by Nikita Nikolaichev
05/23/2003 | 11:14 AM
It seems like just a little while ago we reviewed a 200GB hard disk drive from Western Digital called Drivezilla (see our Roundup: IDE Hard Disk Drives with 8MB Cache-Buffer). Now thanks to Western Digital Company we got an opportunity to test an even larger hard disk drive. The capacity record for 7,200rpm HDDs is 250GB now! Interestingly, the newcomer didn’t boast a unique proper name. Well, it looks as if WD marketing people couldn’t think of anything “cooler” than Drivezilla :).
But WD engineers could do something really cool! The new disk, called “Special Edition 250GB”, boasts the highest per platter density in the industry: 83GB. However, this untypical density was chosen exactly for pure marketing reasons: three platters like that make up the magic number: 250GB!
The fresh piece of candy from WD has a new wrapping as well as a new filling. The package is now made of plastic, not paper as it used to be:
It’s nice that they keep the practice of complementing retail HDDs with a “cost-free” ATA controller. Older mainboards may not recognize such a big hard disk drive, so Western Digital takes it easy: they just present you with a UDMA controller from Promise: Ultra100 TX2. Besides the controller the package includes:
So, if you buy a boxed HDD, you automatically get everything necessary to install it in the computer and make it work.
The exterior of the drive is quite familiar:
Only the shape of the cover is slightly different (you can compare it with WD2000JB in our Roundup: IDE Hard Disk Drives with 8MB Cache-Buffer, for example).
Similar shapes don’t always hide similar content. :)
“Officially”, WD2500JB differs from WD2000JB mostly by larger platter (83GB against 66GB), but we will see that it’s not the only point of difference between them.
You may remember that the last time we compared 8MB-buffer hard disk drives, they were IBM Deskstar 180GXP (IC35L180AVV207-1), Maxtor DiamondMax Plus 9 (6Y160P0), WD1800JB and WD2000JB. Today we will also take four drives, but the cast is quite different.
Firstly, Maxtor DiamondMax Plus 9 6Y200P0 turned up right in time for this review. It is the current leader of the DiamondMax Plus 9 family. By the way, this model is going to be the last in this family and the next capacity bar will be cleared by a drive from MaxLine Plus II family (7Y250P0/7Y250M0).
Secondly, we tested Maxtor DiamondMax Plus 9 6Y200P0 after it underwent some “special” preparations. You may have noticed that the test results we got for last generation Maxtor drives were full of “miracles” and other “peculiarities”.
Thus, we often pointed to a considerable difference in the performance of the drive on different UDMA controllers. A drive was faster with UDMA100 than with UDMA133 in one test and slower in another. For a few months running I have been trying to get repeated results, and, as usual, science defeated common sense in the long rum. In one of our next articles we will share with you the results of our findings. For now, I offer (hopefully) “correct” results for Maxtor DiamondMax Plus 9 6Y200P0.
Thirdly, WD1800JB is not included into this test session (to make it more straightforward).
Overall, we have top-end HDD models from three manufacturers:
The testbed configuration looks as follows:
ATA100 HDDs were tested with Promise Ultra100 TX2 controller (BIOS: 2.20.0.14 Drivers: 2.00.29).
The following software was used:
Before the tests the AAM register of all HDDs was set to OFF position (FAST mode) with the help of Hitachi Feature Tool Utility. For WinBench tests all the drives were formatted in FAT32 and NTFS as one logical drive with the default cluster (to format the drives in FAT32 we used Paragon Partition Manager utility). The tests were run four times each, the maximum result was taken for the diagrams. The drives didn't cool down between the tests. The tests in Intel IOMeter were run in SequentialRead, SequentialWrite, DataBase, WorkStation, FileServer and WebServer patterns. If you are looking for the detailed description of these patterns, please, see our previous articles.
The participating HDDs had the following firmware versions:
The different firmware versions of WD2500JB and WD2000JB already imply significant differences in their performance.
HDTach is a questionable test, in my opinion. So, we don’t dwell upon its results too much and only compare the drives by their average access time and burst rate parameters.
Wow! WD2500JB with its higher-density platters has better access time than WD2000JB! Usually, this happens when the capacity of the platters is not used to the full extent. Once, I got my fingers burnt on Maxtor drives, and now I suspect everyone and everything. :)
So, maybe the design of WD2500JB is better described by the 63x4 formula, rather than 83x3? Read speed values also indicate this…
I abhorred the idea to take such a pricey thing to pieces, but was eager to know the truth. That’s when a new “bloodless” method of checking the number of platters in a drive dawned upon me: weighing! Drives belonging to one family are usually manufactured using the same case. So, the different weight of two models with different storage capacities would indicate a different number of platters and heads used to build it. Having collected weight statistics for various models, you can guess the number of platters in a new drive with a high degree of accuracy.
In our case, we have two WD drives that both have three platters and six heads (at least, theoretically). So, let’s put them onto the SCOUT II scales (SCA210 model) from Ohaus.
As you see, both drives are equally heavy. Since WD2000JB doesn’t have four platter-design for sure, WD2500JB is no longer suspected! Just like in a detective story a-la Erle Stanley Gardner…:)
Maybe the lower access time of WD2500JB can be explained by its higher spindle rotation speed? Western Digital has some experience in “overclocking” WDxxxxAB hard disks to 6,000rpm, so we had to check out this supposition, too. Utilities that measure the spindle rotation speed said that WD2500JB’s spindle rotated 7,200 times per minute. So, there is definitely no spindle “overclocking” here.
To find the answer, let’s take a look at the platter structure of our drives (to our regret, we don’t have any data for the model from Maxtor, but it’s only a question of time…):
As you see, WD2500JB has much more servomarks per track (over 20%!) than WD2000JB.
Higher data density (the data density per track, to be more exact) gave the opportunity to put more servomarks, keeping the same effective track density. Thus, the linear speed didn’t drop down, but even grew a little. And the more servomarks per track there are, the faster the HDD will find a necessary track.
By the way, our study of the platter structures produces an interesting fact: the zero working cylinder of IBM Deskstar 180GXP IC35L180AVV207-1 is actually the 28000th one counting from the platter edge! Yeah, every hard disk has the so-called zero track at a distance from the edge, but this distance is usually 700-800 cylinders…
It means that the effective capacity of the platter by IBM is much smaller than the “theoretically possible”! Let’s try to estimate the capacity of that theoretical platter. The cut-off block of 28 thousand cylinders is about one third of the 70000 cylinders (note also that the highest-density part of the platter is cut off). Thus, the hypothetical capacity of platters in IBM 180GXP equals about 83-84GB. What do you think, what platters Hitachi 250GXP will be based on? :)
Burst Rate results brought no surprises: WD2500JB was a little faster than WD2000JB, but far behind the leaders.
As usual, we are going to test the drives’ ability to process streaming read and write requests. In SequentialRead and SequentialWrite patterns, IOMeter sends read and write requests (request queue depth = 4) and every half a minute changes the data block size. Thus, we get the dependence of read/write speed on the size of the requested block.
Well, unfortunately, we have to admit that WD2500JB is as slow as WD2000JB at processing 2-16KB blocks. The new WD drive also achieves its “nominal” linear read speed only on 32KB blocks.
The new WD is much better at writing. Although WD2500JB didn’t get rid of the family dislike of 1KB blocks, it is clearly better than others at processing data blocks larger than 4KB.
Overall, we haven’t seen anything extraordinary in sequential read and write tests. The higher data density of WD2500JB accounts for its higher speed when working with large data blocks. We also noticed that the new WD is better at writing than at reading.
Database pattern is the next test for WD2500JB.
Unlike WD2000JB, WD2500JB doesn’t flunk the RandomRead mode! This suggests that WD2500JB owes its improved average access time not only to more servomarks per track, but also to firmware enhancements.
Good performance of WD2500JB in the RandomRead mode may result in high performance in the WebServer pattern as it also has no write requests.
Note also the results demonstrated by Maxtor HDD. Our “special” preliminary preparations led to the following picture: the drive shows the same performance on different UDMA controllers in those tests which do not care about the amount of data sent through the interface.
Under higher workload both WD drives behave practically identically, although WD2500JB leaves its counterpart behind when the write requests share is high.
It is very interesting that the drive from Maxtor produces slightly different results on different UDMA controllers.
The workload gets a little higher again, and we see the same effect as under linear workload: WD2000JB doesn’t like the RandomRead mode, while WD2500JB copes with it all right.
Server patterns come next.
The picture tells more than words can say. WD2500JB is a bit faster than WD2000JB, while the IBM drive is much faster than both WD solutions under medium and high workloads. Maxtor lags behind the rest under any workload, so we can’t say anything good about its “server” optimization…
It’s curious, but again we see Maxtor drive performing differently with different UDMA controllers when processing 16-requests.
Let’s calculate the performance rating of the drives by averaging Total I/O values under all workloads:
We get just what we have expected to see: WD2500JB has a small advantage over WD2000JB and they both fall behind the IBM drive.
As the WebServer pattern has write requests, WD drives are deprived of their trump, the excellent lazy write implementation. So, only the most optimistic people, like myself, can expect them to notch better results. :)
I took a look at this graph and all my optimism disappeared instantly. Nevertheless, I should point out that WD2500JB is notably faster than its fellow drive in this pattern, so there is some progress to be mentioned. You can also see it in the ratings diagram:
WD2500JB is targeted at workstations, so here it must do its utmost. The features of the Workstation pattern – a bigger share of write requests and “smaller” average request size – allow WD drives to show their best.
And they did it! The drives from Western Digital are far ahead of the competitors under smaller workloads! Only IBM drive can compete with them, but the strong point of this product (Tagged Command Queuing) works only under high workloads, which don’t often occur during regular work of a desktop PC.
As for Maxtor, it once again showed different performance with different controllers. It’s also characteristic that the biggest gap between the two controllers takes place when the queue depth reaches 16 requests.
Still, we are now interested in the HDDs rating, rather than in the dependence of Maxtor cache-buffer segmentation on the UDMA mode… We will calculate the performance rating in the WorkStation pattern with the following formula:
Performance = 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.
Well, the WD drives pass the WorkStation pattern with flying colors: they won the two top places. And if there were WD1800JB…
After the success in Workstation pattern, let’s see what the WD drives are going to show us in WinBench99 tests.
Well, I have hoped for something more than that. Although WD2500JB outran WD2000JB in the High-End Disk WinMark test, they both have a long way to go before they reach IBM and Maxtor. In the Business Disk WinMark all the drives show similar performance, so there is no definite leader.
To be objective, we should note that very large drives can’t show their true potential here, because we format them for WinBench tests as one logical partition of the full drive’s capacity. So, such gigantic drives have very large FAT (in FAT32) or MTF-files (in NTFS), which negatively tells on their performance…
On the other hand, the competitor drives with similar storage capacities performed much better in WinBench99.
Let’s now turn to NTFS (I hope you remember that this file system “favors” WD drives).
Yeah, WD2500JB is better in NTFS and it nearly caught up with the IBM drive in the High-End Disk WinMark.
In conclusion here is the linear read graph for WD2500JB.
Our file-copy test is the last in the show. I hope you are already familiar with it. If not, please see this article for more details about this benchmark). The HDDs performance was measured with 5 preliminarily prepared file sets. The test names let you guess very easily which folder served as a prototype: Install, ISO, MP3, Programs and Windows.
Why do we use so many different files here? You may think it could be more than enough to take one big file or a few hundreds of small files to measure all the speeds. The matter is that we wanted to take the “real” HDD performance, that is how fast it works with the real and not synthetic files. Since hard disk drives from different manufacturers cope with files of different sizes not equally fast, we need quite a lot of files of the same size to be able to detect the HDD’s native inclination.
The new WD drive demonstrated a significant write speed growth compared with the WD2000JB, especially when working with large files. But still it couldn’t outpace the drive from Maxtor…
WD2500JB is also notably faster than WD2000JB at reading file sets. It even stepped on top of the podium in the ISO pattern.
The Maxtor drive is hard to compete with at copying files. WD2500JB can be considered a worthy rival to Maxtor and IBM only in case of large files.
At far-copying, WD2500JB suffers a greater performance drop than Maxtor. So, there is no competition even in the ISO pattern.
WD2500JB shows somewhat better results in NTFS, but even here it comes on top only at reading large files.
So, the copy tests say that WD2500JB is always faster than WD2000JB, but, regrettably, nearly always slower than the drives from Maxtor or IBM.
We’ve got a surprise for you here! Being an incorrigible optimist, I always tried to check the temperature of WD drives with the help of SMART-reading utilities. And now I was rewarded for the persistence: WD2500JB supports temperature monitoring in SMART!
We measured the temperature following our standard method (one hour and a half of IOMeter tests running). It was 46oC. This is quite a high temperature, so I would advise you to use WD2500JB with additional cooling.
The temperature measurements taken with an infrared thermometer showed the same results as for WD1800JB and WD2000JB from the previous review. And really, why should they be different? The construction of the drive hasn’t changed…
The noise level also remained the same as that produced by WD2000JB.
So, it is time to dot the i’s and cross the t’s.
It’s a meaningful fact that Western Digital marked the transition to 80GB platters with the launch of the world’s highest-capacity drive with 7,200rpm rotation speed. While the competitors are using 80GB platters in low-priced products (to reduce their production cost), WD shows its muscles in the upper price range.
Well, if we try to extrapolate the speed characteristics of WD2500JB to drives of smaller capacities… They are going to very competitive products! :)
On the whole, WD2500JB is quite a successful product. It combines high capacity and high speed. We are also glad to see a few drawbacks noticed in WD1800JB and WD2000JB to be finally eliminated. We are talking about too high access time, low speed in the RandomRead mode, and absence of thermal diodes.