Test Setup



There's not too much to say about the test platform except it's fast. We are utilizing Vista 64-bit SP1 as our OS of choice now. We are still in the process of comparing our NVIDIA-based setup to the Intel X48 for future drive articles. However, performance differences between the two platforms are within 1% of each other with single drives. Native Command Queuing is enabled for all drives and the drive is formatted before each test run along with the prefetch folder being cleared. We run each benchmark five times and throw out the high and low test scores, then average the remainder for our results.

Quick Tests

We are providing PCMark Vantage results today along with initial acoustic and thermal results. The details about the PCMark Vantage HD suite tests and how results are determined can be located here. Our acoustic tests measure the decibel levels while the system is at idle and under load while running the Hard Disk test suite within PCMark Vantage. We take measurements at a distance of 5mm from the rear and front of the drive in a separate enclosure. The test room has a base acoustical level of 20dB(A).

Our thermal tests utilize sensor readings via the S.M.A.R.T. (Self-Monitoring, Analysis and Reporting Technology) capability of the drives as reported by utilizing the Active SMART 2.6 utility. We also utilize thermal sensors and infrared measurement devices to verify our utility results. We test our drives in an enclosed case environment. Our base temperature level in the room at the time of testing is 24C.



The WD 640GB drive generates an excellent score in the PCMark Vantage tests that simulate real-world performance patterns utilizing a variety of actual applications. The Samsung F1 750GB drive sports a 32MB buffer while the WD 640GB "gets by" with a 16MB buffer that has been highly tuned for random seeks and small sequential block sizes. The WD 640GB drive leads in all PCMark Vantage categories except the Windows Defender and Media Center applications where the Samsung exhibits exceptional results. These scores reflect our own off-line application benchmark testing, where we saw similar results because the 32MB buffer greatly assists in the read performance of large data block sizes accessed in sequential order.

The one area that really disappointed us with the WD 320GB drive was its pitiful random access times at 16.4ms. After a lengthy conversation with Western Digital, we now understand why the drive performed so poorly in this test. This also explains the good but not great application results in general. Granted, the drive still performs well, but the initial specifications lead us to believe that performance should have been greatly improved over previous generation drives. Western Digital explained the single platter 320GB drive is aimed at the entry-level market where thermals and acoustics are critical for mass acceptance of the drive by the OEM and retail customers.

WD sacrificed a small amount of performance on the 320GB drive to meet these goals. This was by design and is not an indication of the performance potential of their new technology. In fact, these same 320GB platters will be used in single-sided form eventually for the 160GB drives. According to statistics, the 160GB drives have quickly replaced the 80GB products to become the new "sweet spot" in the general market. As the capacities increase, the typical user for these products normally expects a balance between performance, thermals, and acoustics. With that in mind, WD tuned the 640GB drive for additional performance at the expense of acoustics.

As far as acoustic testing, this drive posted excellent results, although thermals were higher than we expected considering the Samsung drive is carrying an additional platter and set of heads. Even during heavy seeks, noise levels remained muted, and at idle the drive's acoustical footprint was almost silent. The numbers suggest that the 640GB is very close to the 320GB drive in acoustics. In most ways it is, but we could definitely hear a difference in the seek operations between the two drives. The Samsung F1 offered similar acoustics to the WD SE16 640GB drive but during seeks the drive had a slightly heavier tone that showed up in the recordings.

In our first application tests (additional results are coming soon), the drive has performed slightly better than the Samsung F1 in areas such as program loading, gaming, and digital image manipulation. The WD drive is slightly slower in our Nero Recode 2 benchmark and at streaming multi-gigabyte files. This is due to the differences in buffer sizes at this point as far as we can determine. However, the differences between both drives are minor and are something that the majority of users would never notice. This drive excels in providing excellent acoustics and performance while thermal management is acceptable for this capacity range.

Just how well this drive compares to the latest competition in the 750GB range is something we will answer in the near future. Even at this point in testing, the WD Caviar SE16 640GB drive already has our tentative recommendation for users looking for a larger capacity drive for a gaming or audio/video/digital image workstation. The drive is very good, but we need to pound on it for the next several days under varying conditions along with additional analysis before we can give it our full recommendation.

Index
Comments Locked

40 Comments

View All Comments

  • johnsonx - Sunday, March 23, 2008 - link

    Same experience here as the other two replies: WD has rarely failed me, while Seagate and Hitachi have been bad for me. I used to use a lot of Maxtors, but at some point alot of them started failing in customer systems, so I switched to WD.
  • just4U - Friday, March 21, 2008 - link

    I swear by my western digital drives. They have always been good to me. I would feel safer tho if more of them cam with the 5 year warranty.
  • Simon128D - Friday, March 21, 2008 - link

    That's odd, I guess you've just had bad luck.

    I've had the complete opposite experience. All my WD drives never gave problems but Seagate never seem to last more than a year or give some kind of reading problem down the line.

    I've since bought nothing but WD drives and have never regretted it.
  • Xpl1c1t - Friday, March 21, 2008 - link

    Beautifully written article, funny how marketing can destroy the beauty of engineering.

    ...i just wonder how dense a platter could be today if storage needs of end-markets didn't drive the funding of development.
  • Arbie - Friday, March 21, 2008 - link

    There wouldn't BE any platter development without the end markets. It's the sales and marketing that pay the bills back at the lab. Just something to remember before we bash the PR guys - which, as an engineer, I am naturally inclined to do.
  • mindless1 - Saturday, March 22, 2008 - link

    Naw, just about everyone ignores the marketing nonsense except for being offended by what they try to spew to sell the product - this not necessarily pertaining to HDD in particular.
  • Inkjammer - Friday, March 21, 2008 - link

    This seems like a great middle-ground hard drive for an HTPC. Seems like a great amount of storage at a reasonable price. Just gotta wait for PC Cablecards to come out now.
  • silentbob17 - Friday, March 21, 2008 - link

    When do we see a new raptor from WD? Im still inlove with my old raptor from 2006....am I a n00b and not seeing that these new HDD's are better?
  • Casper42 - Friday, March 21, 2008 - link

    They are almost as fast.

    For instance the WD7500AAKS (750GB 7200rpm) is pretty much just as fast as the Raptor 150 in everything except seek time.

    Considering the Raptor 150 costs more than the 750 AAKS, I personally will not be buying Raptors for my next machine. (Note I run 2 74GB Raptors right now in RAID 0, so I am not just a Raptor basher :P )

    The sweet spot from what I can tell is to use 3, 4 or 5 Fast Large drives like the 750AAKS, this new 640 or a Samsung F1 and do a RAID5.
    On an Intel chipset you can go with 3, 4 or 5 and get good performance. However beware that for some reason on nForce chipsets there is a weird performance hit with Raid5 that is well documented in a few places on the net. Someone figured out that you can get around it by using 3 or 5 drive RAID5 arrays and sthen setting block size to a certain number. I think you also have to use a Vista DVD to create your partitions as it does something special with the partition layout that is part of this speed boost. After the partitions are created however you can go back and install XP and just say "Use Existing Drive" or whatever.

    So my original plan was to do 5 x 750 AAKS in R5 which would give me around 2800MB Useable which would be cut into either 800+2000 or 500+300+2000 (Dual Boot).
  • iwodo - Friday, March 21, 2008 - link

    I think the 139 is too expensive. I hope the 640Gb will replace 500Gb version soon.

Log in

Don't have an account? Sign up now