VirtualLarry
No Lifer
- Aug 25, 2001
- 56,554
- 10,171
- 126
Originally posted by: imtim83
The IBM Deathstar died like crazy one after another. I was at storagereview and anandtech at the time and saw day after day more posts of people's IBM Deathstars dying one after enough. It was amazing and horrible at the same time. The funny thing is I was just about to buy one and if I would of not found storagereviews forum I may have. Though once I saw that they were dying in storagereviews forum first I wasn't sure anymore. I then started noticing posts on anandtech forums of the IBM Deathstars dying as well. It was a mess. Almost all you heard about basically.
Me too. The interesting thing is, I owned a 30GB 75GXP model, and it wasn't one that was made in the "bad" factory, and it had firmware that pre-dated the "bad" firmware too. It ran well - over 2.5 years before I sold it to someone. It did develop a couple of bad sectors though, but that was because I let it get too hot a couple of time while being used in one of those plastic removable mobile racks.
The sad thing is, I finally decided that I should update the firmware, because of the well-known issue of "buggy firmware", both in terms of leaving the "write gate enable" during power-off (which will wipe a nice clean spiral path right across the platters, low-level embedded-servo data and all), which was often a problem with W2K at the time, because of MS's stupidity in not properly flushing the IDE HDs cache before executing an ATX soft-off shutoff. (Where did MS expect the unwritten data to go to, anyways? Binary heaven?) Irregardless though, doing that shouldn't wipe servo data, that was a HD firmware bug. Another was that the drive's internal SMART-related "offline scan" mode, could corrupt/overwrite user data sectors accidentally. Since I couldn't verify that my firmware was immune to that problem, I finally chose to upgrade. The upgrade more-or-less "took", but it wiped the entire SMART history (showing zero POH! I should have sold it then... :evil, and soon after, appeared to have a bad sector develop somewhere in the hidden sectors area where it keeps track of .. which sectors are bad! Yeah, just the place to develop a bad sector, right? One time, it finally did a "click-skritch (repeat)" thing when booting up, and that's when I realized that its time had come. I managed to get it to boot up correctly, by installing it upside-down. (My friend suggested that to me, I didn't believe him at the time that it would do anything substantial.) Apparently that was enough to properly read the platters again, so I backed-up everything onto the replacement drive that I had ready, and found that repeated passes of HDTach's write-testing, while upside-down, seemed to make the "spikes" in HDTach's read-testing go away. (Apparently it finally triggered some re-maps.) It still made re-read noises during DFT while "checking internal error logs", which indicated that it still had a marginal sector in the hidden sector area. I discovered that a later version of DFT had a new feature, "security erase prepare", which instead of being a host-controlled "write zeros" operation, it triggered an erase routine built into the drive's firmware itself, which somehow apparently fixed/overwrite/remapped the marginal sectors in the hidden sectors region as well. (Good tip to know!) At that point, repeated passes of HDTach read testing didn't show any noticable spikes, so then I sold the drive, before anything else wierd started to happen. (It was still in warranty for another six months, so I didn't feel too badly about it. The 75GXP issue was more-or-less well-known at the time by then too.)
But to it's credit, those drives were fast, and quiet besides. They really were engineering marvels, even if the firmware was buggy as heck, and many of their were produced in factories with sub-standard procedures and quality-control issues. (The infamous "Hungary" factory. Mine was made in Romania, I think.) Glass platters, ceramic ball-bearings, ramp-load heads for superior non-operating shock protection, advanced firmware commands for A/V streaming and IDE tagged command queuing, optimized head-stepping algorithms, a 2MB cache, man, that thing had it all. Such a shame, really. If they didn't have problems, I'd buy another few and stripe them together for an insanely-fast disk rig. It was like an enterprise-class SCSI HD, trapped in an IDE drive's chassis.