viruses. Unfortunately, the resulting read speed has little to do with reality. Particularly, the results are roughly similar among the tested HDDs and exceed their physical capabilities. If the HDD is requested sequentially located data with small pauses, the data will be given out from the buffer by means of look-ahead reading. This must be the case with this test.
The write speed as reported by PCMark’05 almost equals the possible maximum for sequential writing. The Deskstar 7K1000 even exceeds this theoretical maximum thanks to its huge buffer. So, these two tests, which have replaced one copy test of PCMark04, don’t have anything valuable to tell us.
The storage capacity is the decisive factor for this benchmark, so the overall scores shouldn’t surprise you.
We’ll finish this review by discussing the results of the old, yet still indispensable, ZD WinBench. After we began to limit the size of the formatted partition the test is run on, the results do not depend too heavily on the capacity of the HDD and have become more repeatable as well. It means they have something to say about the HDDs’ performance.
The Business Disk Winmark and Front Page tests in FAT32 report that there has been no progress until the T7K500 and 7K1000 models while these two new drives are equals. AVS agrees with them in general, but thinks that the 7K1000 is a little bit faster. The rest of the tests seem to note that the 16MB buffer of the 7K500 Serial ATA is not a trifle and ensures a considerable performance increase to it. VC++, Adobe Premier and, especially, Sound Forge are all glad at the improvements in processing small data blocks which we noticed earlier in the IOMeter tests.
And finally, the High-End Disk Winmark test says that the first performance jump occurred with the new electronics of the Deskstar T7K250, the second jump was due to the introduction of a 16MB buffer, and the third jump was achieved by modifications in the firmware algorithms for the T7K500/7K1000.
Thanks to their new electronics the Deskstar T7K250 and earlier models acquire a solid advantage in NTFS: 30% in Business Disk Winmark and over 80% in Visual C++! Many applications have reacted favorably to the increase in the amount of cache memory to 16MB. Photoshop and VC++ are also pleased with the improvements implemented in the 1TB model. Thus, a performance increase of 100% occurred in the WinBench NTFS tests in the three years since the release of the Deskstar 7K250!
We would be glad to continue this kind of testing further as it is a real enjoyment. Hard disk performance is currently an annoying limitation of the PC speed in many user tasks like games or professional applications. And our tests show that Hitachi is making a breakthrough after a certain period of stagnation.
They accelerated the electronics with the release of the T7K250 and then they introduced a 16MB buffer in the 7K500. These HDDs were not a total success as they didn’t always show the expected performance benefits, but those 16 megabytes of memory are put to better use in the next generation of HDD electronics, in the T7K500. This somehow affected NCQ support, particularly when it comes to processing two competing read threads, and the copying of large files became somewhat worse, too. The T7K1000 features certain improvements in the firmware and is equipped with 32MB of memory, but still performs oddly with NCQ.
Does the Deskstar 7K1000 support Native Command Queuing at all? We don’t know for sure. Some tests seem to imply a lack of NCQ but IOMeter reports a performance increase as the load grows up, which is typical of NCQ-enabled models. We even suspect that the HDD does not do anything NCQ-related until the load is higher than 2 simultaneous requests. For example, the 7K1000 and T7K500 behave like follows when performing random-address reading:
Otherwise, the newest HDD from Hitachi is almost perfect. It offers a record-breaking storage capacity, an extremely high sequential speed, an enlarged buffer for more performance, and fast processing of multiple data threads. As a result, it performs superbly in all our benchmarks except for IOMeter’s server patterns, but for server applications Hitachi offers the Ultrastar A7K1000. This even gave some hardware reviewers the cause for pitting the Deskstar 7K1000 against the WD1500 aka Raptor II, although the two belong to absolutely different categories. We hope the minor drawbacks of the Deskstar 7K1000 revealed in our tests will help Hitachi’s engineers and programmers to go on improving their products even more.
We want to note that we tested a presale sample of the Hitachi Deskstar 7K1000 and this HDD is still being improved. Hitachi has promised to make its 1TB drive even faster. Samsung, Seagate and Western Digital are ready to begin shipping 1TB HDDs, too, so we hope we’ll have the opportunity to compare serial samples of such products from all the manufacturers in our next review.