In the end of October Seagate Company held a press conference in Moscow. After an introductory speech of Ian D. O'Leary, Senior Corporate Communications Manager, the journalists listened to Bernard Eisman's (Product Marketing Manager) report about desktop storage devices, that is, about IDE drives, and to Shawn F. Hook, Senior Marketing Manager, who spoke about SCSI drives and upcoming HDD interfaces.
Bernard Eisman introduced to the Russian journalists the current IDE HDDs market trends and the position of Seagate in this market. He also described in detail every hard disk drive currently manufactured by the company. Unfortunately he didn't fall for our provocative request to talk about future products. :)
Among current Seagate drives, the Barracuda ATA V (you can read our Seagate Barracuda ATA V Review) got most attention.
Bernard Eisman also voiced the Seagate's intention to give up active promotion of 5,400rpm HDDs. Well, it's no secret that the users' interest to these drives is lowering as well as the difference in prices between 5,400 and 7,200rpm drives. IBM discontinued 5,400rpm HDDs a year ago, now Seagate is going to leave this market sector, too. There is no catastrophe here: just the usual consequences of inevitable progress…
Similar trends can be observed in the SCSI HDD family from Seagate. 7,200rpm drives become a thing of the past. The Barracuda 36ES2 will be replaced with Seagate SATA drives, while the Cheetah 10K.6 will be a natural replacement for the Barracuda 180, as it has a little lower maximum capacity (146GB against 180GB), but smaller dimensions (it's a standard-size 3.5" HDD of 1" height) and much higher speed.
It's all even more interesting with interfaces. Seagate names SerialATA a priority interface for high-end IDE drives. The company thinks the future of IDE HDDs is directly connected with it. The ATA interface won't be forgotten, though, at least during the transition period. But it would be used in low-end models only, so Seagate fans should already start thinking about upgrade options.
Seagate seems to be in doubt so far about the future standard interface for enterprise-drives. On the one hand, Seagate won't use Ultra640 SCSI and will concentrate on Serial Attached SCSI, but if the market thinks differently…
So, the hero of our today's review is the new hard disk drive from Seagate: Cheetah 15K.3. Its name suggests that it belongs to the third HDD generation from Seagate with 15,000rpm spindle rotation speed. Seagate quite appropriately points out in the Cheetah 15K.3 press release that this is the only 15,000rpm drive in the world to boast the proven reliability of three generations in a row. Yes, Seagate is the only manufacturer on earth to have three 15K drives generations on its record list: Cheetah X15, Cheetah X15 36LP and the last one Cheetah 15K.3.
But the number of models launched is not the only trump of Seagate. The sales statistics of 15K HDDs are much more impressive. According to it, every two of the three sold 15K HDDs are manufactured by Seagate! The customers vote for Seagate with their money!
So, what's so good about this cutie?
Compared with the last generation drives, the platter of the Cheetah 15K.3 has twice as high data density: 18GB against 9GB in the Cheetah X15 36LP. Combined with the traditional four-platter design of SCSI HDDs from Seagate, the maximum capacity of the Cheetah 15K.3 is 73GB.
Cheetah 15K.3 comes with Ultra320 SCSI or 2Gb Fibre Channel interfaces. Its cache-buffer is 8MB big. Cheetah 15K.3 model line includes three drives of 18GB, 36GB and 73GB storage capacities.
Cheetah 15K.3 HDD is the first drive of the 15K generation to come into our test lab (we would also like to test later on the Fujitsu MAS and Maxtor Atlas 15K). So, unfortunately, we can only compare it with "older" drives only. So, Seagate Cheetah X15 36LP, Fujitsu MAM and IBM 36Z15 will be the opponents to the Cheetah 15K.3.
The specs of the hard drives tested are listed in the following table:
Testbed and Methods
For WinBench tests the arrays were formatted in FAT32 and NTFS as one partition with the default cluster size. The WinBench tests were run five times each; the average result was taken for further analysis. The HDDs didn't cool down between the tests.
We used the following benchmarking software:
To compare the hard disk drives performance in Intel IOMeter we used the StorageReview patterns described in the third edition of their HDD testing methodology.
These patterns are intended to measure the disk subsystem performance under workloads typical of file- and web-servers.
Our colleague, Sergey Romanov aka GreY, developed a WorkStation pattern for Intel IOMeter basing on the StorageReveiw's study of the disk subsystem workload in ordinary Windows applications. The pattern was based on the average IPEAK statistics StorageReview provided for Office, High-End and Bootup work modes and mentioned in Testbed3 description.
The pattern serves to determine the attractiveness of the HDDs for an ordinary Windows user.
Well, and in the end we checked the ability of the drives to work with sequential write and read requests of variable size.
Our testbed was configured as follows:
- ASUS P3B-F mainboard;
- Intel Pentium III (Coppermine) 600MHz CPU;
- 2 x 128MB PC100 ECC SDRAM by Hyundai;
- IBM DPTA 372050 HDD;
- Matrox Millennium 4MB graphics card;
- Windows 2000 Pro SP2.
To connect the hard disk drives we used Adaptec 29160N controller card with BIOS version 3.10.0 and drivers version 4.10.4002.
The reviewed drives had the following firmware versions:
- Seagate Cheetah 15K.3 - FW: 0002
- Seagate Cheetah X15 36LP U320 - FW: 0022
- Fujitsu MAM 18GB - FW: 0106
- Fujitsu MAM 36GB - FW: 0106
- IBM 36Z15 18GB - FW: S7Z0
Our first test is HDTach from TCDlabs.
The most important of HDTach results are average read/write speeds and the read burst speed.
The new Seagate's product, the Cheetah 15K.3, easily leaves all the others behind. The doubling of the data density per platter compared with the previous generation allows our hero to do this without much effort.
WinBench 99 2.0
We will start with FAT32. First, we would like to offer you a complete table of results achieved in WinBench.
Do you see it? Cheetah 15K.3 processes large files much faster than Cheetah X15 36LP! Of course, it's entirely due to the higher linear read speed of the new solution. By the way, since we mentioned the linear speed:
Not bad, eh? Cheetah 15K.3 outran Cheetah X15 36LP (that is, the previous generation drive from Seagate) by 26%! Note also that the minimum read speed of Cheetah 15K.3 (in inner, less dense tracks) is higher than the maximum read speed of the IBM 36Z15!
Now, let's see the diagrams for two sub-tests of WinBench99 package: Business Disk WinMark and High-End Disk WinMark:
Yes, Seagate Cheetah 15K.3 showed its superiority here, too! But its advantage over the second fastest Seagate Cheetah X15 36LP is only 13% in High-End Disk WinMark: twice as small as the advantage in linear read speed.
The HDDs from Fujitsu showed rather strange results. A drive with higher storage capacity is usually a little faster in benchmarks (due to higher capacity of the cylinder), but this time the situation is just the contrary. The junior model is faster, although by 1% only. Let's consider it a measurement error.
Now it's high time for NTFS results.
The two HDDs from Seagate remained ahead of their rivals in another file system as well:
Let's begin with sequential read and write. The HDD is being "bombarded" with a stream of requests to read/write data blocks. The size of the blocks is increasing. The request queue depth is fixed at four out-coming requests. The table below shows the dependence of data transfer rate on the size of the requested data block:
The table is for those who think in numbers. For people who think in images, here it is:
So, we see that Seagate Cheetah X15 36LP does well in case of small data blocks. But it seems to have some problems with 8KB ones. Well, maybe it just doesn't like this size. :)
Seagate Cheetah 15K.3 has no problems at all, although it processes smaller data blocks slower than Cheetah X15 36LP. It shows its maximum performance starting from 32KB blocks.
Now, let's see what we have with writing:
Both drives from Seagate performed about the same in case of small blocks. Note also that Cheetah X15 36LP is again in trouble with 8KB blocks. Cheetah 15K.3 behaves solidly and wins the test…
Having checked the drives performance with streaming data, let's now turn to quite another mode: simulation of disk subsystem workload typical of a database.
The data block size is 8KB, the address of the block is taken at random, the read/write operations ratio is changing from 100/0 to 0/100 with the stepping of -10/+10.
The results of this test are shown in three workload modes: with request queue depth equal to 1, 16 and 256.
Fujitsu MAM HDD proved the best under such a small workload. Cheetah 15K.3 was close to the leader, but proved faster only in two end modes (Random Read and Random Write).
Things change when the request queue length becomes equal to 16. Cheetah 15K.3 is ahead at any read/write operations ratio, although the Fujitsu's HDD is close behind.
If we examine the curve of the IBM HDD, we will see that it is quite different from what other drives have shown. It's "concave", i.e. the drive gains its maximum speed in the ends of the curve only. The drives from Seagate and Fujitsu have "convex" graphs, i.e. they are at the maximum performance when the read and write operations shares are equal.
We suppose that IBM 36Z15 owes this behavior to its smallest cache-buffer of all the drives tested (4MB). So, it just doesn't have enough memory to effectively perform lazy writing.
Anyway, the IBM drive was good in the Random Read mode. Remember this fact!
After the workload grows up to 256 requests, the Cheetah 15K.3 breaks away from the competitors and leaves them far behind. Its advantage in data density must be very useful here, as the probability to find the data for two requests in a row on the same track increases. So, the drive moves the heads less or moves them over a short distance (the so-called short seek).
So, we learned from the DataBase pattern that Seagate Cheetah 15K.3 likes "big" workloads.
It's time to check the drives' performance in server patterns:
Well, the results are curious. Firstly, we should note that Cheetah 15K.3 did very well, although there is one weak spot. Look at the low performance growth on transition from 16 to 64 requests workload. Yes, Cheetah 15K.3 performs not very well (compared with the competitors) when the request queue depth is equal to 64.
Another problem happened to the IBM 36Z15 drive. It suddenly slowed down in the FileServer pattern at 256 requests queue depth. The cache-buffer must have been too small again.
But in the Webserver pattern the IBM 36Z15 feels all right. In this pattern the HDD receives only read requests and we remember that the IBM drive showed excellent result in the RandomRead mode. So, we shouldn't be surprised at its high performance here.
In order to compare the "average" performance of the hard drives, let's calculate the average of the Total I/O values under all workloads (we assume that all workloads are equiprobable for a server). The ratings we got this way are shown below:
Here we are! Seagate Cheetah 15K.3 lost to both Fujitsu drives because of its failure in case of 64 requests. IBM 36Z15 was even unluckier as it did poorly in the most "profitable mode with 256 requests ("profitable" from the point of view of contribution to the final result). As a result, it ended far behind the others.
But the IBM drive wins the WebServer pattern! Not like we have expected anything like that, eh?
Yes, we see very intriguing results. But believe us, we were not pushing the drive from IBM ahead in any way.
Unfortunately, the renowned Web resource on hard disk drives, StorageReview, didn't test IBM 36Z15 of 15GB capacity. They had a 36GB model and it didn't prove fast in tests. But we know that the junior drive in the family from IBM is often faster than major models (see our IBM Ultrastar 73LZX 9GB Hard Disk Drive Review).
The trick is simple: the junior models are made of more platters than necessary (by economical considerations). The tracks are then spaced out, that is, the "crosscut" density of the platter is reduced. The access time becomes lower, thus accounting for higher speed of junior models.
Moreover, the formula we use to calculate the ratings favors those drives that show high results under high workload (in case of big request queue depth) and all multi-platter HDDs do so. As IBM 36Z15 18GB combines low access time and multi-platter design, it had its chance and used it to the full.
So, we have one more pattern left. It emulates a disk subsystem workload, typical of a workstation.
The HDDs ratings in the WorkStation pattern are calculated differently than those in FileServer and WebServer patterns. As a workstation is more likely to generate smaller workloads, the Total I/O values under small workloads have a bigger weight in the end sum. Here's our suggested formula to calculate the total result:
Overall = 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 + Total I/O (queue=32)/32
Here are the results we obtained in this formula:
If we put the numbers aside, the new generation of 15K hard disk drives from Seagate has one important advantage over the drives of the previous generation. It's the capacity! Thanks to higher density of the platters, the new drives can boast a better capacity-to-price ratio. This way they reduce dramatically the total cost of disk subsystems that demand high capacity as well as performance.
The wide range of models (18, 36 and 73GB) in the Seagate Cheetah 15K.3 family allows choosing most suitable drives (capacity/quantity) for a given task.
You shouldn't be confused by not very high results of the Cheetah 15K.3 in our tests. Our drive had a peculiar firmware number - "0002" (StorageReview had "F501" firmware drive). We can't claim that "F501" is better than "0002", but…
But even the results of our drive prove that:
- Seagate Cheetah 15K.3 is ragingly fast in WinBench tests,
- Seagate Cheetah 15K.3 is excellent at processing sequential requests,
- Seagate Cheetah 15K.3 was the fastest in the DataBase pattern (under considerable workloads),
- Seagate Cheetah 15K.3 was good in server patterns,
- Seagate Cheetah 15K.3 won the WorkStation pattern.
By the way, you shouldn't also be confused by our talking so much about the results of this server-natured drive in the WorkStation pattern. One of the main criticisms to SCSI drives is their noise. That's why they are not very welcome in high-end workstations. But as Seagate Cheetah 15K.3 seems to be using fluid dynamic bearing, the noise from its spindle is subjectively not louder than in any ordinary IDE drive. Of course, there is also some noise produced by the moving heads, but it's not all bad here. During the tests we noticed one curious thing: under high workloads, the HDD nearly stopped producing any noise altogether! A dream? No, it's a feature!
Under high workloads, the request queue depth is big and the HDD controller re-arranges the commands so that the read/write heads would move as little as possible. In other words, instead of mindlessly running the heads from outer to inner tracks and back again, the HDD makes "short steps" and produces no soul-disturbing sounds.
Overall, we have a rather unnatural situation: the higher workload, the quieter is the HDD!
P.S.: The test results of Seagate Cheetah 15K.3 with the U320 SCSI interface will be discussed in one of our upcoming reviews.