Mac OS X and a Beige G3

I was going to spend last Wednesday putting a bigger, faster hard drive in our beige G3/266 and tell everyone what a big difference it made on Thursday. But things didn't work out that way.

Instead, I spent much of Wednesday, Thursday, Friday, half of Sunday, and much of today trying to put a faster, higher capacity hard drive in our beige G3 with very little luck at all. And none of that luck has involved Mac OS X.

Murphy's Law tells us that if anything can go wrong, it will. Finagle's Law says the universe tends toward maximum entropy. In the face of that, things can even go wrong with Macs. And believe me, they did

The Hardware

The beige G3 was released nearly five years ago (November 1997) and is the oldest Mac that Apple officially supports under Mac OS X. Upgrading this hardware for decent X performance is an important issue since there are a lot of them out there and Apple is really pushing Mac users to migrate to OS X.

I had three hard drives I wanted to test in our beige G3/266: a 20 GB Seagate Barracuda, a 30 GB IBM Deskstar, and an 80 GB Western Digital with an 8 MB buffer. All are ATA66 or ATA100 drives, spin at 7200 rpm, and provide 5x to 20x the storage space of the nearly full 4 GB Quantum drive inside the G3.

The 8 GB Limit

The first problem to raise its head was the 8 GB bootable OS X partition limitation of certain Macs, including the beige G3, WallStreet PowerBooks, and tray loading iMacs. In short, if you use a drive larger than 8 GB in one of these models, you must partition it. Mac OS X must be installed on a partition within the first 8 GB of drive space, so the simple solution is to create an 8 GB (or slightly smaller) first partition and leave the rest of the drive for everything else.

The first time around, I forgot about this. My bad. So I repartitioned the drive using Disk Utility on OS X 10.1.5, dragging the size of the first partition to about 8 GB, and then manually typing in 8.0 to set the partition size. That didn't work, either. It seems Disk Utility doesn't care what you type in for the partition size; it only looks at the dragged setting.

Third try I created a 7.81 GB first partition, did the whole thing with Carbon Copy Cloner yet again - and it still wouldn't boot into Mac OS X. I could, however, boot into Mac OS 9.2.2 from any of these configurations.

The IDE Bus

I've been working with Macs since the late 1980s and have installed my fair share of hard drives - internal and external, SCSI and IDE. It's usually easy on the Mac, even when you have to make a change or two to get things working. So I tried putting the drive on the end of the second bus, the start of the second bus, and on the first bus. I set it to slave, master, and cable select.

Nothing made a bit of difference.

Well, the Acard Hcard had arrived, so maybe that would solve the problem. The Hcard is an Ultra66 PCI card for the Mac that has two IDE buses and claims to be OS X compatible. The card actually simulates a SCSI bus as far as your Mac is concerned, but it uses less costly IDE drives. I've worked with similar cards before, and this one has dropped below the US$60 mark if you shop around.

Put in the card, connect the drive, repartition once more just to play it safe, run Carbon Copy Cloner, boot the new drive from OS X.

Same thing - no bootable HFS partition.

I am getting so tired of that error message.

Upgrade Despair

It was so darned easy replacing the drive in my TiBook. And it's always been so darned easy to drop another drive or even drive-and-controller-card into a Mac running the classic Mac OS.

What has Apple done differently with Mac OS X to create such trials and tribulations?

There is a known issue with multiple devices on the IDE bus for Revision A beige G3s (ROM revision $77D.40F2), but this is a Rev. B that works just fine with two devices on the second bus. Besides, I can't get it to work when the drive is the only device on the bus.

Still, the fact remains that these drives boot flawlessly into OS 9, so it can't be a simple hardware issue. I'm at a loss. And at this point, OS X is even broken on the old Quantum drive.

Calling For Help

Running stuck, I asked for help on G-List, the email list we run for people using G3- and G4-based Macs. The best advice seemed to be reinstalling OS X, so that's what I've been trying to do.

Trying. And trying. And trying.

And it isn't working. I've tried installing 10.0 on the existing partition with all the data intact and after completely wiping the partition. The installer loads, runs, and refuses to complete the job. And I don't own a full copy of 10.1, let alone Jaguar, so this is the only way I can reinstall OS X. And I keep getting an unspecified error during installation.

Funny thing is, it worked once upon a time - otherwise I never would have been able to have OS X on this machine in the first place. Something is very, very wrong here.

Sometimes the old beige G3 won't even see the internal CD-ROM drive, so I've tried using external SCSI drives. No joy there either.

I've tried using Disk First Aid, Norton Utilities 7, TechTool Deluxe, and Disk Warrior. Nothing helps. And after all of this, even the OS 9 install has broken.

I miss the good old days of being able to simply drag a System Folder from one drive to another and having it work.

XPostFacto

My last resort is XPostFacto, formerly known as UnsupportedUtilityX. Ryan Rempel's installer not only lets you put OS X on officially unsupported software, but it's supposed to solve some install problems on some supported Macs as well. "Ordinarily, Mac OS X should just work with the Beige G3, but there are some cases in which the Beige G3 has problems that XPostFacto may be able to help with."

By this time, everything had degenerated to the point where I couldn't reliably boot from the hard drive or CD-ROM. The monitor was staying black almost all the time. Time for drastic measures. And even though I'd downloaded XPostFacto, I never did have to use it.

It Finally Works

I shut down and pulled all three sticks of RAM, removed the two PCI cards, took out the battery, reset the cuda switch, put the battery back in, dropped in a single 128 MB DIMM, hit the cuda again, and tried to boot. Still had problems.

Ran out to Meijer, bought a CD-ROM drive cleaner, cleaned the CD-ROM twice (we'd never done it before), and tried to boot. Still had a black screen.

Checked the back of the computer. The monitor, an Apple Multiple Scan 17, was connected - but not quite tight. I tightened it up, and voilà, everything worked!

I managed to boot into OS 9, set the date & time, connect to our home ethernet network, and mount the OS X 10.0 install CD. Run the installer, reboot to launch it from the CD (the 24x Apple drive is loud!), and let it do its thing. It took long enough to enjoy lunch, but it got Mac OS X 10.0 installed.

Next step: Shut down, add the other two RAM modules, drop in the USB card, reboot, and run the 10.1 update CD. It's working perfectly. That done, time to install the security update and the updates to bring it to 10.1.5.

And then I can leave the darn thing alone for a while. I don't want to try another hard drive in it. I don't want to play with CPU upgrades and overclocking. I don't want to try the FireWire card or the Acard Hcard Ultra66 drive controller card. I just want to have it up and running so Tim can use it for a few day.

Maybe then I'll feel like tackling it again.

Closing Thoughts

Of course, none of this explains why Carbon Copy Cloner didn't work. After all, the looseness of the monitor plug or a dirty lens on the CD-ROM drive shouldn't have the least thing to do with the process of moving OS X to another drive.

Nor does any of this explain why sometimes the Mac wouldn't even see that the CD-ROM drive was present. A dirty lens might prevent it from mounting a CD, but the computer should still recognize that the drive is attached.

In the end, I guess we have to chalk it up to corrupt preferences - today's equivalent of blaming SCSI voodoo. Somewhere between zapping the PRAM (dozens of times), removing the battery, and resetting the cuda, whatever had been causing the problem seems to have disappeared.

Yes, it took extreme measures, and I can't imagine why things weren't a whole lot easier, but I'm just glad it's done. We'll get to testing and other updates later. Until then, it's running nicely again.