As you know, SCSI HDDs with 7,200rpm spindle rotation speed are beginning to feel the push of ordinary (and extraordinary) IDE-disk drives. The imminent transition of IDE drives to the new SerialATA interface (with higher 150MB/sec bandwidth) may become the end of SCSI 7,200rpm as one of their strongholds, Ultra160 (160MB/sec) interface, will have an adequate opponent now.

In such unfavorable situation the Seagate Company (and today Seagate only continues to produce 7,200rpm SCSI-disk drives) made a kind of move of the knight.

As a modern HDD must be both fast and convenient-to-use, Seagate engineers laid the emphasis onto the "noiseless" work. They had sufficient experience in this field to rely upon. Seagate started to experiment with fluid bearings as far back as Barracuda ATA III. By the launch of Barracuda ATA IV the FDB-motor technology had been brought to perfection. The noise from working Barracuda ATA IV HDD didn't exceed 2.4bel. Such noise level is not perceptible to the human ear and I immediately checked it. :)

And now that bearing, the motor-bearing block, to be more precise, with the sounding name SoftSonic is used in a SCSI HDD (As you already understood from the title of the review, it's the new Barracuda 36ES2).

Like the previous 7,200rpm SCSI HDD from Seagate, Barracuda 36ES (see our Seagate Barracuda 36ES HDD Review), the Barracuda 36ES2 Winchester disk is "based" on the construction of the Seagate's ATA HDD. As some of the parts are unified, the whole HDD manufacturing cost has been reduced. But it would be wrong to say that Barracuda 36ES2 is just the same Barracuda ATA IV but featuring SCSI electronics! If we give Barracuda 36ES2 a closer look (we'll do it a little later), we'll see that there are a lot of important differences between Barracuda 36ES2 and Barracuda ATA IV. Especially, if we consider their model names. :)

Testing Participants

First, let's see the pictures of Seagate Barracuda 36ES2:

   

It's clear that Barracuda 36ES2 inherited its case from Barracuda ATA IV. The only visible difference between the two HDDs is the shape of the SeaShield. Barracuda 36ES2 has openings for cooling the warmest electronic parts while Barracuda ATA IV has none of the kind.

Let's check the Barracuda 36ES2 specifications:

Note that the 36GB model from Barracuda 36ES2 family "is built" in a totally different way than that of Barracuda 36ES one. While 18GB and 36GB models of Barracuda 36ES feature different number of platters/heads, Barracuda 36ES2 18GB and 36GB use of the same number of platters and heads!

It means that the 18GB Barracuda 36ES2 model has a "narrower" working zone than the 36GB model. One of the effects of it is that 18GB Barracuda 36ES2 model has shorter average seek time than the higher capacity model (it's reflected in the specs).

As you may have guessed already, today, along with Barracuda 36ES2, we are reviewing Barracuda ATA IV (20GB) and Barracuda 36ES, the previous 7,200rpm SCSI model from Seagate based on the Barracuda ATA III design.

Testbed and Methods

We used the following testbed:

• Intel Pentium III (Coppermine) 600MHz CPU;
• ASUS CUBX-E mainboard (1007A BIOS);
• 2 x 128MB PC133 SDRAM by Hyundai;
• Matrox Millennium 4MB graphics card;
• Adaptec 29160N SCSI controller;
• Windows 2000 Pro/SP2.

To test the hard disk drives we used Adaptec ASC 29160N controller. All HDDs were tested in Ultra160 mode. For Adaptec controller we used 1.00 drivers for W2K.

We used the following software:

• HDTach 2.61;
• WinBench 99 1.2;
• Intel IOMeter 1999.10.20.

IOMeter and HDTach were run for an unformatted HDD. For WinBench tests the HDDs were formatted in FAT32 and NTFS as a single logical drive with the default cluster. The benchmarks were run four times each; the average result was taken for further analysis. The HDDs didn't receive any time for cooling between the tests.

Performance

Average Access Time

If we consider the specs, the win should come to Barracuda 36ES2:

Well, that's exactly what happens! The average access time of Barracuda 36ES2 is even less than the calculated one (6.9+4.17=11.07)!

Barracuda 36ES is 2ms behind the leader. Barracuda ATA IV is the last one.

Sustained Linear Read Speed

It was evident that Seagate would "cut off" the less dense inner part of Barracuda 36ES2 platter. But the way it was done just shocked me:

The oldest HDD of the three features the "densest" platter in the beginning of the disk while the newest Barracuda 36ES2 was only the third in terms of the linear read speed in the beginning of the disk. But look at the read speed in the end of the disk: the picture is just the opposite! Barracuda 36ES2 is the fastest here: 38MB/sec. This is incredible speed!

Interesting that I have never encountered such beginning/end read speed ratio (1.07) in any HDD. So it deserves being called the first record.

If we look at the Barracuda 36ES2 linear read speed graph drawn with the help of WinBench software, everything becomes clear:

Well… They did cut the platter of 18GB Barracuda 36ES2 "very deep"… But if it leads to drastic reduction of seek time, what's the use of crying about the cut-off gigabytes. :)

HDTach 2.61

"Only constant and conscientious training in the
martial arts can ensure long and happy life".
Bruce Lee

As the beginning/end read speed ratio of Barracuda 36ES2 is much lower than the speed of the other contestants, the HDTach test becomes mere formality. But attending to formalities is not useless sometimes (see the epigraph).

As you might have expected, the Barracuda 36ES2 HDD is best everywhere: in average read speed, average write speed and maximum read speed from the buffer.

WinBench99 1.2 (FAT32)

Well, let's watch Barracuda 36ES2 performance in WinBench:

The first disappointment: Barracuda 36ES2 loses to its ATA-predecessor both in Business and High-End tests. Interesting that Barracuda 36ES2 was ahead of Barracuda 36ES in Business test, but loses to it in High-End test.

Well, can it be that Seagate software developers have been fooling away for the past nine months? Can't believe.

WinBench99 1.2 (NTFS)

Well, well, well…In NTFS the new Barracuda 36ES2 was ahead of the old one in both Business and High-End tests, but anyway Barracuda ATA IV is faster (and much faster, actually!). It can only suggest that the Barracuda 36ES2 firmware was definitely optimized not for WinBench99.

Intel IOMeter

IOMeter tests are going to take the good half of the review so if you are not familiar with the test and test methods you'd better read a few corresponding articles first.

To all IOMeter experts I would like to offer the results shown by Barracuda 36ES2 in three standard patterns: FileServer, WorkStation and DataBase.

But as the results of a single HDD are less informative, I will provide the table with the results of the three tested HDDs (Total I/O only, but in a lot of patterns).

As you see, Barracuda 36ES2 HDD is always faster than Barracuda 36ES under small workload and is much faster than Barracuda ATA IV under any workload.

In order not to obscure the point with a lot of details, I will only show the diagrams for the first three patterns:

It's clear that with small request queue depth Barracuda 36ES2 is beyond any comparison. After the request queue depth has been increased to 256, Seagate Barracuda 36ES advances to the fore.

Barracuda 36ES looks better in this pattern, but with the request queue depth below 64, Barracuda 36ES2 is faster anyway.

In the pattern that imitates the work of the HDD with a database both SCSI HDDs are "riding high". It is important that Barracuda 36ES2 is again faster than Barracuda 36ES. Well, knowing the performance difference between the heads blocks of these HDDs, it could have been easily predicted :).

Then I was curious to see how the speed of IDE and SCSI HDDs from Seagate depends on the request queue depth and the read/write operations ratio.

We made a pattern that uses 100% random 8KB data blocks and the read/write ratio is changing from 100/0 to 0/100 with the step of 10.

The results we got are listed in the table (the horizontal header of the table indicates the write operations share, the rows are marked with the request queue depth values).

Interesting that there is no big difference between the results of the Barracudas at queue=4 and queue=16 while by the ATA HDD the situation is completely different.

Three diagrams will help us to compare the HDDs' behavior in this pattern:

In this mode the maximum gap between Barracuda 36ES and Barracuda ATA IV takes place in RandomRead and RandomWrite while in the middle of the graph where read and write operations shares are about the same, the results of the HDDs get close.

Increasing the request queue depth drastically changes the picture. The ATA disk drive remained indifferent to the increase in the write operations share while the SCSI HDD is speeding up as the writes share grows (is it due to the inborn ability of SCSI drives to command queuing and reordering?).

At such request queue depth the SCSI HDD seems to be at the end of its capabilities as well as its ATA counterpart, though.

Interesting that in this mode the speeds of the disk drives are close in RandomRead and RandomWrite and the biggest gap is achieved when the read and write operations number gets equal.

In order to check the influence of the interface on HDD performance we conducted one more experiment. We made patterns with sequential read/write of data blocks, which size increased from 512Bytes (the size of the sector) to 1MB.

The tables below show the dependence of the performed requests, data transferred per second and CPU utilization in this mode on the size of the data pack (the request queue depth was the same for all the HDDs and equaled 1 request).

The first thing to be mentioned is the amount of requests processed by ATA and SCSI HDDs when working with 512Byte blocks. The SCSI HDD is only 2.5 times faster! Of course, it could be argued that the controllers are "to blame" for this results gap (Adaptec 29160N and Promise Ultra100 were used in the tests), but it's impossible to measure HDD performance without the controllers :).

In my humble opinion, the gap in the results was caused by the differences in their interfaces. Higher bandwidth of the Ultra160 interface allows Barracuda 36ES2 to perform fast even when the useful information and system information ratio in the data transferred via the cable is far not favorable.

By the way, the CPU utilization compared with the amount of transferred data is much better by the SCSI HDD.

The only thing left is to consider the HDDs performance during write operations:

On the whole, the picture is the same. The SCSI HDD is faster and loads the CPU notably less.

Temperature

Although all decent SCSI HDDs have a built-in thermal diode, it can only notify us of the critical HDD temperature level reached (a decent controller in this case commands to disconnect that HDD from the bus). Unfortunately, it doesn't report the current temperature of the drive. That's why we measured the temperature by means of an external infrared thermometer after 2.5 hours of continuous active work in IOMeter benchmark.

In such conditions, the HDD warmed up to 43^oC. It is usual temperature for 7,200rpm HDDs.

As Seagate HDDs have always been "hot-proof", I don't think an additional cooling is necessary. The fans make noise and Seagate is combating it! Let's not interfere. :)

Conclusion

The tested Seagate Barracuda 36ES2 hard disk drive left, as often happens, an ambiguous feeling.

On one hand, it's certainly the fastest and most advanced 7,200rpm SCSI HDD. On the other hand, I'm used to making the final conclusion about the advantages of a HDD for desktops basing on the results of WinBench. And it's in this test that Barracuda 36ES2 proved not so fast.

The tests tend to become out-dated very quickly nowadays, so let's remember all the facts we know about Seagate Barracuda 36ES2 and evaluate it subjectively.

Barracuda 36ES2 features very fast mechanics, a rather dense platter, excellent linear read graph, and fast and smart interface. One more distinctive feature of this HDD is its noiselessness. The noiselessness is not as "deep" as by Barracuda ATA IV, because the heads of Barracuda 36ES2 are moving quite fast and, naturally, do produce some noise. But in the idle mode it's impossible to tell Barracuda 36ES2 from Barracuda ATA IV by the sound they produce.

Summing up all the advantages we may take a guess at its target field: computers that work with video and "heavy" applications or low-end servers. Wait, that's exactly where Seagate is positioning this solution!

One of the possible disadvantages of this HDD may be its price. But as I don't know their retail price so far, let's not anticipate.

P.S.: Seagate claims that the next generation of SCSI HDDs for high-end workstations will have the spindle rotation speed of 10,000rpm.