Western Digital’s Raptor series remains alone as the only 10,000 RPM ATA drive series. As impressive as the WD360GD was, however, it did come up short in a few areas, namely capacity, idle noise, and multi-user performance. Western Digital claims to have addressed all three issues while also increasing performance. Has the ATA titan delivered? Join SR for a special sneak peek at the next-generation Western Digital Raptor WD740GD.
|
Ironically, it appears the Raptor may have met its greatest success not in the server market but rather with the enthusiast/gaming community itself. The drive lacked a key feature, tagged command queuing, which prevented its performance from scaling up under multi-user loads as adeptly as the typical SCSI drive did. Further, the Raptor’s firmware, designed by a company that had already proved itself as the best at optimizing for the desktop, allowed it to match and even exceed the single-user performance delivered by the latest-generation 10k RPM SCSI units. Retailers such as Alienware, Falcon Northwest, and even Dell offer the Raptor as an option on their high-end desktops. The drive even appeared wrapped in retail packages at certain brick-and-mortar superstores.
StorageReview.com’s initial Raptor was provided by long-time sponsor Hyper Microsystems. As an engineering sample, the drive proved to resellers that WD was capable of manufacturing SATA 10k RPM units. They did not, however, incorporate near-final firmware and also lacked a key feature, device-level write caching. In the end, the beta unit delivered lackluster performance.
Western Digital quickly responded by supplying an evaluation unit that was to represent the performance one could expect from the shipping product. This second sample featured more recent firmware and had write caching fully operational. Further, it featured a more aggressive zone layout that yielded a maximum transfer rate of 63 MB/sec, up from the first sample’s 57 MB/sec. Buoyed primarily by enabled write caching, the Raptor seized a top spot in the StorageReview Desktop DriveMarks.
As users got their hands on retail drives, however, they quickly noticed that their shiny new hotrods topped out with 57 MB/sec transfer rates rather than the 63 MB/sec that SR’s WD-supplied sample boasted. Sequential transfer rates (STR), unfortunately, are the easiest HD performance metric to measure and also the least indicative of how the drive will perform under real-world situations. It was obvious from the lower STR figures that the shipping Raptors incorporated the more conservative zone-layout found in our original sample rather than the one used in the following unit.
Conspiracy theories ensued in the community. Eventually, a flawed consensus was reached: If the shipping Raptors featured the lower transfer rates exhibited in SR’s first sample, then the write-caching, firmware, and performance of shipping units must also be similar to that of the first drive’s. This logic, of course, is fallacious. Zone configuration (and thus transfer rates) is independent of write caching and firmware.
We eventually contacted Western Digital on the issue. The firm responded by stating that while they had initially hoped to ship Raptors featuring the more aggressive zone layout, yields on such platters were too low and as a result would knock the Raptor above its targeted price. To meet the drive’s price point, WD reverted to the more conservative setup.
Afterwards, Hyper Microsystems supplied us with a third, retail sample. Around the same time, to fix unrelated problems in another serial ATA drive review, we were forced to update our normally-static Testbed’s SATA controller’s BIOS and driver. Hence, a proper comparison between the second (higher STR) and third (retail) samples required a retest of the second drive. Let’s briefly recap the tested differences:
| 2nd Sample | 3rd Sample | |
|---|---|---|
| Random Access Time | 8.6 ms | 8.6 ms |
| Maximum STR | 63.0 MB/sec | 57.4 MB/sec |
| SR Office DriveMark 2002 | 500 IOs/sec | 483 IOs/sec |
| SR High-End DriveMark 2002 | 495 IOs/sec | 467 IOs/sec |
| SR Bootup DriveMark 2002 | 465 IOs/sec | 410 IOs/sec |
Basically, while the third sample’s transfer rate slid by about 9%, most application-level marks fell by about 3%. Only the Bootup DriveMark, heavily dependent on transfer rates, fell by a more significant margin. The average loss of 3% elsewhere likely comes from more conservative firmware, as a change from 63 MB/sec to 57 MB/sec simply does not impact today’s applications in a meaningful way.
The bottom line? Despite the lower-than-anticipated transfer rates, the Raptor nonetheless was easily the fastest ATA drive ever and the ideal mix between price and performance when both ATA and SCSI drives were considered.
Since then, however, Hitachi Global Storage Technologies introduced the marvelous Deskstar 7K250. Though incorporating a more conventional 7200 RPM spindle speed, the 7K250 nonetheless features a wide capacity range combined with performance that approached WD’s speed demon. Does Western Digital have anything else up its sleeve?
Enter the second-generation Raptor! Announced September 15th, WD’s revised offering promised to address nearly all of the complaints leveled against the firm’s first entry. Maintaining the line’s unique 10,000 RPM spindle speed, the Raptor WD740GD features the following improvements:
- 74-gigabyte capacity – perhaps the most significant improvement is the migration to a two-platter flagship design. WD also plans to introduce a revised single-platter, 37 GB unit, though perhaps not until most WD360GD units sell through the market.
- 37 GB Platters – the aggregate areal density of the new Raptor will remain the same as the first. Linear density, however, has been increased, to achieve:
- 72 MB/sec outer-zone transfer rates – though STR remains non-consequential in the large majority of uses, some folks were disappointed with the 55 MB/sec that the first Raptor delivered. WD is confident enough with new yields to spec a transfer rate that rivals the best available from today’s disks.
- 4.5 millisecond seek time – the Raptor WD360GD specs at 5.2 milliseconds.
- Firmware-level TCQ – matching a feature available on all contemporary SCSI drives, the new Raptor will feature tagged command queuing… that is, device-level reordering of outstanding requests for more efficient service times.
- FDB motors – though quiet from an emitted sound-pressure perspective, the original Raptor emitted a slight high-pitch idle whine that could irritate sensitive ears. WD has been on the slow side when it comes to migrating to fluid bearing motors when compared to other manufacturers. Fortunately, the new Raptor uses quieter and ostensibly more reliable FDB motors.
Maintaining a 5-year warranty and adding command queuing, with the new Raptor WD once again sets its wanton eyes on the low- to mid-end server market currently enjoyed by today’s 10k RPM SCSI drives. Of course, the manufacturer must also take care to ensure that the drive’s blazing desktop performance does not suffer.
Interestingly, the new Raptors will still use an onboard PATA-to-SATA bridge disdained by many enthusiasts. When hot-swap functionality, command queuing, and the potential for blazing performance remain, however, why should we care? Performance and functionality, not PCB layouts, remain the bottom line. As has been the case with all bridged solutions (that is to say, all serial ATA drives except Seagate’s Barracuda series), the new Raptor features both a newer 15-pin power connector as well as the more traditional 4-pin molex receptacle.
Western Digital recently supplied StorageReview with an engineering sample of the 740GD upon which we have based the following preview. We have been putting it through its paces over the last two weeks in a variety of objective and subjective scenarios. Before we turn to the numbers, let us keep a key point in mind- this drive is an evaluation sample. It lacks the tagged command queuing that WD promises will be enabled in shipping units. What kind of impact does the lack of TCQ have in our tests? In the server suites, this sample will not scale upwards as effectively as a SCSI drive under heavier loads. Under single-user situations, ramifications are not quite as clear. Even with heavy multi-tasking, generated queue depths remain relatively light. As a result, command queuing may not help- it may even exact a penalty. More generally, remember that all things are of course subject to change between this engineering sample and final retail products.
That said, let us see how this eagerly anticipated drive stacks up!
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. |
|
||||||||||||||||||
|
WD740GD Average Read Service Time |
|||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
The Raptor WD740GD’s measured access time weighs in at 8.3 milliseconds. Subtracting 3.0 ms to account for the rotational latency of a 10k RPM spindle yields a measured access time of 5.3 milliseconds. Though this figure bests the mark set by the drive’s predecessor (the WD360GD), it nonetheless misses Western Digital’s 4.5 ms claim by some margin.
|
Note: Scores on top are better. |
|
|||||||||||||||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||||||||||||||
WD740GD Transfer Rate |
|||||||||||||||||||||||||||||||||||||||||||||||
Transfer rates, on the other hand, pan out as anticipated. In its outer zone, the new Raptor delivers 72.2 MB/sec, a score that surpasses all 10k RPM SCSI drives. Inner zone rates are even more impressive, decaying down to 54.1 MB/sec, a figure previously delivered only by 15,000 RPM disks.
This said, we must take one more moment to emphasize that sequential transfer rates are not important in most single-user and multi-user scenarios. While interesting from an academic standpoint, overuse of such “benchmarks” as SiSoft Sandra, HDTach, and ATTO PowerTools has contributed to an unwarranted fixation on transfer rates. Consider, for example, the Maxtor DiamondMax Plus 8 (not shown). The manufacturer makes no pretense as to the drive’s market- the most entry-level of machines. This offering, however, delivers an outer-zone transfer rate (which is more or less the -only- thing ATTO and Sandra measure!) of 60.1 MB/sec. Those who erroneously depend on benchmarks such as these would thus draw the conclusion that the DiamondMax Plus 8 is a “faster” choice than the original Raptor, a drive which tops out at 57.4 MB/sec. In reality, the Raptor is twice as fast in typical real-world cases. Do not rely on transfer rates to judge a drive’s performance.
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. |
|
|||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||
Weighing in at an astounding 575 I/Os per second in the StorageReview Office DriveMark 2002, the Raptor WD740GD ranks among the industry’s elite when it comes to drives servicing a single-user. To put this gain in proper perspective, consider the difference between the 740GD and 360GD (19%) versus the 360GD and the 7200 RPM Caviar WD2500JD (16%).
The Raptor’s showing only improves in the SR High-End DriveMark 2002. This suite is influenced slightly more by transfer rates due to its applications’ reading and writing of large files- though even here, firmware optimizations are paramount. The Raptor’s decent transfer rates, combined with single-user-friendly programming, allow it to dominate with a score of 568 I/Os per second. Even Fujitsu’s mighty MAS3735 (not shown), the previous record holder in -every- one of our performance measures, has been unseated.
Western Digital’s newcomer continues to impress in the Bootup DriveMark, a scenario that combines higher-than-normal queue depths with largely sequential access. Here the WD740GD places squarely in the middle of today’s pack of 15k RPM drives.
Finally, the manufacturer has not forgotten the warm reception it has received from the gaming community. Under the SR Gaming DriveMark, a weighted average of 5 PC games, the new Raptor achieves 715 I/Os per second, once again besting all 10,000 RPM drives and keeping elite 15K’s company.
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. |
|
|||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||
The engineering sample’s lack of tagged command queuing catches up to it when pitted against today’s SCSI units in our server suite. While the Raptor commences quite competitively at low queue depths (not shown- viewable via drive comparison in the Performance Database), its lack of TCQ causes it stumble as queue depths increase.
These two SR Server DriveMarks, a weighted average of a drive’s performance under varying load levels, demonstrate that though it offers significantly superior multi-user performance when contrasted to 7200 RPM drives and though it also improves upon its predecessor, without command queuing the WD740GD can not keep pace with today’s 10,000 RPM units.
It is these results, however, that are most subject to change when a final retail product is mated with an appropriate controller. Although the Raptor’s TCQ will be implemented in firmware (as oppose to hardware in most SCSI drives), its presence will dramatically increase the WD740GD’s viability as loads scale upwards.
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. |
|
|||||||||||||||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||||||||||||||
While it objectively measured in at just 42.5 dB/A from a distance of 18 millimeters, the original Raptor nonetheless emitted a continuous high-pitched squeal that was glaringly obvious under our measurement test conditions. Even when placed in a case with (fairly quiet) power supply and CPU fans, the drive’s whine remained audible. We’re thus very pleased to report that the Raptor WD740GD offers a whisper-quiet noise floor. The objective numerical difference between the 42.5 dB/A WD360GD and 40.7 dB/A WD740GD simply does not do justice to the improvements the latter delivers. Though it features two platters and a 10,000 RPM spindle speed, WD’s first exclusively fluid dynamic bearing- based drive delivers noise floors every bit as competitive as Seagate’s Barracuda ATA series. That’s pretty darn quiet!
Despite the powerful actuator necessary to deliver sub-9ms access times, the WD740GD’s seek noises remain impressively muted. Even with a full random seek test running, the Raptor’s noise weighs in as only slightly audible over our testbed’s Antec PP412X power supply and retail P4 heatsink fan.
The WD740GD’s 10,000 RPM operation, two-platter design, and perhaps its new FDB motors all combine to create an operating temperature of 22.7 degrees Celsius over ambient room temperature. As a result, the Raptor places hotter-running than all other ATA drives and yet slightly below today’s 4-platter SCSI designs. Unlike with virtually any other ATA drive, care must be taken to ensure adequate cooling and ventilation when installing the new Raptor. Considering the drive’s target market of enthusiasts and servers, however, this should not be a problem.
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 |
According to filtered and analyzed data collected from participating StorageReview.com readers, a predecessor of the
Western Digital Raptor WD740GD, the
Western Digital Raptor WD360GD
, is more reliable than
of the other drives in the survey that meet a certain minimum floor of participation.
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.
ConclusionThe single-user performance delivered by this WD740GD engineering sample ranks among the most impressive we have yet measured. Further, the new Raptor’s noise floors when both idle and seeking compete with those of the quietest 7200 RPM ATA drives. The lack of command queuing, however, causes WD’s drive to stumble a bit in multi-user scenarios when contrasted with SCSI units. Remember, however, that the manufacturer’s press release and literature has explicitly referred to the advantages that queuing will deliver- we’d be surprised if the feature did not make its way into the shipping product. |
Western Digital indicates that the drive’s anticipated street price will be about $300. It is always difficult to exactly quantify the relationship between estimated street price and the price delivered by competitive yet reliable resellers. Taking the estimate at face value and contrasting it with HyperMicro’s (at the time of this writing) price of $359 for either the Seagate Cheetah 10K.6 or Maxtor Atlas 10k IV yields a savings for the SATA product of about 20%… this does not factor in the significantly lower cost of SATA adapters.
One can not draw a definitive conclusion from an engineering sample. However, if the final product’s command queuing capability yields server performance rivaling that of SCSI drives while maintaining the exceptional single-user speed and quiet operation we’ve already witnessed, the Raptor WD740GD will rank as one of the most impressive drives we have seen in StorageReview’s five-and-a-half year history. Let’s hope for the best!
