Frankly speaking, when we were writing the previous roundup we couldn't even suppose that the developmentof IDE hard disk drives manufacturing technology would go full bat. We do understand why all manufacturersstarted integrating Ultra ATA/66 support - they tried to be in the fashion. But besides that, the data densityshot up every month. Of course, we can ask: "And who cares about 37GB HDDs besides the people dealing withdigital video?" We can also complain about the effort we have to make in order to find a smaller HDD (sincethe data density has exceeded 6GB per platter, disks with smaller capacity will be hardly manufactured at all).But no one will deny the reality: there is another generation of IDE harddisk drives. And in the current roundupwe will try to take a look at a few innovations.

We reviewed some products provided by the following manufacturers:

• IBM
• Maxtor
• Seagate
• Western Digital

And in the near future we are going to add another bit on Fujitsu and Quantum as well.

IBM

It has already become a tradition to consider two main families: Deskstar 37GP (5400rpm) and Deskstar 34GXP(7200rpm) with the letter code DPTA. When we were working on the roundup we had the following models at our disposal:DPTA-353750/ 353000/ 352250/ 351500 (5400rpm, 37.5/ 30/ 22.5/ 15 GB correspondingly) and DPTA-373420/ 372730/ 372050/371360 (7200rpm, 34.2/ 27.3/ 20.5/ 13.6 GB). Compared to DJNA family there was only one revolutionary change introduced:two youngest models 352250 and 351500 got a smaller buffer. Its size was reduced from 2MB down to 512KB. And the restremained the same:

• Giant magnetoresistive - GMR - heads
• Vibration protection system - TrueTrack Servo
• Sectors formatting system - No-ID
• Drive Fitness Test (DFT)

The company preferred evolutionary methods of data density increase: they managed to achieve 7.5GB per platter atspindle rotation speed equal to 5400rpm and 6.8GB per platter - at 7200rpm.

We tested the eldest models: DPTA-353750 and DPTA-373420. We have to admit that IBM Company appeared slightly aheadof time having produced the disks of this size. The thing is that the maximum allowed disk size for most today's BIOSversions is only 33.6GB (as you see, not only we underestimated the HDD manufacturers). That's why if your super cool37GB HDD you are so proud of provides you only with some wretched 33GB, don't be in a hurry to blame the seller for that.The best thing for you to do is to look for a new BIOS version, replace your old BIOS with it and enjoy: no more problems.Besides that IBM harddisks prepared no other surprises, even the outer look remained as it had been before.

Maxtor

We can't say that we've had a lot of products by this company to review. We hope things will change for the betterin the future, and in the meanwhile we are glad to introduce Diamond Max 6800 family. It includes harddisks of thefollowing capacities: 27.2, 20.4, 17.0, 13.6, 10.2, 8.4 and 6.5GB. It's pretty interesting that 10.2 and 8.4GB disksare equipped with the same amount of platters/heads - 2/3 that is why in the first case the data density makes 6.8GB perplatter (one of the two sides isn't used) and in the other case - only 5.6. Why so?

As for the technological solutions introduced in the models of this family, we would like to point out two things:

• Antishock protection in the form of ShockBlock technology, which helps to reduce the tendency for the headto lift from the disk during a shock load and to eliminate the most common forms of damage, which may occur ifthe head does lift and then touch the platter. The company announced that during the testing the disks could inmost cases withstand short-duration shock impacts up to and over 1,000G! Read more about this technology here.
• Proprietary DualWave architecture increases the controller performance due to the fact that all functionsare now distributed between two processors: a high performance Texas Instruments DSP (Digital Signal Processor)that controls the actual disk functions (servo, reads and writes, ECC, etc.) and a Maxtor designed 50 MIPs RISCprocessor that is optimized for storage I/O and is dedicated to processing host commands. Read more about thistechnology here.

Actually, seems quite interesting, although Maxtor slightly yields in data density to another participants ofour roundup. Though the spindle rotation speed is 5400rpm, the data density is equal to that of harddisks rotatingat 7200rpm.

Seagate

Barracuda ATA family is a real dash forward for Seagate Company into the IDE HDD market. And to tell the truth,this attempt strikes as quite impressive. Just take a look. They offer models with capacities of 6.8, 10.2, 13.6, 20.4and 28GB. The spindle rotation speed is 7200rpm. The data density - 6.8GB per platter. These disks can withstand shockimpacts of up to 300G for 2ms in off state (just for a better comparison: IBM can boast only 200G, while Maxtor announcedup to 1000 though didn't mention the exact time period). Besides, they also introduced their super system of data integrityprotection SeaShield, which includes the following components:

• The electro-static protection is carried out by means of a pretty plate called Soft SeaShield Cover, whichcovers the PCB at the lower side of the disk. And to save room, there is a detailed installation instruction onthe plate surface.
• DiskWizard installation utility. DiskWizard on a floppy disk accompanies every HDD. It's a very nice programwith a very simple and understandable animated interface, which serves to help a beginner to install his new HDD.The program allows determining the place of installation (master/slave and channel), making a partition table,formatting and copying the info from the old disk to a new one.
• Of course, S.M.A.R.T. level III and Safe Sparing system, which proactively identifies and eliminates defects.
• It has already become pretty popular to provide some special means, which allow reducing the number of pseudodefectivesector returns. The system consists of Drive Self Test (DST) disk diagnostic tools included in the firmware and of a single-stepdiagnostics program, which can be downloaded from the company's web-site. It is quite convenient, actually, though in the firstplace it is intended for beginners (the first steps you are asked to carry out include checking the power and the cableconnection).

All in all, everything you may only need is all here, which is already becoming a standard.

We tested ST320430A model with 20.4GB capacity. We liked the way the disk looked. The exterior case was very similarto the old Medalist, and seemed to be thoroughly done. Let's take a look at it at work.

Western Digital

The situation here is very similar to that of IBM: we have the chance to watch the evolution of Expert family.And as for the main difference from the HDDs of the blue giant, it has totally disappeared: like IBM, WD is now alsousing SEC memory microchips. We managed to get a retail version of WD273BARTL model with 27.3GB capacity. The end-user,who buys it will get a pretty large colorful box with:

• Hard disk drive in plastic foam holders
• Rather thick multilingual installation manual (although no Russian :-( ). Even if you have never seen a PCbefore, you will manage to successfully install the HDD with this detailed manual.
• A small box with the most exciting content: a converter, which allows installing the disk into a 5-inch section,a small bag with bolts, Ultra ATA/66 cable, a floppy disk with Data Lifeguard Tools (including EZ-Install, diagnosticsand Ultra ATA/66 enabling/disabling software) and a telephone card to contact the technical support for free (which ofcourse, doesn't work in Russia).

So, the package is pretty funny, actually and some items are even very helpful (a converter and a cable).

Well, summing up all the technical parameters we composed the following table:

Tests

The testing system was configured as follows:

• Chaintech 6BTM (Intel 440BX) mainboard;
• Ultra ATA/66 Abit HotRod66 controller (HighPoint HPT 366);
• Intel Pentium II 400MHz CPU;
• 128MB SDRAM SEC PC100;
• ASUS V2740 (Intel 740) graphics card;
• Windows 98 SE.

The contents of all the disks tested was absolutely identical. We cloned it with Norton Ghost utility. And as for thetests, the set was rather standard: WinBench 99 (Business WinMark and Disk Inspection Tests) and Adaptec ThreadMark 2.0.Disk CPU Utilization benchmark may be carried out in two different regimes: with the fixed read rate and with the maximallyachievable one. All the benchmarks were run with the speed fixed at 4000000Byte/sec. We repeated WinBench 99 three times,and Adaptec ThreadMark 2.0 - five times so, the tables contain only average values. The Windows DMA support was enabled. Thebenchmarks were undertaken only for "cold" disk drives, i.e. we let the harddisk cool down till the room temperature beforestarting the next test. The hard disk drives were defragmented before each attempt with the help of Defrag utility includedinto the Windows 98 SE set. Since Ultra ATA/66 controllers are not so widely spread yet, we tested every HDD in two regimes:Ultra ATA/33 (the HDD was connected as the only master device to the first channel of the built-in mainboard IDE-controller)and Ultra ATA/66 (the HDD was connected the same way to HotRod controller).

So, to begin with take a look at the results obtained in WinBench 99:

Check the table below for more details. The numbers in brackets denote the CPU utilization during the test run. The firstvalue is given for Ultra ATA/33 and the second - for Ultra ATA/66 regime.

So, what have we got? First we would like to say a few words about HDDs with 7200rpm rotation speed. No one will besurprised to find out that the results shown by the disks from IBM and WD practically coincide, because the harddisks arealmost alike. The only fly in this ointment is the linear read rate graph for IBM DPTA. And there used to be such beautifultimes (with IBM DTTA, for instance) when there were hardly any notches (remaps) noticeable on it. The things with DJNA are abit worse now. But we have to pay for higher data density.

But as we come to Barracuda, everything appears much more interesting. We look at the graph obtained in Ultra ATA/66 regimeand... we see an ascending line! And even on top of it the transfer rate is far over the well-known limit of 33MB/sec.And in Ultra ATA/33 regime the situation is quite classical and the read rate in the beginning of the disk is absolutelydifferent. However, we got so greatly interested in this phenomenon that we decided to write to Seagate asking them tomake the matter clear for us. We received a very kind response from Seagate Disk Presales Representative Mr. Bart D. Hereis what he wrote:

"In theory there should not be a performance difference in the drive running on different IDE controller, as the drivedoes not break the 33MB/Sec barrier of Ultra ATA/33. However it has been observed that the Barracuda ATA will have aslightly slower Sequential Read Transfer rate on the Ultra ATA/33 bus. The reason for this has not been addressed as towhy this happens yet. One suggestion as to why this slower transfer rate may happen is the standard IDE cable, opposed tothe 80-conductor cable, is effecting performance or as newer faster drive hit the market, this may be a reoccurring sideeffect to running the drive on a slower ATA Bus. This phenomenon has not been found to effect random reads or writes. Sothere should not be any real performance loss to run the drive on a 33MB/sec bus."

Well, as for the last statement, we will probably agree to it, but concerning the cable... And as for our personalthough not absolutely ungrounded opinion (the arguments will be given further), it sounds as follows: unlike some othercompanies, Seagate designed a special controller directly for Ultra ATA/66 not caring about the way it would work withUltra ATA/33.

And now directly to the results.

So, Ultra ATA/66. The indisputable leader in terns of read rate from the platter and access time is Barracuda.And then the small buffer inevitably tells on the rates. The "fish" manages to succeed over its competitors only inapplications working with larger files, such as Photoshop, SoundForce. And in the rest it totally fails. So, this issimply another proof of the buffer size importance: it is as significant as the read rate from the platter.

Ultra ATA/33. Barracuda fails everywhere (well, this is the first argument backing up our opinion mentioned above).

And the Barracuda read speed graph looks just cool, best of all.

Now a bit about the harddisks with 5400rpm rotation speed. We think that everybody have noticed that we obtained someresults for Maxtor with a different controller - Promise FastTrak66 RAID Card (we are going to tell you a bit more aboutit very soon). And the reason for that was quite sad, to tell the truth. Having honestly run all the tests in ATA/33, wesat down to testing with HotRod. But then suddenly - oh god! - horrible things started happening! During High-End DiskWinMark 99 test both harddisks simply crashed the system dead. And no matter how the tests were combined: the crashoccurred at one and the same time as soon as we reached somewhere close to 57% done. The first thing that occurred tous was overheating. However, having taken the temperature of the CPU, mainboard and the HDD we didn't notice anythingabnormal. Then we disabled the mainboard controller just in case. But the drivers and system reinstallation also didn'thelp. So, we had to sacrifice the unification and to use a Promise controller. We didn't have any more problems (here)and managed to finish the test. But the most unpleasant thing about it was that there were real applications involvedin this test and some programs might simply not work in such a configuration. However, even if all the tests successfullyfinished, the results were still much worse than those obtained for ATA/33. And which is most striking, the worst resultsappeared exactly in those situations where all the other HDDs made the best use of all ATA/66 advantages: in applicationsworking with smaller files. We got the impression that the problem lied in work with the buffer. As for IBM, here everythingfollowed the traditional logic: they were slightly falling behind the others with a bit higher rotation speed. It appearedpretty interesting that when working with FrontPage, for instance, the fastness was almost independent from the rotationspeed: everything seemed to have moved to the buffer :-)

However, the best-looking graph belongs to Maxtor here.

All in all, the results seem much better compared to the previous generation IDE HDDs (see IDE HDD Roundup. Part 2),which is quite natural. The only strange thing we can't comment on is the results achieved in Adobe Premiere, which aresteadily better for ATA/33.

And now Adaptec ThreadMark.

The second column of the table below represents the average transfer rate in MB/sec judging by 64 benchmarks, whichincluded reading and writing the data into a number of blocks with the following size: 512Byte, 1, 2, 4, 8, 16, 32 and64KB. These operations were carried out through one, two, three or four threads simultaneously calling Win32 API in astandard way. And the third column shows CPU utilization during the benchmarks. As we have already mentioned, the firstvalue is given for Ultra ATA/33 and the second - for Ultra ATA/66.

The funny situation, we have already described in our HotRod Controller Review repeats here as well. All the disks butone proved much worse in Ultra ATA/66. And this lucky fellow - Seagate Barracuda ATA - performed just brilliantly! This isanother fact to prove our point about the disk controller, which was specially designed for Ultra ATA/66 regime. And besidesthat, nothing particular, only Maxtor disks prepared another gift for us: now they altogether crashed the system during the32nd test when working with... Promise controller! And with HotRod everything was just perfect. It seems really strange sincethe guys from HardwareReview have successfully tested the same harddisks with the notorious Promise controller. In other words,we are puzzled, but we'll try to find out what's going on.

Temperatures

The temperature of the hard disk drive was taken in the middle of Adaptec ThreadMark test by means of Motherboard MonitorLite v.2.1 utility. Yes, these disks warm up very good. Sharing our personal purely subjective impression, we can say thatwhen we removed the IBM one with 7200rpm rotation speed after running Adaptec ThreadMark (about 40 minutes of active work)we nearly burnt our fingers. And you should bear in mind that we didn't close the case during testing and besides, the roomtemperature was not so high then. That is why we would strongly recommend getting some additional cooling.

Of course the results are quite approximate, but they are very close to what should actually be: Maxtor heats least ofall. And among those rotating at 7200rpm Barracuda appears the most "cold-blooded fish".

Acoustic Characteristics

We would say right away that all harddisks are quiet. And the difference is so insignificant that you may disagreewith our positioning: the quietest ones are those from Maxtor, then comes Barracuda and IBM and WD are a bit louder.

Conclusions

So what is the outcome? The general conclusion will sound as follows: IDE HDDs made another step forward and with theirtoday's features they can easily compete with SCSI HDDs. And in terms of prices… Isn't it high time SCSI disks became cheaper?

And to be more precise, here is our opinion.

WD Expert is a good low-cost universal hard disk drive.

IBM is a combo of good speed and impressive capacity. Though some problems with BIOS may still turn up.

Seagate Barracuda ATA is the best choice for the applications requiring high data transfer rate from the platter (forthe digital video for instance). And we would strongly recommend getting an ATA/66 controller. And if you install RAID 0then...

Maxtor Diamond Max 6800 is a good disk to work with a standard ATA/33 controller. And as for the change to ATA/66, wewill do our best to make it clear.