Long-time readers of StorageReview.com no doubt realize that we've been rather down on Western Digital's Caviar series of drives. In our initial 6.4 GB ATA Drive Roundup, the Caviar placed in the middle of the pack. Since then, however, we've noted a downward slide in performance with the Caviar series when compared to similar drives from other manufacturers. Thus it was with some surprise that we noted WD's latest marketing efforts when it came to the new 13 gig AC313000. For the first time in recent memory, the press release for the drive quoted performance excellence based on a popular benchmark (in this case, ZD's own WinBench 98 v1.0). The drive's box further hammers home the point, with a large circular sticker declaring the drive 35% faster than Maxtor's DiamondMax 3400. While it may be argued with some validity that the AC313000 should be compared with something more up-to-date than the DiamondMax 3400, there's no denying WD's pride in its newest ATA drive's performance.

The AC313000 presents perhaps the last iteration of WD's aging 3-platter design. Each platter in this 5400rpm drive packs a rather contemporary 4.3 gigs of data (See Areal Density). The drive features a now-standard 9.5 millisecond random seek time. A 512k buffer rounds out the package. The AC313000 is the first drive to hit the market that incorporates the new ATA-66 standard. Proper ATA-66 operation requires an 80-conductor cable along with a motherboard that supports the spec (as of the writing of this article, there aren't many that do). Though it'll eventually become important in allowing future drives to realize their maximum transfer rates, ATA-66 doesn't do much for today's drives. Transfer rates for state-of-the-art ATA drives are hovering right around 15 MB/sec, just high enough to signal the call for ATA-33 (UltraDMA) rates. ATA-66 is a spec for the future, not the present.

The new interface did, however, manage to make installation of the Caviar an atypical procedure. StorageReview.com's testbed features the venerable Abit LX-6 motherboard (though the LX chipset is no longer the latest available, the PIIX4 [includes ATA] controller is still the most current available from Intel) which, like most mainboards, does not support ATA-66. Out of the box, the AC313000 had ATA-66 operation enabled. Thus, the Award BIOS found on the Abit detected the Caviar as a "UDMA, Mode 4" drive. Upon booting, the Windows 95 Device Manager placed a big exclamation point on the ATA controller. Something was obviously wrong. It turns out that the Award BIOS has a bug of sorts- it can detect the drive as an ATA-66 unit even when its accompanying chipset can't support ATA-66. The result is a drive running in MS-DOS compatibility mode, i.e., slow as molasses. WD was gracious enough to provide a utility to disable ATA-66 support in the drive and notes that the AC31300 units shipping in retail packages have ATA-66 disabled by default. Once disabled, installation continued smoothly.

According to WinBench 98, the AC313000 is one fast drive. Even when compared to the "revised" figures of the Maxtor DiamondMax 4320, the Caviar sails by the Maxtor by a margin of 15% with a Business Disk WinMark 98 (Win95) score of 1890. Keep in mind that this is just a hair less than the figure turned in by the original Seagate Cheetah! The WD's High-End WinMark score also nudged by the Maxtor by 6%. Maxtor drives always perform very well under Windows NT, however. Here the WD lagged its red-boxed competition by up to 16%. Even so, however, WinBench 98 NT figures were quite respectable. The Caviar's performance in ThreadMark under Win95 was also commendable, edging past the previous ATA champ, the DiamondMax Plus 2500.

It's the WinBench 98 Business Disk WinMark score, however, that draws attention, easily the highest ever produced by a non 10k rpm drive. It's so high, in fact, that a dose of skepticism may be healthy. Good thing we've begun our transition to the new WinBench 99.

Looking at the WinBench 99 results, we're starting to hear a different tune. And though the AC313000 managed to best the DiamondMax 4320 by 15% in the Business Disk WinMark 98, it lagged the 4320 by 10% in the 99 version. Ditto for the High-End WinMarks: Under 98, the Caviar led by 6%, but under 99, it fell behind by a substantial 18%. NT figures got even uglier, with the Caviar falling behind the DiamondMax by margins of up to 29%.

What's the deal? Unfortunately, we see this happen repeatedly as the year goes by. When a well-designed benchmark is freshly released, it tends to provide accurate, transportable results. As time passes by, however, hardware/firmware/software programmers "crack" the benchmark, and find ways to produce higher scores that may not necessarily translate into noticeably better real-world performance. This is the number one reason for the reactionary, knee-jerk reaction along the lines of "Benchmarks don't tell us anything! They're useless and totally unreliable." We're also treated to trite phrases such as "There are lies, damned lies, and benchmarks." Such subjectivists then go on to prove that so-and-so piece of hardware is "faster" through even less verifiable methods. While we lend such camps little credence, we do realize that benchmarks can eventually be "beat." That's why reliable sources continue to refine their benchmarks yearly. Ok, enough soapbox .

An educated guess would be that the WB99 figures correlate better to actual performance than WB98. This is confirmed by informal trials where we've "lived" with the drives in our personal systems. I hate to say it, but the AC313000's performance simply does not match the figures it turns out under WB98.

In terms of heat and noise, the AC313000 is par for the ATA course: relatively unobtrusive. The drive runs only slightly warm to the touch even outside of a drive cooler. Seeks are a rather soft, hollow metallic sound that would probably bother only the most sensitive of users.

It unfortunately appears that WD has engaged in a game of "benchmark specsmanship" with the AC313000. They've found a way to turn out fantastic figures in WinBench 98 and have set forth with a marketing campaign to emphasize this "strength." WB99, however, defeats this attempt and reveals that the AC313000 lags behind the only other ATA drive we've tested so far, the DiamondMax 4320. On a positive note, however, this escapade does show that WD cares about performance and is looking for ways to deliver more of it to the consumer. Though the method used with the AC313000 is probably not what most users are looking for, there's a bright star in the distance. The AC31300 is the last of WD's "3-platter/9.5ms" designs that have been hovering around for over a year and a half. A deal struck with IBM last year finally seems to be reaching fruition, with substantial benefits to be reaped. WD has already announced 4 platter drive designs which will finally allow the company to push the envelope in regards to capacity. Further, 8.5 millisecond seek times and 7200rpm designs have also been announced. All from a company that's been rather stagnant over the last couple years. So, though it hasn't happened yet, we may soon see state-of-the-art designs from Western Digital that will truly give the competition a run for their money. Time will tell.