"Can't Get Out of My Head"

Just like a well-known Australian pop-singer, whose song has become a title to the review introduction, I also couldn't get rid of one fixed idea. As you may have already guessed, my most inmost wish was to try an IBM hard disk drive from the Deskstar family: 120GXP. But, just like last year IBM sustained a pause like a gifted student of the great Stanislavsky… Really why should IBM speed up time? The hard disk drives from 60GXP family have been showing their excellent performance almost everywhere leaving behind numerous competitors. They were in firm demand in the OEM and Retail markets and very rarely required warranty repair. A real idyll! However, in our busy times, it is really hard to sell HDDs bringing in low revenues, even if they are fast and reliable. It's true that the competitors' HDD solutions of the same storage capacity were built with fewer components (platters and heads) and as a result cost less.

I hope no one will argue about the fact that IBM is currently the most high-tech company in the world and the only reason why its solutions are not years ahead of the competitors, is because the market is not ready for them. But in our case the market is already well-prepared :)

Probably, once IBM realized it, they launched their new HDD model: Deskstar 120GXP. As usual, the number in the model name stands for the maximum storage capacity in the family. The HDD features 120GB storage capacity, which is achieved due to the use of 3 40GB platters, i.e. IBM uses the today's largest platters in its products, just like its rivals do. However, we all do remember pretty well about the 5-platter HDDs from IBM: DTLA307075 from the 75GXP family. If we try to imagine a HDD made of 5 platters 40GB each, then the overall storage capacity of this solution will make 200GB! Do we need a HDD like that nowadays? I don't think so. Besides, such a great number of platters used will produce too much noise during operation and will result into inevitable overheating and painful death…

So, 120GXP HDD was announced. However, it meant nothing but the beginning of another waiting period: waiting for the drive to start selling (especially for us, as IBM doesn't send any samples to Russia, unlike Seagate or WD).

Luckily, no pause is eternal. And one day some very enterprising guys brought several IBM 120GXP drives to Moscow. I was lucky to get one of those. So, here you are:

  

As you can see, this HDD is hardly any different from the 60GXP model.

Testing Participants

First of all I would like to say a few words about our today's hero: IBM Deskstar 120GXP HDD.

So, glass platters, which surface can be made extremely smooth, plus the notorious "magic pollen", i.e. the thinnest 3-atom layer of Ruthenium metal (Ru belongs to the platinum group, was discovered by Karl Klaus in 1844 and named Ruthenium for Russia, as the ancient name for Russia is Ruthenia) - this is the entire secret of IBM engineers. This is how they managed to achieve such impressively high data density per platter (29.7Gbit per square inch).

At the same time, the heads positioning speed didn't change! The claimed average seek time equals to the traditional for IBM HDDs 8.5msec.

The HDD supports ATA/100 interface. This is not the newest feature of the storage market. However, as we have already found out, the new ATA/133 interface hardly tells on the performance of the today's HDDs. Since the 120GXP family doesn't have any solutions with the storage capacity exceeding 137GB, they don't need the Big Drive specification.

For my testing session I decided to take all the best HDDs of the today's storage market: Seagate Barracuda ATA IV (ST3008021A), WD800BB, WD1000BB and Maxtor D740X-6L (6L080J4). Since we got hold of the Deskstar 120GXP with the 80GB storage capacity (its full model name sounds like IC35L080AVVA07-0), then all the other testing participants, except WD1000BB also features the same 80GB storage capacity.

Here is a comparative table with the HDD features for your reference:

Note that Full stroke of IBM 120GXP is only 15msec!

Testbed and Methods

We tested the newcomer in the following test system:

• Intel Pentium III (Coppermine) 600MHz CPU;
• ASUS CUBX-E mainboard, bios 1007A;
• 2 x 128MB PC133 SDRAM by Hyundai;
• Matrox Millennium 4MB graphics card;
• Windows 98/Windows 2000 Pro.

In order to check the HDDs performance in different UDMA modes, we had to use the following controllers:

• UDMA33: the controller integrated into i440BX chipset;
• UDMA66: Promise Ultra66 controller;
• UDMA100: integrated Promise Ultra100 controller on ASUS CUBX-E mainboard.

For Promise controllers we used the drivers ver. 1.60 (build 33).

The disk drives were connected as Master-units to a separate IDE-channel. DMA support in Windows was enabled. 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

Performance

Average Access Time

Let's find out how well IBM engineers managed to "reconcile" the data density with the seek time:

Well, the average access time we measured appeared even lower than the results we calculated: 8.5 + 4.17 = 12.67ms. However, Maxtor HDD turned out a bit faster.

Sustained Linear Read Speed

And now let's see what the IBM 40GB platter is worth:

Amazing! The previous winner in the read speed (beginning) was WD1000BB, but not any more: now it is completely defeated! 47MB/sec for a HDD with 7,200rpm spindle rotation speed is an impressive result. of course, it became possible only due to IBM's "noID sector format" technology.

By the way, IBM 120GXP differs from the other HDDs by twice as many zones with similar number of sectors per cylinder. Our hero has 31 zones like that! If you are curious to double check this number, please, see the linear read graphs, which will come later in this review (in the WinBench section). You will be able to see these zones yourselves.

HDTach 2.61

HDTach test will help us to find out the average read and write speeds and the maximum read speed from the drive buffer.

At first you may get the impression that in read and write tests only that hard disk drive wins, which has more data written onto the outer tracks. However, things are far not that simple. The speed is measured on the entire disk surface (to tell the truth it is not quite correct: the speed is measured with 64MB step along the HDD address array). As a result, the entire drive, and not just some selected sectors, contributes to the end-result, which is obtained as an average of all the intermediate results. This averaging gives us an approximate drive speed at a very short period of time. This is also very important, because read/write operations performed on the entire HDD with its huger storage capacity can take a really long time.

Let's compare the read burst speed of our testing participants:

As we see, the HDDs show almost equal results here. Now we would like to cast a glance at the linear read speed:

This is another proof of the just discussed theory. WD1000BB, which read speed in the beginning of the drive was much slower than that of IBM 120GXP, has caught up with the rival in the average read speed due to faster reading in the end of the drive.

As for the average write speed, IBM baby again won the laurels, but the second prize now belongs to Maxtor solution, and not to WD1000BB.

Well, IBM appeared an indisputable leader in the synthetic HDTach benchmark. Let's see what it is worth in WinBench.

WinBench99 1.2 for Win98

As usual, at first I would like to offer you the following table:

Now here come the diagrams:

Well, IBM HDD is leading the race, however, it hasn't got too far ahead of Maxtor.

At the same time, I suggest you pay attention to the extremely low results shown by IBM 120GXP in UDMA33. Very strange, especially since we have never seen IBM HDDs being unfriendly to this protocol.

However, in High-End test IBM HDD performs just brilliantly! It is 12% faster than the nearest rival, Maxtor HDD. And Seagate Barracuda ATA IV, which won the third prize, is 17% behind the leader.

Threadmark 2.0

This is one more benchmark, which may seem absolutely non-systematic. I suppose that you can still find some system in the results obtained. For instance, WD hard drives have been winners in this benchmarks lately. As you remember, WD HDDs also win in those benchmarks, which focus on defragmentation and copy operations.

Yes, I nearly forgot to remind you that Adaptec Threadmark emulates the work of several applications creating "data streams".

Well, no more WD harmony in these benchmarks!

But what do we see? Hey, the results of IBM 120GXP in UDMA33 mode are again very poor. Is it a tendency?

WinBench99 1.2 for Win2000 (FAT32)

Now we will run our benchmarks under Win2000. Due to more advanced core, this operation system loads the disk subsystem much more efficiently, so that Win2000 tests show considerably higher performance. However, the users also state that it is much nicer to work under Win2000. It must be Gates' hypnosis :)

DTR: Beginning

Graph

Graph

Graph

Hm, IBM 120GXP seems to be too fast here…

Look at the linear read graphs for this HDD, especially at the one for IBM 120GXP in UDMA33.

Unlike the tests for Win98, IBM HDD in Business WinMark performs quite well. It is almost 6% ahead of Maxtor HDD, which won the second prize this time. And the curious thing is that in UDMA33 mode IBM is not bad at all. We would even say that it managed to outperform its competitors in UDMA33 much more than in UDMA66 or UDMA100. Does it mean that IBM 120GXP doesn't have any idiosyncrasy towards UDMA33?

[Censored!] My god! It looks as if the 7,200rpm HDDs felt really hurt by their SCSI fellows with higher spindle rotation speeds! What made me speak of SCSI right now? Because only SCSI HDDs, and only the fastest of them, could boast performance like that.

Well, IBM once again showed the world what one can squeeze out of a 7,200rpm HDD with 2MB cache. Note that in UDMA33 IBM solution nearly beat WD800BB in UDMA100!

Winbench99 1.2 for Win2000 (NTFS)

We wonder if the same gap between IBM HDD and the others will remain in NTFS.

Hm… IBM is again ahead of all, however, the performance difference between the leader and the rest of the racers is not that great, like in Win98. Just in case I have to add that this time IBM faced no problems in UDMA33.

No matter what, but IBM 120GXP managed to leave its competitors quite far behind (8% ahead of the closest rival, WD1000BB).

Well, I have to admit that IBM tore its competitors to tatters in Winbench99 tests in all operation systems. The only upsetting thing is the behavior of the winner in UDMA33 in Win98. Though some optimists may say that Win98 is a pure anachronism now :)

Intel IOMeter

And in conclusion our favourite benchmark. If you are looking for more details about the methodology and main idea of this test, please see our earlier articles. We will pass over to the results:

Well, everything is more than clear: IBM 120GXP is always a leader when the number of requests exceeds 4. It is most successful in WorkStation pattern. Anyway, let's have a look at the graphs:

You can clearly see that when the workload is low (1 and 4 requests), IBM HDD performed slower than Maxtor D740X-6L, however, the higher grew the workload, the faster it performed and left the Maxtor counterpart behind.

Here IBM 120GXP won the leader's laurels in the very beginning and retained the leading position till the very end of the test. To be fair I should say Wd hard drives fell behind it a bit less than in the previous pattern.

In this pattern the situation got very tense in case of low workload. However, at higher workload, over 16 requests, IBM 120GXP hardly had any more competitors left.

Temperature

To measure the temperature of the working HDDs we resorted to DTemp utility. The room temperature stayed around 22^oC, and the measurements were made after more than 2.5 hours of IOMeter tests running. In this benchmark the HDD actuator works very intensively so that we could state that the drive works at the utmost of its power.

This way, the temperature of the working IBM 120GXP HDD we measured made 48^oC. This is quite a lot for a dual-platter HDD that is why I would recommend those of you who decide on this solution, to think about some extra cooling for your IBM 120GXP.

Conclusion

Well, having run all the tests I have to admit that this hard disk drive was worth waiting for. Its unbelievable performance in WinBench tests showed that all the modern HDDs have room for performance increase, and needn't speed up their spindles to 10,000rpm to get faster.

In IOMeter tests we could also see the advantage of IBM 120GXP over the competitors in case the workload exceeded 4 requests.

Two major lows of this solution (I have to stress that these lows may exist only for the piece I had at my disposal) are high working temperature and some unpleasant squeaking noise when the heads were moving from one position to another. Though you can here these noises only if you listen attentively.