I've looked into it and read several IBM documents regarding BGA devices.
Essentially, it IS possible to do with a heat gun and a NEW PPC 740, but don't try to do it with a pulled 740 unless you've had it re-balled with all new solder balls.
Ideally, one would have a new chip and a specialized workstation that uses very hot gas, the temperature of which is precisely controlled. A precision tool would be used to remove the old BGA device and clean the pads on the board, including removing any balls that remained stuck to the board. Then, the new chip would be mounted precisely and melted into place.
Reworked BGA devices use the same setup, but the device to be reworked is pulled and the balls are removed. Then, the chip is placed in what looks like a garlic press, and new balls are installed thru the holes in the bottom and adhere to the chip. Then, the chip with its new solder balls is replaced onto the board and melted down again.
Some of the IBM documents showed that on each subsequent removal of a BGA device from a board, it gets messier every time. Typically, more and more balls stick to the board as opposed to remaining on the BGA device, and the pads can become worn and reluctant to hold solder.
It's a neat idea, and I still wanted to try, only more with replacing the 250 or 292MHz 750s on Wallstreet daughter cards with 750L chips in the 400-500MHz range. I really don't think any upgrades from Newer or anybody utilized the 83MHz system bus. I also thought about replacing the BGA 603e in a PowerBook 3400 with a faster one, but it's hard to find a 603e over 300MHz, so it's a bit pointless.
Anyway, the OS X issues on the Gazelle models (and Tanzania machines like the 4400 or Starmaxes) are related more to the other system hardware rather than the 603. I mean, a 3400 and a 2400 can run OS X to 10.2 without much issue (though slowly), and they're both 603e machines.
Essentially, it IS possible to do with a heat gun and a NEW PPC 740, but don't try to do it with a pulled 740 unless you've had it re-balled with all new solder balls.
Ideally, one would have a new chip and a specialized workstation that uses very hot gas, the temperature of which is precisely controlled. A precision tool would be used to remove the old BGA device and clean the pads on the board, including removing any balls that remained stuck to the board. Then, the new chip would be mounted precisely and melted into place.
Reworked BGA devices use the same setup, but the device to be reworked is pulled and the balls are removed. Then, the chip is placed in what looks like a garlic press, and new balls are installed thru the holes in the bottom and adhere to the chip. Then, the chip with its new solder balls is replaced onto the board and melted down again.
Some of the IBM documents showed that on each subsequent removal of a BGA device from a board, it gets messier every time. Typically, more and more balls stick to the board as opposed to remaining on the BGA device, and the pads can become worn and reluctant to hold solder.
It's a neat idea, and I still wanted to try, only more with replacing the 250 or 292MHz 750s on Wallstreet daughter cards with 750L chips in the 400-500MHz range. I really don't think any upgrades from Newer or anybody utilized the 83MHz system bus. I also thought about replacing the BGA 603e in a PowerBook 3400 with a faster one, but it's hard to find a 603e over 300MHz, so it's a bit pointless.
Anyway, the OS X issues on the Gazelle models (and Tanzania machines like the 4400 or Starmaxes) are related more to the other system hardware rather than the 603. I mean, a 3400 and a 2400 can run OS X to 10.2 without much issue (though slowly), and they're both 603e machines.