IBM Deskstar 75GXP Click of Death: Real Fix or Replace?

What You're Actually Dealing With

The IBM Deskstar 75GXP earned the nickname "Deathstar" for a reason. Between roughly 2000 and 2003, IBM shipped millions of these drives and a huge chunk of them failed — some straight out of the box, some after a few months. The click of death on a 75GXP is not a cable problem, not a jumper problem, and not something a firmware flash will fix. It's a mechanical failure inside the sealed drive assembly, usually the GMR (giant magnetoresistive) read/write head slamming against the platter because the head stack assembly lost its calibration.

If your drive is clicking in a rhythmic pattern — click-click-click, pause, click-click-click — and the BIOS doesn't detect it or detects it as 0MB, you're hearing the head trying to load, failing to find the servo tracks, and retracting. It will do this until the motor burns out or the head scratches the platter into garbage. Every power-on cycle makes it worse.

Before you touch anything: if the data on this drive matters, stop powering it on right now. Skip to the advanced section. If the data doesn't matter and you just want the drive working, you're going to be disappointed — but read on anyway.

The 30-Second Fix: Rule Out the Stupid Stuff

I've seen plenty of "clicking drives" that turned out to be a dying power supply or a bad Molex connector. It's rare with the 75GXP specifically, but check it before you write the drive off.

1. Power down the machine completely. Pull the power cord.
2. Reseat the 40-pin IDE ribbon cable on both ends. The 75GXP uses a standard ATA/100 interface, and those ribbon cables fail constantly — the connectors crack, pins bend, and the strain relief pulls loose.
3. Swap to a different Molex power connector. If your PSU has a splitter, go direct.
4. Set the jumper correctly. If it's the only drive on the channel, Master or Cable Select. If it's sharing with another drive, Master/Slave. Wrong jumper setting can make a healthy drive click as it hunts for an address.
5. Boot and listen. If the click is gone and the BIOS sees the drive at its full capacity, you got lucky. Back up your data tonight anyway.

If it's still clicking, the problem is inside the drive. Move on.

The 5-Minute Fix: Confirm It's Mechanical

You want to hear the drive with the case open and the system running. Do this carefully — don't touch the PCB while it's powered.

1. Pull the side panel and power up. Put your ear near the drive (not touching it).
2. Listen for the spin-up. A healthy Deskstar spins up to 7200 RPM in about 3 seconds with a smooth whine. A dying one spins up, clicks, spins down, tries again. That cycle is the firmware's retry loop.
3. Feel the drive. If it's vibrating hard or getting hot within 60 seconds, the spindle bearing is gone.
4. Check the BIOS. If the drive shows up as 0MB, 8GB, or with garbage characters in the model field, the drive's own controller can't read the servo data it needs to report its identity.

Any two of those symptoms together means the drive is toast. Don't bother with a low-level format, don't run chkdsk, don't try SpinRite. None of those tools can fix a physical head crash, and running them makes it worse by forcing the drive to keep spinning.

One Thing That Occasionally Works

There's a trick some techs use on 75GXPs and other Deskstars from this era: the "freezer trick." Seal the drive in a Ziploc bag, put it in the freezer for 4-6 hours, then pull it out and immediately connect it. The cold shrinks the metal slightly and can temporarily restore head alignment. It's a last-ditch data recovery move, not a repair. It works maybe 1 in 5 times, and if it works, you get 10-30 minutes of drive access. Use that window to copy your critical files off. Don't waste it watching a full disk image run.

I've personally pulled a 60GB 75GXP back from the dead long enough to grab 40GB of client data using this method. The drive died permanently on the next boot. That's the normal outcome.

The 15+ Minute Fix: Data Recovery and Replacement

If the data matters, stop reading the click-of-death troubleshooting and start treating this as a data recovery job.

Option A: DIY Recovery (if the drive is still readable at all)

1. Get a known-good machine with an IDE port, or a USB-to-IDE adapter — the StarTech USB2IDE is the one I still keep in my kit.
2. Boot into Linux (a live USB works fine). Linux is far more forgiving with failing drives than Windows. It'll retry reads instead of just giving up.
3. Use ddrescue, not dd. ddrescue logs which sectors failed and lets you resume. It's the right tool.

sudo apt install gddrescue
ddrescue -n /dev/sdb /mnt/backup/deskstar.img /mnt/backup/deskstar.log

Run the first pass with -n (no scraping). It skips bad areas fast and gets the healthy sectors first. Then run a second pass without -n to try and pull the marginal sectors back. If the drive is clicking during this, power down between passes and let it cool.

Option B: Professional Recovery

For a 75GXP specifically, professional recovery runs $500-$2000 depending on how bad the platters are. The lab has to open the drive in a cleanroom, swap the head stack assembly from a donor drive, and read the platters. Donor 75GXPs are getting rare and expensive because everyone's drives already died. Get a quote before committing — sometimes you'll find the cost is more than the data is worth to you.

Option C: Just Replace It

If you don't need the data, do yourself a favor and put this drive in the recycling bin. Do not try to use it as a secondary drive, do not use it for a Windows page file, and for the love of everything do not put it in anything you care about. A 75GXP that's clicked once will click again. The average lifespan after first click is measured in hours, not weeks.

Replace it with literally anything. A cheap SATA SSD plus a $10 IDE-to-SATA adapter if you're keeping an old machine alive. You'll get 10x the performance and none of the heartburn.

Why the 75GXP Failed So Badly

IBM pushed the 75GXP to 15GB per platter at a time when everyone else was at 10GB. Higher density means the head flies closer to the platter, and the tolerances get tighter. IBM's glass platters (yes, glass) also turned out to be prone to shattering under heat stress. Combined with a firmware bug in the servo tracking that let the head drift out of alignment, the drives ate themselves.

IBM paid out a class-action settlement over these drives and eventually sold its entire storage division to Hitachi in 2003. The 75GXP is the reason "Deathstar" is still a word in the IT lexicon 20+ years later.

Quick Decision Tree

Bottom line: the click of death on a 75GXP isn't a fixable problem, it's a countdown. Move fast, prioritize your data, and don't throw good hours after a bad drive.