Just by chance, we have recently looked at the contents page of our Storage section and were very ashamed to see very few articles devoted to SCSI solutions. We have to assure you that we don't belong to SCSI haters and we didn't tackle this topic only because we were afraid that we wouldn't manage it since we lacked the experience. In fact, we don't think that we've acquired enough experience for a review like that since then, however, we got brave enough to try. :-)

The second thing that inspired us to test SCSI HDDs was the banner, we saw every day on one of our favorite sites StorageReview. Here it is:

Frankly speaking, we do respect Quantum storage solutions, however, this banner naturally attracted our attention and we got very curious to find out if it reflected the real state of things or not.

And finally the third reason why we decided to sit down to SCSI testing, is the newest Seagate solution - Cheetah 36XL. And in order to make it a fair competition and a fair analysis, we also took a closer look at some other SCSI hard disk drives. So, we tried to kill two birds with a stone… And if you wonder how we managed it, then go ahead and read on!

Testing Participants

First of all let's take a close look at the newest model from Seagate Cheetah Family - Cheetah 36XL:

Seagate Technology Company announced this drive on 27 November, 2000, which is quite a while ago already. However, it is only now that we managed to get our hands on it.

This HDD represents already the fifth (!) generation of Cheetah disk drives comprising all the currently available Seagate's technological achievements.

The Drive Defense is provided by Seagate's brand G-Force Protection and SeaShell systems. G-Force Protection implies system approach towards the development of the HDD basic components, such as engines, heads, head holders, platters and platter holders, so that to ensure high shock robustness of the device. Besides, all Seagate hard disk drives are shipped in a new SeaShell. It is a transparent clamshell drive container, which replaces the ESD bag and protects the HDD against external non-operating shock. According to the tests carried out by the leading OEMs, SeaShell, unlike the common soft ESD bag, reduces the shock by nearly 10 times when the HDD falls from the 30cm height.

Data Defense technology ensures the highest data integrity, which hinders data losses when working with system applications. The technology includes the following functions: Advanced Multidrive System (SAMS), Error Correction Code (ECC), Safe Sparing and End-to-End Data Path Protection.

Diagnostic Defense is provided by the following software: SeaTools, Web-Based Tools and Drive Self Test, which analyze various parameters of the system thus helping the user to diagnose and predict some potential breakdowns.

According to the specification, Cheetah 36XL HDD features 5.2ms seek time, 46.1MB/sec formatted data transfer rate and low power consumption - only 9W, which makes Seagate's new solution heat considerably less than the previous generation HDDs.

Very fast heads positioning as well as high spindle rotation speed guarantee (theoretically) perfect performance in server tests, and high data density should ensure good performance when handling large files.

Seagate SCSI HDDs have been always known for high reliability. We believe that the new model will once again prove that the company deserves this reputation. Especially, since the developers claim that Cheetah 36XL has 1,200,000 hours MTBF (Mean Time Between Failure), which is over two times higher than by any average IDE hard disk drive. In fact, reliability is one more important feature of SCSI HDDs besides fastness. It is exactly due to high reliability (which results from high quality components and impeccable mounting) that SCSI drives cost so much.

Cheetah 36XL family includes models with the following capacities: 9.2GB, 18.4GB and 36.7GB. That is why all users should be able to find a model, which will satisfy their needs.

Brief Technical Specs

In order to make our investigation an exciting read for you we made up our mind to invite the following strong participants:

• Seagate Cheetah 36LP
• Seagate Cheetah 18XL
• Quantum Atlas 10K II 36GB

Also just for your reference we included the results obtained for the following three HDDs:

• Seagate Barracuda 18XL
• Quantum Atlas 10K II 18GB
• Quantum Atlas V 18GB

Now we suggest taking a look at the most important specs of the today's heroes:

As you can see from the table given above, new Seagate hard disk drive features fewer platters due to higher data density per platter. As a result, it dissipates less heat and produces less noise during operation. Also the track-to-track seek time has doubled. We wonder how it will tell on the performance of this drive…

And now come the specs for Quantum HDDs:

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 ASC26160N SCSI controller;
• Windows 98/Windows 2000 Pro.

To test the HDDs we used Adaptec ASC26160N controller and the ordinary LVD cable. All the drives were tested in Ultra160 mode.

When testing we noticed that all HDDs with 10,000rpm rotation speed needed additional cooling that's why we used an extra cooler for them. However, now this situation forced us to give up the thermal tests.

For Adaptec controller we took the drivers version 1.00 for Win2000 and 1.11 - for Win98. We used FAT32 and NTFS file systems to format each of them 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:

• Windows 98:
  • WinBench 99 1.2
  • Adaptec Threadmark 2.0
• Windows 2000
  • WinBench 99 1.2
  • HDTach 2.61
  • IOMeter 1999.10.20

Access Time

The Average Access Time parameter seems to be much more important for SCSI drives rather than for IDE drives, which is determined by the application field for SCSI solutions. In our previous reviews we have already shown the evident connection between the average access time and the results obtained in IOMeter benchmark, which is the "most server benchmark" we have at our disposal currently.

As you can see the new Cheetah 36XL showed much better results than the previous Cheetah 36LP and Cheetah 18XL. Fewer platters and hence, higher sector density per track, save the heads time and trouble moving too far when they need to shift to another sector. In case it is still necessary to make a noticeable move to reach another sector, Cheetah 36XL acts a bit slower than its predecessor, which features shorter track-to-track time. This is one more consequence of higher data density by Cheetah 36XL: tracks have become narrower that is why the heads should be positioned more precisely, which can't be done as fast as before.

Cheetah 18XL showed higher (worse) time than Cheetah 36XL and lower (better) time than Cheetah 36LP, which corresponds to their parameters actually. As we have already mentioned, all the three models feature the same rotation speed and positioning speed, while the number of platters and hence the data density per platter differ.

The results shown by Quantum Atlas 10K II HDDs strike as outstanding: the time shown is just excellent and the performance meets the stated specs 100%. We really wonder how Quantum engineers managed to put together very high data density and beautifully fast heads positioning…

Among the hard disk drives with 7,200rpm rotation speed, Quantum Atlas V nearly ruined Barracuda 18XL, having surpassed it by about 0.5ms. Even though the specs of Barracuda 18XL look more attractive.

Data Density

In fact, we don't measure the data density per platter as it is, but the linear read speed. However, this value reflects the real data density per platter and the linear read graphs give us a perfect idea of the data density in the beginning and in the end of the drive and show how many sectors feature similar data density.

As we see, Cheetah 36XL appeared considerably ahead of its counterparts here and nearly caught up with Quantum Atlas 10K II.

Among the HDDs with 7,200rpm rotation speed, Quantum Atlas V again appeared ahead of all.

WinBench99 for Windows 98

The diagrams below show the results obtained for Business Disk WinMark and High-End Disk WinMark:

Even though Seagate Cheetah 36XL has almost leveled with Quantum Atlas 10K II in terms of data density, the latter managed to leave its competitor behind in these tests, especially in High-End Disk WinMark. It is most likely to owe this victory to its 8MB cache buffer, which is twice as large as that by Seagate Cheetah 36XL. We have always said that cache buffer can never be too large :-)

We would also like to point out that all the drives tested showed pretty modest results in Business Disk WinMark. Is it because of the data density, which is lower compared to that of the today's IDE-hard disk drives, or because of the poor controller drivers optimization?

Adaptec Threadmark 2.0

Here is what we obtained in this benchmark:

New Seagate Cheetah 36XL proved here much faster than its predecessors, running 25% ahead of Cheetah 36LP and Cheetah 18XL. However, at the same time it fell about 22% behind Quantum Atlas 10K II.

Surprising as it might seem, but Barracuda 18XL fell even more behind its competitor: almost 60%!

By the way, we didn't see any linear dependence on the data density here…

HDTach 2.61

First, we would like to offer you all the results obtained in this benchmark:

Graphs	Graph	Graph	Graph	Graph	Graph	Graph	Graph

However, we are going to focus only on Read Average and Write Average, as they seem the most informative to us.

Well, the situation seems to have changed drastically. Our today's hero is ahead of all. Even Barracuda 18XL has got quite close to Quantum Atlas V.

If the average write speed of Cheetah 36XP is only 5% higher than that of Atlas 10K II, then the 12% superiority in the read speed seems a really good achievement to begin with. :-)

Also we would like to draw your attention to Read Burst. It shows how quickly the HDD can return the data from the cache buffer:

Hm… We were pretty surprised to see Ultra160 SCSI HDDs show results like that. Especially, since we usually got about 93-94MB/sec in case of IBM DTLA 307015 with Promise Ultra100 controller.

WinBench99 1.2 FAT32

Well, the results obtained are quite interesting. The winner in Business Disk WinMark is Seagate Cheetah 36XL, which easily beats its predecessors, Cheetah 18XL and Cheetah 36LP, and the winner in High-End Disk WinMark appears Quantum Atlas 10K II.

Note that the overall performance shown by all testing participants in Windows 2000 appeared almost twice as high as that shown in Windows 98. Since we didn't observe the same difference when we tested IDE hard disk drives, we think we have every reason to blame "raw" Adaptec controller drivers for that. In fact, this is quite natural: what's the use of polishing the Win98 controller driver, if the SCSI hard disk drives are usually used under WinNT or Win2000.

WinBench 99 1.2 NTFS

Here we made a very interesting observation. When we shifted to another file system (maybe it is a different cluster size that matters?) Seagate Cheetah 36XL didn't run ahead of Quantum Atlas 10K II any more, lagging about 6% behind instead.

Intel IOMeter

Yes! Finally we will use this benchmark in the proper way: for SCSI-HDDs evaluation. If you are unfamiliar with the principles and methods used in it, please check our previous reviews, where we have already discussed it.

To tell the truth, this benchmark generates a lot of numbers and results. However, in order not to overload you with them, we suggest concentrating on Total I/O parameter, showing the amount of transactions made. The sum total is calculated taking into consideration the results obtained with Light, Moderate and Heavy loads. Just for a better comparison we added here the best results shown by some IDE HDDs tested before, namely Seagate Barracuda ATA III and Quantum FireBall Plus AS.

Well, smaller Average seek time by Quantum Atlas 10K II results into the advantage of 100 points over the competitors in FileServer mode. And the results shown by all Cheetahs appeared very close to one another.

In WorkStation mode, faster heads positioning of Atlas 10K II doesn't help it retain leadership. Its place is now occupied by Cheetah 36XL.

In Database mode Cheetah 36XL fell a little bit behind the previous Cheetah models and a great bit behind Quantum Atlas 10K II. Actually, this is pretty strange since Database pattern differs from WorkStation only by higher percentage of write operations (33% against 20%). And as we have already seen in HDTach, Cheetah 36XL copes with write operations just excellently…

The matter probably lies with the heads of Cheetah 36XL, which are switched into the write mode too slowly. The average track-to-track (write) time makes only 1.5ms by Cheetah 36XL against 0.9ms by Cheetah 36LP.

Anyway, the gaps between the HDDs of one and the same class aren't that large. Quantum Atlas V and Seagate Barracuda 18XL appeared very far behind the drives with 10,000rpm rotation speed, while being almost twice as fast as the drives with 7,200rpm. Well, this is an excellent proof of the fact that IDE hard disk drives are absolutely unsuitable for hard server work.

Conclusion

All in all Seagate Cheetah 36XL made a very good impression. It proved a worthy successor of the Cheetahs family and showed that the evolution of Seagate's SCSI HDDs has taken the right way. Fewer platters used in the HDD construction allowed to improve the acoustics, reduce heat dissipation (which we didn't manage to check, unfortunately) and will hopefully tell positively on the prices as well. The server market keeps growing day by day in a very high tempo that is why reliable and fast storage solutions will not remain undemanded.

Seagate Cheetah 36XL represents an excellent combination of high data density and fast mechanics, which makes this solution somewhat universal. It can be used in high-performance desktop PCs as well as in mainstream servers (which are the most numerous today). And if some of you aren't satisfied with its performance, there will be always a way-out: to create a RAID array… Wait! This is another article already :-)

In conclusion we would like to return to banner ads. We believe that Quantum has every right to be proud of its Atlas 10K II HDD. Even though it didn't win all the races, it proved one of the best among 10K drives.