Our initial look at Western Digital’s 10,000 RPM Serial ATA Raptor left many readers disappointed. However, the drive, an early pre-production concept model, lacked numerous features and refinements that substantially impact performance. For this final review, WD provided a production model unit. How large of a difference do the changes make? Join us for a comprehensive review of the most highly anticipated drive in ages!
|
Western Digital is in many ways the perfect company to lead ATA to a next-generation spindle speed. Ever since it introduced the Caviar WD400BB, WD has consistently led the field in ATA performance. That is a 2.5-year run at the top- very impressive in the competitive computer hardware field. More importantly, however, the firm has no SCSI business to protect. The last thing that established SCSI powerhouses such as Seagate, IBM, and Maxtor want to see is the erosion of the relatively cushy margins associated with SCSI drives. Now that WD has opened this veritable Pandora’s Box, the competition is sure to follow.
According to WD, the key factor holding back higher spindle speeds was parallel ATA’s lack of specification-level hot swap functionality. To be successful (initially, at least), any 10k RPM ATA drive must gun for the enterprise market. And in the enterprise, a sector that views minimal downtime as critical, the ability to seamlessly swap out a failed drive another is crucial. Serial ATA provides for such hot-swap functionality. Now that SATA is trickling into the channel, WD believes 10k RPM ATA’s time has arrived.
The Raptor comes in just a single configuration- a single 36-gigabyte platter. WD specifies the drive’s seek time at just 5.2 milliseconds, solidly within SCSI territory. An 8-megabyte buffer accompanies the unit. Some folks may be disappointed with the disk’s relatively paltry capacity- after all, today’s SCSI drives deliver 147 GB of storage in a low-profile chassis. Much like its namesake made popular by 1993’s Jurassic Park, however, WD envisions Raptors in multiple-drive configurations running off of relatively inexpensive SATA RAID controllers. Reflecting its enterprise orientation, the Raptor claims a 1.2 million hour MTBF spec and features a five-year warranty.
It is important to note that the market for the Raptor is primarily the entry- and mid-level server segments and not the enthusiast desktop sector. When Western Digital raised the bar nearly 1.5 years ago, we repeatedly pointed out that the Special Edition (JB series) Caviar was what readers really wanted when they speculated over 10,000 RPM ATA drives. Equipped with an 8-megabyte buffer and accompanying firmware aggressively tuned for single-user scenarios, the WD1000JB easily matched and even exceeded the performance that the best 10k RPM SCSI drives of the era delivered when it came to desktop performance. While SCSI disks feature superior mechanics, their server orientation forces them to trade away firmware optimized for highly-localized patterns in favor of strategies that maximize returns in random access scenarios. In the Raptor, WD faces much of the same quandary. With its enterprise-class warranty and seek time, however, the firm attempts to target server markets.
Initial tests run on a pre-production sample supplied by long-time sponsor HyperMicro yielded disappointing results. Immediately afterwards, however, Western Digital was eager to get a production-class unit in our hands. Retrospectively, there were three key factors that prevented the beta sample from achieving maximum performance. First was the lack of write caching, especially apparent by the model’s poor showing in the SR High-End DriveMark 2002, a pattern that emphasizes write performance. Secondly, the pre-production drive utilized platters that were not optimally low-level formatted. Lastly, the drive suffered from an older firmware revision.
The production sample, of course, addresses these concerns. Our initial beta sample figures were drawn in conjunction with a Promise SATA TX4 controller. Though we also ran tests with a Silicon Image 3112A chipset, the combination resulted in exceptionally poor results likely arising through some unpredictable interactions between the drive’s firmware and the SI chipset. The production unit, fortunately, does not exhibit the same problems. Though Promise’s controller delivers superior results with the Raptor, we have presented results with SI’s chipset due to its prevalence in the controller market.
How does the Raptor perform? Let us turn to some numbers!
Low-Level ResultsFor diagnostic purposes only, StorageReview measures the following low-level parameters: Average Read Access Time– An average of 25,000 random accesses of a single sector each conducted through IPEAK SPT’s AnalyzeDisk suite. The high sample size permits a much more accurate reading than most typical benchmarks deliver and provides an excellent figure with which one may contrast the claimed access time (claimed seek time + the drive spindle speed’s average rotational latency) provided by manufacturers. WB99 Disk/Read Transfer Rate – Begin– The sequential transfer rate attained by the outermost zones in the hard disk. The figure typically represents the highest sustained transfer rate a drive delivers. WB99 Disk/Read Transfer Rate – End– The sequential transfer rate attained by the innermost zones in the hard disk. The figure typically represents the lowest sustained transfer rate a drive delivers. |
For more information, please click here.
|
Note: Scores on top are better. |
|
||
WD360GD Average Read Service Time
|
||
|---|---|---|
The Raptor turns in an average measured read access time of 8.6 milliseconds when operating on Promise’s controller. Subtracting three milliseconds to account for the rotational latency of a 10,000 RPM spindle speed yields an average measured seek time of 5.6 milliseconds, a bit above Western Digital’s 5.2 ms claim. Interestingly, however, when paired with the Silicon Image adapter, average access time drops to 8.3 milliseconds. The resulting seek time of 5.3 ms is right on the money.
|
Note: Scores on top are better. |
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||
WD360GD Transfer Rate
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||
With either controller, the final Raptor exhibits significant improvements in sequential transfer rates. Outer zone rates hit 63.1 MB/sec; while not quite up there when compared to today’s state of the art SCSI drives, the Raptor nonetheless manages the best transfer rates ever for an ATA drive. The unit’s rates decay less than normal as we move towards inner zones, however, resulting in an impressive 45.0 MB/sec minimum transfer rate.
Single-User PerformanceStorageReview uses the following tests to assess non-server use: StorageReview.com Office DriveMark 2002– A capture of 30 minutes of actual computer productivity use that exactingly recreates a typical office-style multitasking environment. The applications include: Outlook XP, Word XP, Excel XP, PowerPoint XP, Calypso (a freeware e-mail client), SecureCRT v3.3 (a telnet/SSH client), CuteFTP Pro v1.0 (an FTP/SSH client), ICQ 2000b), Palm Hotsync 4.0, Gravity 2.3 (a Usenet/newsgroups client), PaintShop Pro v7.0, Media Player v8 for the occasional MP3, and Internet Explorer 6.0. StorageReview.com High-End DriveMark 2002– A capture of VeriTest’s Content Creation Winstone 2001 suite. Applications include Adobe Photoshop v5.5, Adobe Premiere v5.1, Macromedia Director v8.0, Macromedia Dreamweaver v3.0, Netscape Navigator v4.73, and Sonic Foundry Sound Forge v4.5. Unlike typical productivity applications, high-end audio- and video- editing programs are run in a more serial and less multitasked manner. The High-End DriveMark includes significantly more sequential transfers and write (as opposed to read) operations. |
StorageReview.com Bootup DriveMark 2002– A capture of the rather unusual Windows XP bootup process. Windows XP’s boot procedure involves significantly different access patterns and queue depths than those found in other disk accesses. This test recreates Windows XP’s bootup from the initial bootstrap load all the way to initialization and loading of the following memory-resident utilities: Dimension4 (a time synchronizer), Norton Antivirus 2002 AutoProtect, Palm Hotsync v4.0, and ICQ 2000b.
StorageReview.com Gaming DriveMark 2002– A weighted average of the disk accesses featured in five popular PC games: Lionhead’s Black & White v1.1, Valve’s Half-Life: Counterstrike v1.3, Blizzard’s Diablo 2: Lord of Destruction v1.09b, Maxis’s The Sims: House Party v1.0, and Epic’s Unreal Tournament v4.36. Games, of course, are not multitasked- all five titles were run in a serial fashion featuring approximately half an hour of play time per game.
For more information, please click here.
|
Note: Scores on top are better. |
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Equipped with firmware designed by an engineering team that has had a lock on the StorageReview.com 7200 RPM Leaderboard slot for over two years, the Raptor really flexes its muscles when it comes to single-user performance. The drive hits 500 I/Os per second in the SR Office DriveMark 2002 when paired with Promise’s controller, virtually matching the score turned in by the Maxtor Atlas 10k IV and significantly besting Seagate’s Cheetah 10K.6.
The Raptor is even more breathtaking in the High-End DriveMark 2002. One may recall the dismal 300 I/Os per second that the beta sample, crippled with no write caching, produced. The production unit, however, delivers an astonishing 524 I/Os per second, easily the best score we have ever recorded. In this test, the Raptor crushes the Atlas by an 18% margin!
WD improves upon an already strong Bootup DriveMark showing. Here the Raptor scores 526 I/Os per second to again set a new record. In the Gaming DriveMark, the drive finally lets up a bit, trailing the Atlas by about 3%. The Raptor’s score of 625 I/Os per second nonetheless is an impressive showing, however.
The Raptor’s relatively small 36-gigabyte capacity underscores its surprising results. The SR Desktop DriveMarks, as real-world, high-level disk accesses, exhibit better scores on larger drives. In the above graphs, the 36 GB Raptor faces off against 147-gig SCSI drives and ATA drives featuring up to 200 gigabytes of capacity. One can only speculate on the spectacular single-user showing a 108- or 146-gig Raptor would deliver.
Multi-User PerformanceStorageReview uses the following tests to assess server performance: StorageReview.com File Server DriveMark 2002– A mix of synthetically-created reads and writes through IOMeter that attempts to model the heavily random access that a dedicated file server experiences. Individual tests are run under loads with 1 I/O, 4 I/Os, 16 I/Os, and 64 I/Os outstanding. The Server DriveMark is a convenient at-a-glance figure derived from the weighted average of results obtained from the four different loads. StorageReview.com Web Server DriveMark 2002– A mix of synthetically-created reads through IOMeter that attempts to model the heavily random access that a dedicated web server experiences. Individual tests are run under loads with 1 I/O, 4 I/Os, 16 I/Os, and 64 I/Os outstanding. The Server DriveMark is a convenient at-a-glance figure derived from the weighted average of results obtained from the four different loads. For more information click here. |
|
Note: Scores on top are better. |
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||
In the SR File Server DriveMark 2002, a normalized average of typical file server access run under 4 different queue depths, the production Raptor scores 193 I/Os per second, an improvement of 9%. While it significantly outpaces all other ATA drives, WD’s newcomer still can not touch the performance of a 10,000 RPM SCSI drive. Similar margins occur in the Web Server DriveMark.
Though the serial ATA spec supports command queuing, the Raptor does not incorporate the feature. As a result, the drive lacks the ability to efficiently reorder requests at the device level, a key feature when it comes to multi-user performance. This omission is a bit strange considering that the Raptor guns for the enterprise server market.
Overall, when it comes to server performance, the Raptor delivers the performance one would expect from a 7200 RPM SCSI drive. While it doesn’t hit 10k RPM SCSI levels, the drive can service as segment left out in the cold due to the death of entry-level SCSI disks.
Legacy PerformanceeTesting Lab’s WinBench 99 Disk WinMark tests are benchmarks that attempt to measure desktop performance through a rather dated recording of high-level applications. Despite their age, the Disk WinMarks are somewhat of an industry standard. The following results serve only as a reference; SR does not factor them into final judgments and recommends that readers do the same. |
|
Note: Scores on top are better. |
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Heat and NoiseIdle Noise– The sound pressure emitted from a drive measured at a distance of 18 millimeters. The close-field measurement allows for increased resolution between drive sound pressures and eliminates interactions from outside environmental noise. Note that while the measurement is an A-weighted decibel score that weighs frequencies in proportion to human ear sensitivity, a low score does not necessarily predict whether or not a drive will exhibit a high-pitch whine that some may find intrusive. Conversely, a high score does not necessarily indicate that the drive exhibits an intrusive noise envelope. Net Drive Temperature– The highest temperature recorded from a 16-point sample of a drive’s top plate after it has been under heavy load for 80 minutes. The figures provided are net temperatures representing the difference between the measured drive temperature and ambient temperature. For more information, please click here. |
|
Note: Scores on top are better. |
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Objectively speaking, the beta Raptor turns in impressively low noise floors, likely due to its single-platter design. A score of 42.5 dB/A approaches the noise floor delivered by the latest Barracuda ATA drives. Subjectively, it is interesting to note that while the final version of the drive weighs in with a higher noise floor than the beta, it tempers the high-pitched whine exhibited by the latter and thus is less obtrusive. It nonetheless still emits a very slight squeal reminiscent of third- or fourth-generation 10k RPM SCSI units. Though the sound does not bother me personally, it may irritate those who insist in ultra-quiet operation from their machines. The Raptor’s seek noises definitely land significantly higher than today’s ATA drives. While not quite as loud as a SCSI disk, it is nevertheless easy to tell when the drive’s actuator is in motion.
The Raptor’s single-platter configuration also yields a relatively low operating drive temperature. Our measurements reached 20.8 degrees Celsius above ambient room temperature- on the high side for an ATA drive but well below the typical SCSI unit.
ReliabilityThe StorageReview.com Reliability Survey aims to amalgamate individual reader experiences with various hard disks into a comprehensive warehouse of information from which meaningful results may be extracted. A multiple-layer filter sifts through collected data, silently omitting questionable results or results from questionable participants. A proprietary analysis engine then processes the qualified dataset. SR presents results to readers through a percentile ranking system. According to filtered and analyzed data collected from participating StorageReview.com readers, the |
Note that the percentages in bold above may change as more information continues to be collected and analyzed. For more information, to input your experience with these and/or other drives, and to view comprehensive results, please visit the SR Drive Reliability Survey.
ConclusionOur preview noted the following:
|
In the end, ironically, the Raptor is a world-class desktop unit and a decent but unspectacular server performer. As we have already remarked, the Raptor’s server performance falls somewhere between that of a 10k SCSI drive and 7200 RPM ATA units. It can admirably fill the niche formerly served by the defunct 7200 RPM SCSI segment. To have a shot at higher-end multi-user applications, however, future Raptors will have to incorporate command queuing, a feature that allows SCSI drives to scale well beyond the Raptor’s showing as I/O depths.
WD’s drive, however, delivers single-user performance that matches and even exceeds not only 10k RPM SCSI but also that of 15k RPM units. It is the answer to many years of prayers from power-users and enthusiasts worldwide.
