Well, it turned out that the birth of this hard disk drive was announced at least twice. For the first time it was announced by Quantum, then it sank into oblivion for some time. And not so long ago this HDD was introduced to the world once again but under a totally different brand name. Of course, we are talking about Atlas 10K III, which attracted all our attention.

Since it is not quite correct to test a single HDD, we had to involve some other SCSI models with 10,000rpm spindle rotation speed. In the end we got enough testing participants for a roundup. :-)

Well, let's get started!

Testing Participants

So, let us tell you a bit about the hard disk drives, which we included in our roundup. First of all, please, meet Maxtor Atlas 10K III:

Maxtor Atlas 10K III

Although it looks as if Maxtor hadn't yet made up its mind about the brand name:

Or still has a plenty of stickers like that in stock. :-)

We haven't yet had the chance to take a closer look at Fujitsu SCSI HDDs, however, this company has very reliable reputation and we decided to take advantage of the situation and to get acquainted with its SCSI solutions. We would like to start with MAN3367MP:

Fujitsu MAN3367MP

We were also very unpleasantly surprised to find out that we haven't yet ever tested any SCSI hard disk drives from IBM! Hurrying to correct this awful unfairness we would like to introduce to you DDYS-T36950, which will have to compete with a bit more up-to-date drives from other manufacturers.

IBM DDYS-T36950

As for other testing participants, we have already tested them before: they are Quantum Atlas 10K II, Seagate Cheetah 36XL and Seagate Cheetah 73LP (see our Seagate Cheetah 36XL HDD Review for details).

As for the last HDD mentioned, Seagate Cheetah 73LP, we would like to make a few additional comments. As a rule we included in our comparisons the drives with the same storage capacity. It allowed us to create a fair competition for different models of drives from different manufacturers in the first place, and in the second place, we wanted to give you the possibility to select the most suitable solution for your particular needs from a number of products with the same storage capacity. In the current investigation we had to veer a bit away from our traditions and to let Cheetah 73LP take part in the race. The matter is that this drive is available with only one possible storage capacity of 73GB, though its architecture (18GB per platter) corresponds to that of the latest Maxtor drives (Atlas 10K III) and Fujitsu (MAN). And if Seagate only wished, they could easily develop 18Gb and 36GB solutions as well. Since Maxtor and Fujitsu HDDs with 73GB storage capacity are unavailable at the moment (though they are present in both product families), we can only make some approximate estimates concerning their performance basing our forecasts on what we manage to squeeze out of 36GB models.

Unfortunately, we didn't manage to get hold of any IBM 73LZX HDDs, that is why you should forgive their absence in this roundup (as well as the absence of Fujitsu MAJ).

Brief Technical Specs

As you may see, the major differences between the new Atlas and the old one are higher data density, Ultra320 interface and slightly better access time rate. Judging by the specs, IBM DDYS seems to be the only outsider here, as 6GB platters by a hard disk drive with 10,000rpm spindle rotation speed look somewhat awkward, we believe…

Testbed and Methods

Our test system was configured as follows:

• Intel Coppermine 600MHz CPU;
• ASUS CUBX-E mainboard with 1007A BIOS;
• 2x128MB PC133 SDRAM by Hyundai;
• Matrox Millennium 4MB graphics card;
• Adaptec 29160N SCSI controller;
• Windows 2000 Pro.

To test the HDDs we used Adaptec ASC29160N controller with 1.00 drivers for W2K and the ordinary LVD cable. For SCA HDDs (Maxtor Atlas 10K III and IBM DDYS-T36950) we used a special SuperMicro HS-chassis. All the drives were tested in Ultra160 mode.

We used FAT32 and NTFS file systems to format each of the hard disk drives as one logical drive of the maximum size with the default cluster. All the tests were run 4 times and then the average results were taken for the diagrams. The HDDs didn't rest for cooling down between the tests.

Here are the benchmarks used:

• WinBench 99 1.2
• HDTach 2.61
• IOMeter 1999.10.20

Performance

Access Time

To measure the Average Access Time we used two tests (as usual): HDTach and Disk Inspection Test from WinBench99 package.

Despite the better parameters claimed in the specs list, Atlas 10K III let its predecessor, Atlas 10K II, take the lead in HDTach. However, in WinBench99 Atlas 10K III leaves Atlas 10K II behind. So, what shall we trust then? :-)

IBM solution appeared the slowest hard disk drive of all, which si quite understandable, to tell the truth: it's hard to be fast with 12 heads…

Sustained Linear Read Speed

Sustained Linear Read Speed is the quintessence of the per platter data density (we have already discussed it many times in our articles). Let's see how the positions from the data density point of view got distributed between the testing participants:

We can clearly see which HDDs belong to the latest hard disk drive generation, and which are somewhat outdated, to put it mildly. Seagate Cheetah 73LP, Maxtor Atlas 10K III and Fujitsu MAN3367MP feature 18GB platters (which is the maximum density possible for SCSI drives), which makes them look much more attractive against the competitors' background.

The highest read rate belongs to Cheetah 73LP, while Fujitsu MAN boasts the smallest difference between the read rate in the beginning and in the end of the drive. It usually implies that the average read rate is quite high (which we will check in the next HDTach benchmark).

The worst result again belongs to IBM DDYS for the same reasons: it is completely outdated…

HDTach 2.61

The diagram below contains the results for three tests: read rate from the HDD buffer, average read speed and average write speed. This way we can kill three birds with one glance :-)

As we have expected, Fujitsu won the first prize in the average read speed test, though it didn't manage to get too far ahead of Seagate Cheetah 73LP. Maxtor Atlas 10K III turned out the third in the read speed test, and the second in the write speed test (Fujitsu MAN is ahead of all here as well). However, Atlas 10K III managed to win the first prize as well: it leads the race in burst read test.

WinBench99 1.2 for Win2000 (FAT32)

We have already mentioned many times in our reviews that the today's IDE hard disk drives perform better in this benchmark than SCSI ones. Now let's see if it is true or not:

DTR: Beginning

Graph

-

Graph

Graph

Graph

Graph

Hm, things are far not that bad! Four SCSI hard disk drives with 10,000rpm rotation speed managed to break the 10,000 barrier in Business Disk WinMark and 30,000 barrier in High-End Disk WinMark.

The winner in both tests is Maxtor Atlas 10K III. The second position in High-End Disk WinMark belongs to its predecessor, Atlas 10K II. And the third - to Seagate Cheetah 73LP.

WinBench99 1.2 for Win2000 (NTFS)

Again Maxtor Atlas 10K III won all the laurels in High-End test. The second winner appeared Seagate Cheetah 73LP, which turned out the No 1 in Business test.

Intel IOMeter

We have nearly got used to the fact that every benchmark helps us discover something new and interesting. And this benchmark was also not an exception.

First of all, we can clearly single out two hard disk drives, which appear the leaders at low workload (1-16 requests). They are Quantum Atlas 10K II and Seagate Cheetah 36XL.

Secondly, Atlas 10K II retains its leading position at higher workload and only when the number of requests exceeds 176, it yields to Fujitsu MAN.

As for Maxtor Atlas 10K III, it turned out slower than Atlas 10K II for some reason when the number of requests got greater than 16.

It might be called "optimization for multi-HDD RAID-arrays. At least Atlas 10K III is faster than Seagate Cheetah 73LP and IBM DDYS in all cases.

In this pattern we can also watch Quantum Atlas 10K II and Seagate Cheetah 36XL beat all the other competitors at low workload. The greater gets the workload, the greater differs the performance of these hard drives. The performance of Atlas 10K II grows up quite nicely up to 64 requests and then this growth gets significantly smaller. The performance of Cheetah 36XL also grows up, but it gets more or less stable at 16 requests already. Further growth is nearly negligible.

Fujitsu MAN and Maxtor Atlas 10K III HDDs, which yielded in the interval between 16 and 64 requests to Quantum Atlas 10K II and Seagate Cheetah 36XL, rush ahead of them when the workload exceeds 64 requests.

In this pattern the situation repeats again. Fujitsu leaves everyone else behind in case of 1-4 requests and 150-256 requests (judging by the graph). Quantum Atlas 10K II performs brilliantly between 16 and 150 requests. The interesting thing here is the fact that Maxtor Atlas 10K III managed to catch up with the predecessor, Atlas 10K II, only in case of maximum workload. Does it mean that 0.2ms difference between the two models grants Atlas 10K II such a tremendous advantage over the competitor? Or is it still the firmware the matters here?

Temperature

To measure the temperature of the working HDDs we used a special distant infrared thermometer, which we have already told you about in our Roundup: 6 Leading 7200rpm Hard Disk Drives of the Year 2001. The room temperature stayed around 22oC, and the measurements were made after more than 2.5 hours of IOMeter tests running. Here is what we've got:

The results turned out quite logical, we should say: the more platters and heads the HDD features, the higher rises its temperature. We were pleased to discover that the today's dual-platter hard disk drives with 10,000rpm spindle rotation speed heat up almost as much as the regular ATA-drives with 7,200rpm do. It should be the bearings quality and the casing shape that matter then.

The "hottest" HDD appeared IBM DDYS, and the "coolest" - Fujitsu MAN.

Conclusion

Well, our tests showed that Maxtor Atlas 10K III performs much better than Quantum Atlas 10K II in WinBench tests, however, in IOMeter it turned out considerably slower than the predecessor (especially at lower workload). Why? Is it because of the increased data density per platter or some other firmware optimization? Hard to say…

However, there is something to make up for the slight performance lag of Atlas 10K III: it is quieter and cheaper to produce. And lower production costs result into lower retail price, that's quite logical :-)

Anyway, Maxtor Atlas 10K III yielded only to one rival, Fujitsu MAN, out of the entire majority of 10,000rpm HDDs.

Quantum Atlas 10K II is a very interesting solution with excellent "server performance". Unfortunately, it will become very hard to get soon.

Fujitsu MAN3367MP is the first SCSI hard disk drive from Fujitsu we got our hands on. And we liked it a real lot, we should say. No matter that it didn't win in file tests, the results shown in server benchmarks were strikingly high. This HDD appeared especially efficient in case of highest workload.

Seagate Cheetah 73LP revealed impressively high performance some time ago, but time passes and it doesn't seem that fast any more. Its major trump, a combination of large storage capacity (73GB) and high speed, are no longer a threat to the competitors, which hard disk drives boasting the same maximum storage capacity are as fast as our hero or even faster. So, we are very curious to find out what could we expect from Seagate next year.

Seagate Cheetah 36XL is a hard disk drive, which showed very interesting results. The graph showing the dependence of the performance in IOMeter on the workload is very far from standards and hence arouses a lot of questions. Maybe it has something to do with SAMS (Seagate's Advanced Multi Drive System optimizes performance in server and RAID applications). However, we don't know anything for sure…

IBM DDYS-T36950 is an old solution from IBM, which can't stand any serious competition against younger and more progressive rivals. The company's reputation can be saved only by the new IBM 73LZX family, which should appear in the market very soon now.