Original Link: http://www.anandtech.com/show/1082
Western Digital's Raptor - Part I: The World's Fastest Desktop Driveby Anand Lal Shimpi on March 11, 2003 7:46 PM EST
- Posted in
Just last week we previewed Western Digital's brand new enterprise-class drive - the Raptor. The 10,000 RPM hard drive garnered quite a bit of attention, not because of its spindle speed, it's 8MB buffer or its 5 year warranty; the Raptor was thrown into the limelight because it can deliver all the specifications of a high-end SCSI drive, but with a Serial ATA interface.
For years, performance enthusiasts have asked why there hasn't been a 10,000 RPM IDE drive for the desktop, and for years the manufacturers have responded with a lack of demand for such a product.
Western Digital continues to believe that, which is why the Raptor is classified as an enterprise drive and not a successor to their highly successful Caviar line of desktop hard drives. But there are a number of enthusiasts that are using 15,000 RPM Seagate drives in their desktop systems, so the Raptor has a viable user base in a non-enterprise market as well.
All of this discussion about the potential uses for the Raptor is meaningless if the drive cannot deliver world class performance and based on our preview, the Raptor had a lot of growing up to do before it could achieve that goal.
Shortly after we published our preview, we were contacted by a number of other OEMs and system integrators that had been playing around with similar beta samples of the Raptor drive. Everyone that contacted us had similar disappointing performance results, but one company in particular mentioned the possibility that Western Digital disabled write caching on all early Raptor samples. We worked on seeing if there was a way to enable the write cache on our sample, but the end result was that the drive would need a firmware update.
The new and old Raptor
Western Digital contacted us to arrange for a new sample, with write caching enabled, to be provided. They confirmed what we had heard, as well as mentioned that there were other optimizations present in the shipping firmware that weren't there before; by far the biggest performance improvement would be due to enabling write caching. A day later, we had a new Raptor drive in our hands and started putting it through the paces
Breeding a Raptor
In talking to Western Digital, we managed to get a bit of clarification about exactly what technology was borrowed from previous drives in order to maintain the Raptor's lower-than-SCSI price point.
As you know from our original article, the Raptor is a single platter design with an initial capacity of 36GB. The reason Western Digital went for a relatively low capacity for the Raptor is because of the target market (we'll address this more in the next section). With the vast majority of SCSI drives in use in the enterprise market limited to 18 - 36GB size ranges, Western Digital figured a 36GB introductory capacity would make sense. Since there's only one platter in action on the Raptor, there are clear chances to increase capacity simply by moving to more platters, but for the near future you won't see more than a 36GB capacity available.
Western Digital expects 36GB Raptor drives to sell for around $160, compared to $200+ for most 36GB 10,000 RPM SCSI offerings. Where's the 20% lower price coming from?
It's not coming from smaller capacities because we're comparing apples to apples here, and it's definitely not coming from the interface, because implementing a Serial ATA connector vs. a SCSI connector doesn't reduce the overall price of a drive by 20%.
It turns out that Western Digital was able to save a significant amount by reusing the drive electronics from their 7200 RPM Special Edition Caviar drives. A combination of essentially "free" electronics from their Caviar line and a slightly slower actuator assembly than what the fastest SCSI drives use left Western Digital with a SCSI class drive, at a 20 - 30% lower cost.
A Raptor Loose in the Enterprise Market
Along with talking to Western Digital about the technical aspects surrounding the Raptor drive, we also asked about positioning. As we just alluded to, Western Digital is marketing the Raptor as a low-cost 10,000 RPM drive for the enterprise market; in essence, they are going after the SCSI market, something they are able to do because they have no SCSI line to cannibalize by doing this.
Western Digital has no delusions about being the fastest performer at 10,000 RPM, but what they are promising is a much lower total cost of ownership and competitive performance from their Raptor.
The lower total cost of ownership comes from the drive itself, which we already know will be between 20 and 30% cheaper than an identical SCSI drive, but there are a couple of other factors to take into account.
For starters, Serial ATA cabling is significantly cheaper than 68-pin SCSI cabling. Not only is it cheaper though, it is also much better suited to improving cooling within a chassis because of the very thin cables.
Next we have the cost of controller cards; remember that by the end of the first half of this year, Serial ATA will be integrated into all Pentium 4 I/O Controller Hubs (ICH5) courtesy of the 865 and 875 chipsets. That transition will also happen with server specific chipsets, although somewhat later. With integrated Serial ATA, the controller basically comes free of charge, as opposed to the pricey Ultra320 SCSI add-in cards.
The final point is that the cost of RAID will be significantly cheaper with Serial ATA as well, once again going back to the price of the controllers.
We mentioned in the previous section that Western Digital's decision to stick with a 36GB capacity was because of the fact that the majority of 10,000 RPM SCSI drives in use are either 18GB or 36GB in size. Western Digital's research shows that 36GB should be the target capacity for the rest of this year, but as we mentioned before, if there's room for competition with higher capacities then the Raptor will adapt.
You will be able to begin ordering the Raptor drives through resellers (not retail outlets however) in the next week or so.
Now it's time to find out if Western Digital's claims of improved performance on the updated Raptor are valid. Because of the enterprise target for this drive we've split our review into two parts, the first part is a rerun of all of the tests from our Raptor preview with the new drive. The second part will introduce a few more enterprise level tests and will focus entirely on server performance.
Our benchmarking methodology has already been outlined in our Raptor preview; we strongly suggest reading our one page discussion on hard drive benchmarks and how we're going to be doing things here on AnandTech before going any further with this review, it will help you in understanding the benchmarks.
Windows XP Professional Test Bed
Intel Pentium 4 3.06GHz (Hyper-Threading Enabled)
Intel 850E Chipset
2 x 256MB PC1066 Kingston RIMMs
|Hard Drive (Boot)||
120GB Western Digital Special Edition 8MB Cache HDD
Image Sil3112ACT144 SATA Controller Card
|Video Cards (Drivers)||
ATI Radeon 9700 Pro (128MB) - CATALYST 3.1
A special thanks to Hypermicro for providing us with a number of the hard drives and controller cards for this review.
Access Time & Transfer Rates
Although the focus of this review is entirely on real-world performance numbers, we felt the need to start out with Access Time and Transfer Rate tests to illustrate some basic differences between the drives being compared here today.
The two purple bars are the two 10,000RPM SCSI drives in this comparison, blue is for all of the ATA/SATA drives and red is for the new comer WD Raptor.
We measured the access time of the drives by conducting 25,000 randomly distributed read requests, the average service time for those requests is reported below:
As you can see, the access time improved slightly for the production version of the Raptor, probably due to improvements in the firmware.
The production drive allowed for some fairly impressive increases in sequential transfer rates, as is evident by the graphs above and below.
The Raptor is now at the top of the transfer rate chart, however it didn't do too poorly in these tests initially either - what about in our real world application test suite?
Content Creation Performance
In order to measure Content Creation performance we used Multimedia Content Creation Winstone 2003 and Internet Content Creation SYSMark 2002. First we'll start off with the Winstone benchmark, which is best described by its creators at Veritest:
Multimedia Content Creation Winstone is a system-level, application-based benchmark that measures a PC's overall performance when running top, Windows-based, 32-bit, multimedia content creation applications on Windows 2000 (SP2 or higher), Windows 98, Windows ME, and Windows XP. Multimedia Content Creation Winstone 2003 uses the following applications:
Adobe® Photoshop® 7.0
Adobe® Premiere® 6.0
Macromedia® Director 8.5.1
Macromedia® Dreamweaver 4
Microsoft® Windows MediaTM Encoder 7.01.00.3055
NewTek's LightWave® 7.5
Sonic Foundry® Sound Forge® 6.0
Following the lead of real users, Multimedia Content Creation Winstone 2003 keeps multiple applications open at once and switches among those applications. Multimedia Content Creation Winstone 2003 is a single large test that runs the above applications through a series of scripted activities and returns a single score. Those activities focus on what we call "hot spots," periods of activity that make your PC really work--the times where you're likely to see an hourglass or a progress bar
We start out by isolating all of the disk accesses that take place during a run of the benchmark and running them on each of the individual drives.
How's that for an improvement? Enabling write caching and a few firmware tweaks left the Raptor as the fastest drive for Content Creation applications. What we end up having is an enterprise drive that is excellent for non-enterprise applications as well.
The performance improvement of the Raptor translates over into the actual Winstone benchmark itself, it is now just as fast as the Special Edition Caviar, although not significantly faster.
Content Creation Performance (continued)
Next we have ICC SYSMark 2002; for this test we only have the trace results, but we're still able to draw some conclusions based on them. First, a listing of the programs used in the benchmark from the developers (BAPCo):
Adobe Photoshop® 6.01
Adobe Premiere® 6.0
Microsoft Windows Media Encoder 7.1
Macromedia Dreamweaver 4
Macromedia Flash 5
The Raptor continues its dominance in the ICC SYSMark 2002, now outpacing the Special Edition Caviar by 31%.
General Usage Performance
Although not as performance-critical as content creation applications, it is the set of every day applications like Office and other general usage programs that the majority of users find themselves interacting with the most and also happen to be very disk intensive, after all, here's where most users find themselves complaining about I/O performance.
We start with VeriTest's Business Winstone 2002:
The Business Winstone tests are "market-centered" tests. Business applications are the popular applications employed by most users every day.
Five Microsoft Office 2002 applications (Access, Excel, FrontPage, PowerPoint, and Word)
Microsoft Project 2000
We start out by isolating all of the disk accesses that take place during a run of the benchmark and running them on each of the individual drives:
The production continues to do very well, setting a new record for our Business Winstone 2002 Disk Performance test of 781 I/O operations per second, a full 19% faster than the Special Edition Caviar and over a 2x improvement over the beta drive we tested.
The real world performance of the Raptor ends up being on par with the 180GXP and the WD1200JB, much better than the poor showing of the first drive we tested.
General Usage Performance (continued)
For the desktop side of things, we conclude with Office Productivity SYSMark 2002. The applications tested include:
Microsoft Word 2002
Microsoft Excel 2002,
Microsoft PowerPoint 2002
Microsoft Outlook 2002,
Microsoft Access 2002,
Netscape Communicator® 6.0
Dragon NaturallySpeaking Preferred v.5
McAfee VirusScan 5.13.
Once again we only have the trace results for Office Productivity SYSMark 2002, but we use them to confirm our previous findings:
The Raptor sets another record, this time outpacing the Caviar by 122 I/O operations per second.
AnandTech Forums Database Server Performance
For our final performance test we have a trace of all I/O operations that occur during our AnandTech Forums DB Server trace. For those of you that aren't familiar with this test, we've historically used it in only our enterprise server CPU and platform reviews, but we decided to introduce it here because of its relevance to the topic at hand.
Here's a description of the test:
The database we tested is the Forums DB, which is by far the most transaction intensive database in the AnandTech Network. While the vast majority of the requests to the DB are in the form of selects (users reading categories and threads), there are significantly more inserts and updates (posting, thread/post counts, etc ) than in either of the other DBs. This database is also our largest, weighing in at just under 3GB during the testing and well over 10GB today (we used an older version of the DB from over a year ago).
We recorded a 30 minute trace of all activity to the AnandTech Forums DB Server, and used that as an input to our I/O trace recording. The end result, is what we have here today:
This is a teaser of what is to come in Part II of our coverage of the Raptor, here we see that it is faster than all ATA/SATA drives, but it is still a bit slower than the two SCSI offerings. Since WD is targeting the enterprise market, performance in these tests is critical to finding out how well they deliver on the "similar performance to SCSI, but at a lower cost" promise. We will reserve judgment in that area until part II, where we focus exclusively on enterprise performance, we just wanted to give you all an update right away on the drive's performance in our original suite.
With write caching enabled and with the production level optimizations present in the drive's firmware, the Raptor is now the fastest desktop drive we've laid our hands on. This is a side effect that Western Digital wasn't really shooting for, but it is a result of the physical characteristics of the drive and Western Digital's extensive experience in producing high-performing desktop drives. Right now, the Raptor is absolutely the only performance-oriented reason to migrate to Serial ATA.
From an enterprise standpoint, we're going to hold off on making any sort of conclusions until we get a better idea of the overall performance of the Raptor in server environments. We're hard at work at translating our other SQL database tests into I/O only traces as well as new web server tests, stay tuned for our continued coverage on the Raptor...