I don't know about you, but by me every mention of FireWire associates with the fired fuse… However, the more correct should be an association with Apple Company, which has been the only one for a relatively long time who was pushing forward this interface in the market (IEEE 1394 specification was approved as an industry standard by the Institute of Electrical and Electronics Engineers only in 1995). The efforts of Apple Company definitely paid back: take for instance the Emmy award for the contribution to the television industry.
For the PC market FireWire interface was something exotic for a long time. However, as the number of digital cameras supporting this interface in the market grew bigger, the PC platforms also required broader FireWire support. We all saw very well the PCI cards supporting IEEE 1394, which then got replaced with the more common integrated IEEE 1394 controller chips (you can see them not only on mainboards, but even on the sound cards now!).
And if the things stand like that, then why not sell external HDDs with IEEE 1394? You may be wondering: what for, as there is USB 2.0! The matter is that IEEE 1394 interface was designed to transfer streaming video, and USB interface was initially designed for low-speed peripheral devices. In version 2.0 the bandwidth grew up by 40 times (to 480Mbit/sec), so that USB 2.0 appeared faster than IEEE1394 with its 400Mbit/sec. However, the ability of the interface to transfer big data packs is not always determined by its maximum bandwidth (it is actually the size of the transferred data pack that tells more on the performance).
Well, as usual the box is the first thing we get acquainted with when we get a new product:
The box of our WD FireWire External Hard Drive is pretty classical in design, which we are little by little getting used to. :)
Inside this big beautiful box we found a very elegant silver-transparent case with a FireWire logo on top:
At the front panel of the case we can see a WD logo and two indicators: power and HDD addressing.
At the rear panel of the case we can find two connectors for FireWire cables and a power supply connector:
At the bottom of the case there are a few thin slots letting the airflow for HDD cooling circulate inside and outside the case:
Of course, I couldn't resist the temptation of opening this case very carefully:
And what do you think was inside? I found a hard disk drive WD 1200BB, as we have actually expected and told you in our news some time ago.
Note that in the back part of the case (the left part of the photo) all electronic parts are covered with a foil. I didn't dare open it :)
The package coming with the drive includes the following items, besides the drive itself, of course:
Testbed and Methods
We ran all the benchmarks in the following test system:
- Intel Pentium III (Coppermine) 600MHz CPU;
- ASUS P3B-F mainboard;
- 2 x 128MB PC100 ECC SDRAM by Hyundai;
- IBM DPTA 372050 HDD;
- Matrox Millennium 4MB graphics card;
- Windows 2000 Pro SP2.
To test WD External Hard Drive we used Adaptec AUA-3121 controller, which is a complete solution for powerful graphics stations allowing the use of external hard disk drives due to a combination of the USB 2.0 controller from NEC and IEEE1394 controller from Texas Instruments on a single PCB.
WD1200BB hard drive was tested with promise Ultra100 TX2 controller.
For WinBench tests we used FAT32 and NTFS file systems to format each of the hard disk drives 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:
- HDTach 2.61;
- WinBench99 1.2;
- IOMeter 1999.10.20;
- FC-Test v.0.3.
As you may have noticed, now we've got a new benchmark in our testing set: FileCopy Test (FC-Test). You can find its detailed description alongside with the working principles and testing methodology in this article.
First we would like to offer you the results obtained in HDTach:
As for the maximum read speed from the cache (read burst), our hero, WD FW (hereinafter I will call WD External Hard Drive WD FW) appeared nearly twice as slow as the regular WD 1200BB. However, the average read and write speeds of WD FW were not much slower. But since we know that there is the same WD 1200BB inside WD FW, then the performance difference can be explained only by some limitations imposed by FireWire interface. Which we will beautifully show in Intel IOMeter tests in SequentialRead and SequentialWrite patterns. :)
Since the external drive remains the same hard disk drive despite its "externalness", we can store some utilities and work files on it as well. And not only store, but also work with them!
|Linear Read Graphs||Graph||Graph|
Judging by the results obtained in WinBench99, WD FW drive yields to its ATA/100 counterpart about 10% of the performance in Business Disk WinMark and around 16% in High-End Disk WinMark, which corresponds to the speed of fast 5,400rpm drive.
The linear read graphs for WD FW and WD 1200BB show that the linear speed of WD FW on fast tracks is limited by the bandwidth of FireWire interface.
Intel IOMeter: WorkStation Pattern
We are using this pattern for our tests for a not very long time that is why I decided I'd better remind you where it actually comes from. The respected site about HDDs and other storage devices, StorageReview.com investigated the workload upon the disk subsystem when the user works with the regular Windows applications we are all using every day (under NTFS5!). The results of this investigation were summed up into a pattern for IOMeter test. Since we focused on the workload created by office applications, the pattern was called WorkStation:
WD FW showed pretty curious results in this pattern, I should say. Note that in case of low queue depth, which sets the workload, the hard drive with slow FireWire interface outpaced its ATA/100 counterpart. However, as soon as the queue rises up to 8, ATA/100 HDD rushes ahead.
Intel IOMeter: StorageReview Patterns
In patterns emulating server workload the situation is completely different:
ATA/100 hard disk drive appeared faster than WD FW from the very beginning, i.e. when the workload was very small. And its advantage keeps growing (approximately up to 10%) as the workload increases. However, then as the workload reaches 64 and 256 requests, the performance gap between the testing participants starts reducing. Of course, these workloads are very unlikely to take place for a single hard disk drive, so let's simply take note of these results, and use the performance rate shown under lower workloads to estimate the HDDs' performance in these patterns.
Intel IOMeter: DataBase Pattern
The old DataBase pattern has been recently replaced with this monster. Testing with the new pattern takes over 10 ours of my life. However, the results are worth it, really, as they allow us to judge about the HDD performance in a good ten of key knots.
To compare the hard drives we use the results for three different queue depths (1, 16 and 256 requests):
As we see, the HDD speeds do not differ that much from one another, but ATA/100 HDD is nevertheless faster than WD FW.
Here the advantages of ATA/100 HDD are much more noticeable and almost do not depend on the read-to-write ratio.
Intel IOMeter: SequentialRead Pattern
Well, we have actually expected something like that to happen. The wider is the road, the more cars can go along it… :)
Note that the graphs showing the dependence of the data transfer rate on the size of the data pack feature three parts. On the first part, where the data pack is still very small, the performance is influenced by the data bus width, because the amount of auxiliary info the HDD should exchange with the controller is comparable with the size of the data pack transferred.
On the second part the data transfer rate is limited by the hard disk drive, as the speeds of ATA/100 and FirwWire HDDs level out.
On the third part of the graph, where the data pack size is the most optimal for both: HDD and interface, the data transfer rate is limited by the bus width and the quality of driver optimization.
The diagram actually explains very well the results we have just seen in HDTach. Since this test reads and writes 1MB blocks and in this case the ATA/100 drive boasts the maximum advantage over its FireWire competitor, no wonder that FireWire solution looks not very successful against the background of the ATA/100 HDD.
Intel IOMeter: SequentialWrite Pattern
In case of writes, no wonder happens and everything is pretty predictable, except a significant performance drop by both hard drives when the data pack they worked with was 1KB big. I will tell you something secretly now: this is a family trait of all WD hard drives.
Note that in case of reasonable data packs the write rate shown by FireWire drive is twice as low as that of ATA/100 WD 1200BB drive. The performance reduction here is evidently caused by the insufficient bandwidth of the FireWire interface, which appears unable to tackle the data stream produced by the contemporary HDD.
By the way, in this respect I suddenly recalled an interesting case. Some time ago at one of the WD seminars Brian Frederiksen, WD technical specialist, was nearly attacked by one of the Russian journalists. He very aggressively asked why they still used 5.400rpm spindle rotation speed. He was pretty sure that 5,400rpm is far not enough for work with video…
As we see, WD decided to meet the wishes of the Russian press and now they are using 7,200rpm drives in their FireWire solutions, which definitely tells on the price of the external drives, as we all understand. Contemporary 5,400rpm HDDs are capable of writing the data at 20MB/sec, so it would be correct to say that FireWire interface and them are simply made for one another.
In conclusion I would like to offer you the first public results (this time of only two hard disk drives) obtained in our new test. You can find the description of this test and our testing methodology in this article.
Briefly, the main idea of this test implies that the HDD tested was formatted as two equal logical drives (in this review we have the results for file-copy speeds only under FAT32). Then we created a test set of files on the first logical drive and measured the time we needed to complete this task. To make it easier to estimate the HDDs performance, we put not the time but the average copy speed for all files of the patterns in the tables below:
To make the whole thing more illustrative, I suggest building diagrams:
As we see, the maximum gap between the drives with ATA/100 and FireWire interface takes place in case of larger files (in ISO, Install and MP3 patterns). When we have a big set of smaller files, the performance of hard drives with different interfaces differs insignificantly.
Then, the set of files on the first logical drive is "read", i.e. all the files of the pattern are stored in the memory (again we measure the time spent):
During reads WD FW hard drive fell behind its ATA/100 counterpart by about 40-50% in case of larger files and by 13-20% in case of many smaller files. Please take note that when working with larger files the FireWire interface appeared a bottleneck for the WD FW drive, and ATA/100 HDD managed to reach the maximum speed of 49.5MB/sec!
Theoretically, the problems of the hard disk drive with slower interface could get aggravated during file copy operations, but as we may see, this never happened in any pattern except ISO:
I would like to draw your attention to the fact that ATA/100 HDD reached its top speed when copying files from the ISO pattern, while FireWire model copied ISO files slower than MP3 files, for instance.
When the files get copied to the second logical drive the copy time increases, as the HDD need more time to move the heads to the middle of the drive (and farther). Here the performance gap between FireWire drive and ATA/100 drive gets reduced in some mysterious way (of course, since the HDD may still transfer something while the heads are flying to the required tack…)
Thank you, guys, for reading that far!
I have to state that our first acquaintance with FireWire interface and a very interesting external drive from Western Digital has been successfully completed.
The benchmarks showed that FireWire bandwidth of 400Mbit/sec is quite enough to prevent a contemporary ATA/100 hard drive connected to your PC via FireWire interface from becoming a hand brake. As we saw in most tests, the performance difference between the FireWire WD drive and the same ATA/100 drive never exceeded 20%, which corresponds to the difference between 5,400rpm and 7,200rpm HDDs. Yes, the external drive appeared slower than its ATA/100 fellow, but we shouldn't forget that WD1200BB is one of the today's fastest IDE drives. And it is not shameful to lose to a worthy rival (here I shed a few tears remembering the football match between Russia and Belgium…)
We also discovered that you'd better not use NTFS file system to format your FireWire HDD with the 4KB default cluster, as this will not make the use of this drive highly efficient and it will not be able to show the best performance. However, Western Digital has already taken care of it: the user receives WD External Hard Drive already formatted under FAT32 (at least we received it like that).
As for the user features of the drive, I will try to sum them all up below:
- Stylish looks (the product is designed in Apple-like style);
- High performance;
- Large storage capacity.
- High price.
In conclusion I would like to thank Western Digital for the 120GB WD External Hard Drive.