IBM DeskStar

As stated above, the IBM DeskStar is a family of ATAPI consumer hard drives. One particular line, the 75GXP, earned the appelation "Deathstar" because they broke down often. Apparently the deskstars affected are the DTLA, IC35, AVVA, or AVER models of the 75GXP line. They are the first of the IBM drives to use GMR (Giant magnetoresitive) head technology, a good overview of the parts of a hard drive, and a good description of how they work can be found at www.storagereview.com I'll point out here that IDE drive lines are where the cutting edge technology is tested. The inferiority shown in the capacities of SCSI drives is usually attributed to letting IDE drives work out the kinks before letting the new techniques loose on drives that go in computers that will go on to do more then surf the web, play games, or write emails. I'll also point out, that one of the excuses IBM has used is that desktop IDE drives are rated with a MTBF which gives the customer an idea of how long a drive will operate before it fails. In a similar vein, if you read the documentation behind an optical drive, you'll find that you are not buying a drive that will read optical disks until the end of time, you are buying one that will read maybe 300,000 times, and then there will be a 1:2 chance that it'll drop dead. One excuse IBM has used, is that desktop (IDE) drives (with the exception of "enterprise" IDE drives, like the Western Digital Raptor, or the WD RAID line made for the serious user on a budget) are not designed to be used every hour of every day. The mean time before failure figures given for desktop drives assume that you use your drive between three to seven hours a day. IBM contends that the failure rate of deathstars used in the 3-7 constraints are truthfully advertised, and that they are not libel if you run their drives out of spec.

As the storage review site will tell you, the data density is so high, that because of the heat expansion that occurs in the envelope of time between when a drive starts up, and when it reaches a relatively ambient temperature, you will simply find bit "B" where you put bit "A", and bit "A" is now where you put bit "C". If you go further into the details at the SR site, you'll find that modern drives use a technique called thermal Calibration to tackle this problem. If you aren't doing anything with your computer, and you hear that the drive is seeking a lot (not the "clicking" that people think of as a harbinger of drive failure, but the noises of drive seeking) that noise is probably the sound of your drive calibrating itself.

What's thermal calibration got to do with the 75GXP? The firmware on the 75GXP doesn't really work well to combat the expansion. Some people say there's a particular problem with writing sectors, the pheuron site says that one of the drive heads catches some platter "lubricant" (probably not lubricant but the magnetic material that the platters are covered with so that they can hold data), some say it's got to do with the drive wiping vital servo data by accident, and anandtech says it's because the drive losses track of where it puts things then (presumably in some write verify cycle or something) the drive seeks for something, but it fails to get a good read, and then it retries until it resets, and the frequent resetting of the drive head is thought to be the cause of the clicking.

IBM released a firmware update that moves the heads around when the drive is idling, which either solves the problem, or delays the demise. And for the record, I read in one place that it seemed that drives from Hungary were more likely to fail.