While I was working on our Western Digital 1200JB Review, I was very much upset about the absence of the WD1200BB models among the pieces we had at our disposal. It was really hard to draw logical conclusions from the comparison of HDDs with 40GB platters and a WD monster with a 8MB cache-buffer, as I couldn't be sure what influenced the performance of WD1200JB more: a four times larger cache-buffer or higher data density per platter. Now that we've finally got our hands on WD1200BB we can answer this question with hundred percent certainty.
As we have already compared the performance of a good many hard drives with 40GB platters (except for Samsung ones) in our Western Digital WD1200JB HDD Review, I made up my mind to have a look at only three products this time: WD1200BB, WD1200JB and IBM Deskstar 120GXP (IC35L120AVVA07). All of them boast 120GB storage capacity and 7200rpm spindle rotation speed.
Well, it looks as if only two companies dared produce hard disk drives with 40GB platters: Western Digital and IBM. On the one hand, it indicates their technological leadership over the competitors, on the other, it means that these two companies are not planning to update their product range too soon. At least, Western Digital gave us to understand that we will see no new HDDs before SerialATA comes into being.
|WD1200BB||WD1200JB||IBM DeskStar 120GXP|
|Spindle Rotation Speed||7,200rpm||7,200rpm||7,200rpm|
|Average Seek Time (Read)||8.9ms||8.9ms||8.5ms|
|Average Track-to-Track |
Seek Time (Read)
|Average Full Stroke |
Seek Time (Read)
As we see, many positions in the specifications of WD1200BB and WD1200JB are similar. The only exception is the cache-buffer size, of course. IBM HDD boasts a bit faster mechanics, but I wouldn't call it a victory grant so early.
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
- Intel IOMeter 1999.10.20
Average Access Time
As usual, we measured the average access time on addressing the random sector in two tests: Winbench99 and HDTach:
Sustained Linear Read Speed
When we tested WD1200JB the linear read graph was of such unusual shape that the DTR: Beginning and DTR: End results didn't reflect the actual per platter data density of the drive. Now that we have WD1200BB at our disposal, we can figure out the "real" linear read speed from a WD's 40GB platter.
It is hard to say why WD1200JB produced such a weird liner read graph. Maybe it was because of some "experimental" software reflashed into the sample we've got :)
In HDTach benchmark we considered the performance of our testing participants in three cases. The first test we carried out dealt with checking the maximum data read speed from the cache-buffer of the drive:
As we see, here IBM 120GXP wins the first prize. It managed to outperform both WD solutions in all three protocols.
Another interesting example is the average data read speed from the HDD platter. WinBench test shows us only the minimum and maximum read speed from the platter, while HDTach provides a much more illustrative average value:
Here both WD hard drives prove faster than IBM 120GXP.
And the average write speed is better by IBM solution, as the diagram shows.
WinBench99 1.2 for Windows98
So, now let's see what our giants are worth in Windows98 (the drives were formatted by Paragon Partition Wizard).
As we have expected, WD1200JB appeared ahead of all. The second position got occupied by IBM 120GXP, which managed to elbow its way between two WD fellows. Moreover, its performance is closer to that of WD1200JB rather than to that of WD1200BB drive.
But, just as the previous time, IBM 120GXP doesn't feel very well in UDMA33 protocol.
Adaptec Threadmark 2.0
This benchmark measures the HDD performance in multi-thread mode. As we see, all three hard disk drives cope with this benchmark pretty successfully.
WinBench99 1.2 for Win2000 (FAT32)
As we shifted to another OS, IBM 120GXP got completely cured from its dislike towards UDMA33 protocol in a wink of an eye. Was it the work with 64KB data blocks that the drive didn't like most then?
The performance of IBM 120GXP in UDM33 grew up, but just look how greatly WD1200JB improved in UDMA66/100:
Wow, there is a tremendous gap between WD1200JB and both hard drives with 2MB cache buffer. And among the ordinary drives, IBM baby appears an indisputable leader, including UDMA33 protocol.
In High-End tests IBM 120GXP managed to get strikingly close to WD1200JB, though it didn't prove powerful enough to beat the monster with 8MB cache-buffer. However, look how greatly it defeated WD1200BB: by the good 13% (in UDMA100).
WinBench99 1.2 for Win2000 (NTFS)
In NTFS file system both WD drives feel absolutely at home. As well as IBM HDD, which was also designed in NTFS epoch :)
That is why I was not surprised to find out that the situation had changed. WD1200JB is the No.1, then comes IBM 120GXP and the last one is WD1200BB.
In High-End benchmarks IBM 120GXP HDD worked wonders. In UDMA100 mode it caught up with WD1200JB having left WD1200BB far behind. However, in all other UDMA protocols IBM solution appeared a bit slower than WD1200JB though still faster than WD1200BB.
So, summing up the results obtained in all WinBench99 tests we can conclude that Wd1200BB is definitely slower than IBM 120GXP (with the only exception of UDMA33 in Windows98). It means that we have every reason to state that larger cache-buffer of WD1200JB determined its indisputable victory over the competitors.
In conclusion, we would like to offer you the results of Intel IOMeter tests. At first come three standard patterns:
Only in case of queue depth equal to 1 WD1200JB managed to outperform IBM 120GXP. However, as the workload kept growing, IBM 120GXP easily defeated both WD fellows. Note that Wd1200JB is much faster than its 1200BB counterpart.
And in this pattern WD1200BB didn't fall too far behind WD1200JB solution. IBM 120GXP performed nearly equally with WD1200JB in case of queue=1, and outpaced it in all other cases.
And in this pattern Western Digital HDDs look far better, especially WD1200JB. Of course, the drive with larger 8MB cache-buffer appeared faster than IBM 120GXP in all modes except queue=256. But even WD1200BB turned out fast enough to beat IBM 120GXP in case of queue below 4 requests! DataBase pattern differs from all the previously considered ones by larger number of Writes (33% from the total number of operations). That is why pretty good performance of WD HDDs might be connected with better implementation of lazy write algorithms (This algorithm serves to update data on the hard drive with maximum efficiency. It implies that file systems supporting lazy write place the changes of the stored data into the cache buffer and then move it to the platters, very often in background mode).
Let's find out if this supposition is true.
Just like in our WD1200JB HDD Review we created a pattern with random 8KB data blocks. The reads-to-writes ratio in this pattern changed from 100/0 to 0/100 with the step equal to 10.
Here are the results obtained:
Now I suggest taking a look at the diagrams with queue equal to 1, 16 and 256 requests.
You can clearly see that both Western Digital HDDs process the requests faster as the number of writes grows. You can notice the performance increase by IBM 120GXP only when the number of writes reaches 60%. Besides, you can also clearly see that WD1200JB with a far larger cache-buffer than that of WD1200BB is more successful with lazy write algorithm.
By the way, WD1200BB and IBM 120GXP run almost neck and neck in RandomWrite pattern (in case of 100% writes). It is also worth pointing out that IBM 120GXP outperforms WD HDDs once the writes get below 20%.
As the queue depth reaches 16 requests, IBM drive speeds up a little bit, though it is still unable to compete with WD1200JB in case of over 20% writes. As for WD1200BB, IBM drive competes with it pretty successfully up to 45% writes, but after that WD hard disk drive defeats it. However, in RandomWrite mode WD1200BB and IBM 120GXP drives again show similar results.
The apotheosis of our today's race is the queue=256. In this case IBM 120GXP again gets a bit faster and moves the cross-points with the graphs for WD1200JB and WD1200BB farther to the right along the X-axis.
Here I would also like to draw your attention to the dog-leg on the WD1200BB graph in 70% writs: very interesting :)
Well, WD HDDs turned out to feature very well implemented lazy write algorithm (this also explains good results obtained by these drives in benchmarks with HDD de-fragmentation). However, IBM 120GXP boasts better queue sorting (changing the requests processing order) intended to create the most optimal route for the heads moving along the platters.
We measured the temperature with the help of DTemp utility (for IBM HDD) and infra-red thermometer (for WD HDDs) after 2.5 hours of Intel IOMeter tests running. In this test all hard disk drives work very hard as actuators, so that we can say that they are running under the heaviest workload from the heating point of view.
The results obtained during this test session let us draw the following conclusions:
- WD1200BB HDD is slower than IBM 120GXP in all tests except UDMA33 in Windows98.
- WD1200JB owes its huge advantage over the competitors to larger cache-buffer and of course, to the firmware using it.
- WD HDDs can boast very successful implementation of lazy write algorithms.
WD1200BB proved fast enough to be called one of the today's fastest IDE HDD. Its 40GB platter boasts outstandingly high per track data density, which has never been the case for WD Caviar 7200 HDDs (as we remember, WDxxxBB drives used to have pretty low data read speed from the platter). Unfortunately, Western Digital doesn't share any plans about moving slower WD Caviar 7200 models to 40GB platters and providing them with large cache-buffers yet…
In this case, it would be fair to state that an indisputable leader among the most popular HDDs is IBM 120GXP.