The drive

The whole 15K.3 range is a low profile 1" form and this one is no different. The drive itself is a 68 pin variant, the ST373453LW with 73GB (drive manufacturer GB) of space. Nothing out of ordinary for most workstation users there (well, except the size). It is also a U320 SCSI drive. Unfortunately, I don't have access to a U320 controller at the moment so tests of the drive were done using the onboard Adaptec (AIC-7899W) U160 channels of a Tyan Thunder K7X. There should be negligible difference between U160 mode and U320 mode but that difference will need to be looked into at another time. U320 does have a few nifty features such as CRC checking upon commands rather than just data as well as a packet protocol to allow for multiple transfers during a single drive - host connection. Adaptec has a fairly good summary of all the new features right here. With most of the new motherboards coming out with U320 onboard and U320 controllers available in volume then you will have the chance to take full advantage of whatever benefits U320 brings to the table. I'm not going to bore you with pictures of the drive or more specifications. It's a hard drive. It looks like a hard drive. You've all seen a hard drive. Let's see what it does.

Testing

Since a disclosure of the test system is called for now would be an opportune moment to introduce 'Sidious' (blame Hooz, he named it). Sidious is a bit of a beast. Tyan Thunder K7X, twin MP2100+, 512MB of Crucial Registered ECC and then there's the reason for its existence and the real reason for the name: the storage subsystem. 10 x 36GB X15 36LP's, 1 x 73GB 15K.3, Mylex eXtremeRAID 2000 quad channel RAID controller, IBM LTO Ultrium 100GB/200GB tape drive, Adaptec 2000S ZCR, twin onboard Adaptec U160, twin Intel SC5100 hotswap backplanes and it's growing....... Hopefully I can find some time in the not too distant future to run some more tests on the the rest of it. However, as Hooz has already alluded to in news posts, it's not without issues. PCI throughput is the principal one. At this point, the base motherboard isn't one I would recommend if you have more extreme PCI throughput requirements. Sidious currently runs Windows 2000 SP3. Motherboard BIOS is 4.03b, ZCR firmware 3.80b, 36LP firmware 0004, 15K.3 firmware 0002. All drive testing was using unformatted drives so that any effects of Windows' known reluctance to produce decent SCSI performance were minimised. I did try Windows 2000 SP2 for a few runs and the results were identical to when under SP3.

For the series of tests I placed the 15K.3 alongside the previous flagship, the X15 36LP. The X15 36LP comes in a maximum capacity of only 36GB so I plucked one from the gaggle and set it to work. Having the two side by side made things a little easer to see and feel the differences that aren't readily ascertainable from manufacturer specifications. The testing software of choice is Intel's IOMeter 99.10.20. Tests were run 4 times and the run time per step was two minutes with a 20 second ramp. Doesn't sound like much but it adds up to a lot of testing hours. Read: days and days and days spread out over weeks and weeks. Anyway, this software is freely available on the wider 'net and a rather powerful tool for analysing I/O. In addition, the folks at SCSI Toolbox also provided me with their somewhat daunting suite of tools, SCSI Toolbox 32. A time limited demo of the product is available to the public upon request and it's worth checking out as I intend to use the full toolbox to a greater extent in the future. The harsh reality is that it's not going to be a product that you purchase unless it's for your business. Well, unless you have $2495 to spare for the complete suite. It's got some pretty amazing tools inside but I don't think its drive throughput figures are entirely useful for this test so I'll stick with IOMeter for this review. To give you a small hint of what's in store for the future with SCSI Toolbox 32 here are a few screen dumps: