Home / Opinions / Hard Disk Roundup / How We Tested

How We Tested

In an ideal world the performance of your hard drive would have little effect on your PC as you would have stacks of memory and all the operations would run in solid state. Windows would read files in the background and have them ready for the moment when you need them, and all would be well.

Unfortunately, in the real world you access your hard drive fairly randomly as you open and close applications, and Windows kicks in with autosave and System Restore. None of that is a problem if you have a fast PC, a decent processor and plenty of memory, but as soon as you start to transfer a large amount of data the speed of your hard drive becomes apparent. An older drive will thrash away, sounding like a CD-ROM and although a newer hard drive will be faster, the mechanical nature of the drive means that it is many orders of magnitude slower than operations that occur in memory.

With the previous generation of Parallel ATA hard drives there was a theoretical performance difference between drives with ATA66, ATA100 and ATA133 interfaces, but it was very unusual to be able to detect these differences in actual use.

The first generation of Serial ATA drives all have a nominal ATA150 interface, but the drives themselves are very similar to conventional Parallel ATA drives.

We tested these drives using an AOpen AX4C Max motherboard with 875P chipset and ICH5R Southbridge. The AOpen also has a Promise PDC20378 controller chip so there's plenty of scope for us to connect up test drives. We used a 3.2GHz Pentium 4 Northwood C processor on an 800MHz FSB with 1GB of Crucial PC3200 memory in four modules to take full advantage of dual channel memory, so the PC itself was no bottleneck to data transfer. In addition we used a 120GB WD drive as our Windows C drive and a pair of WD Raptor's in a RAID 0 array connected to the Promise controller, leaving the native Intel controller free for the hard drive we were testing.

We ran HD Tach from TCD labs ( www.tcdlabs.com ) along with our own 'real world' file transfer test where we created a folder containing 4.32GB of small files on the Raptor array which we first copied to the test drive and then wrote back to the Raptors.

In each case CPU usage was about 10% (fluctuating between 6% and 20%) We also noted a strange occurrence to do with the capacity of the drives. When we added a test drive to the PC we ran HD Tach on the unformatted drive, and then we used Disk Management in Windows XP to format the drive. The 200GB Maxtor, 160GB Seagate and 250GB WD were all detected as being much smaller than their true size. The WD, for instance, was reported as 128GB in size. We then ran the Windows set-up CD as though we were installing Windows on the new drive and all the 'missing' capacity was detected correctly, with the WD appearing as 234GB of formatted capacity. We hate mysteries and this one bears further examination. As soon as we’ve figured it out this disparity you’ll be the first to know.

comments powered by Disqus