I was actually doing computer repair when the IBM PC came out. It had a backplane, but most of the system was on the motherboard. Repair people were horrified that many if not most of the logic chips were soldered in, not in sockets like the Apple II. At the time, I had repaired Apple II's, and dumb terminals that had soldered IC's. Half of my time was spent painstakingly de-soldering chips from boards so a replacement chip could (hopefully) solve the problem.
On the other hand, most Apple II repairs consisted of what we jokingly called "the laying on of hands," which meant reaching in and pushing all of the chips back into the sockets with the palms of your hands. That was always accompanied by rubbing the gold fingers on the disk controller card with a pencil eraser because the gold plating was inferior. Many Apple II repairs were 5 minutes. Third on the list was power supplies, which we never figured out how to repair at a component level, and were justifiably afraid of.
At the time, IBM published some data based on their own years of computer service. They serviced a lot of their own boards, and made their own memory chips. Their data showed that chips were actually more reliable than sockets, so they limited sockets to those which had potential for fallout in production (the 8088), and the memory chips which were an optional add-on. And the ROMs, which could potentially be upgraded.
On the other hand, most Apple II repairs consisted of what we jokingly called "the laying on of hands," which meant reaching in and pushing all of the chips back into the sockets with the palms of your hands. That was always accompanied by rubbing the gold fingers on the disk controller card with a pencil eraser because the gold plating was inferior. Many Apple II repairs were 5 minutes. Third on the list was power supplies, which we never figured out how to repair at a component level, and were justifiably afraid of.
At the time, IBM published some data based on their own years of computer service. They serviced a lot of their own boards, and made their own memory chips. Their data showed that chips were actually more reliable than sockets, so they limited sockets to those which had potential for fallout in production (the 8088), and the memory chips which were an optional add-on. And the ROMs, which could potentially be upgraded.