3 ms·

They've pulled off major hardware transitions twice before without hiccups. But this time they make the whole platform which is quite impressive imo.

scarface74 5y ago
The 68K to PPC transition was pretty good. But the operating system was running emulated 68K code for five years.
- astrange 5y ago
  That was more or less intentional. The 68K interpreter was quite fast and 68K asm is smaller in memory than PPC asm, so converting all the cold code over would've made the system slower. I'd like to see the same idea used today. One I can think of is C++ exception unwinding which uses a DWARF interpreter instead of generating tons of code to do it.
  - scarface74 5y ago
    You’re looking through rose colored glasses. My PowerMac 6100/60 ran 68K apps slower than my LCII with a 68030-40Mhz accelerator. The original 1st gen 68K emulator was really bad. Connectix became famous for shipping SpeedDoubler, a much better emulator. But my 6100/60 could still barely keep up with the 68030-40Mhz. The emulator performed even worse on the PPC 603.
    - astrange 5y ago
      That could be, I think it was explained to me by an former engineer on Mac OS 8 or so. But there was still 68K asm in there up until Carbon.
adolph 5y ago
Gosh I had to think back a moment to remember 68k to PPC. I wonder if that transition could be considered “botched” in that it happened at instead of going directly to x86. Outside hindsight I recall it was considered a questionable choice at the time.
- GekkePrutser 5y ago
  It wasn't really though. Motorola was going nowhere. As to why not x86, that's another story but Intel has gone down the wrong path several times. Like with the Pentium 4.
- zozbot234 5y ago
  Not really, PPC was a reasonable choice for a high-performance architecture back then, and arguably a better fit for former 68k coders than x86. And a move was necessary because the 68k was becoming a dead platform by then.
- doctor_eval 5y ago
  My recollection is that the PPC at launch was much faster than x86. Jobs talked about the road map a bit, and there was a lot of press about it too, but the road map didn’t pan out, and their partners dropped the ball. And many other companies made the transition to x86 (Data General was one I worked with) and subsequently died.
  - atdrummond 5y ago
    The biggest problem was Motorola/Freescale saw that their future was in embedded, not the (relative) high performance required for personal computing. So the chip provider for Apple's mobile lineup no longer was producing performant chips. Unfortunately, IBM's implementation of PPC led a dual life as a server/workstation and desktop chip meant that getting it's power consumption, heat profile and performance optimized for a mobile device was an extremely difficult proposition. It would be interesting to see where we'd be if IBM had ensured Apple they could deliver a G5 laptop and had done so at a price and spec competitive with, or superior to, Intel.
    - GekkePrutser 5y ago
      > The biggest problem was Motorola/Freescale saw that their future was in embedded, not the (relative) high performance required for personal computing. Ironically that's also the niche where PowerPC ended up when Apple dumped them :)
ghaff 5y ago
The striking thing about this time is that it is essentially transparent to most people. Probably more third party programs have been broken by security changes that Apple has made over the past five years than have been broken by the M1 transition. Yes, there are performance implications but M1 is sufficiently fast and most important performance sensitive programs are being quickly ported that it doesn't matter that much. (And, for most people, ultimate performance is mostly not a big deal on laptops these days.) Transitive's tech, combined with processors that can afford some inefficiency, is pretty much magic to anyone who remembers what ISA transitions used to look like. (As a hardware product manager, I lived through a couple of them including a Motorola 88K to x86 one on Unix.)
- quitit 5y ago
  I feel like more software died during the 32->64 bit transition than any other time in recent history. Lifting the rug is also par for the course for Apple, even artificially so with their latest crackdown on apps that don’t run well.
  - astrange 5y ago
    That mostly seemed to be video games (at least their developers complained the most), but I just had to wonder why they couldn't make their games 64-bit-ready with 15 years of warning. Btw, 32-bit Windows binaries still run through WINE, just not 32-bit Mac binaries.
    - stormbrew 5y ago
      Games just don’t have a long enough tail to make this kind of effort worth it. They aren’t like business software where you try to keep people on the hook for years. There’s little reason to keep updating a game past its first few post release patches. So you can assume most of their code bases have been untouched for years, were probably written by people who are no longer around, and haven’t been compilable by modern toolchains for a long time. The tech debt is just insurmountable most of the time.
    - jedberg 5y ago

scarface74 5y ago

The 68K to PPC transition was pretty good. But the operating system was running emulated 68K code for five years.

astrange 5y ago
That was more or less intentional. The 68K interpreter was quite fast and 68K asm is smaller in memory than PPC asm, so converting all the cold code over would've made the system slower. I'd like to see the same idea used today. One I can think of is C++ exception unwinding which uses a DWARF interpreter instead of generating tons of code to do it.
- scarface74 5y ago
  You’re looking through rose colored glasses. My PowerMac 6100/60 ran 68K apps slower than my LCII with a 68030-40Mhz accelerator. The original 1st gen 68K emulator was really bad. Connectix became famous for shipping SpeedDoubler, a much better emulator. But my 6100/60 could still barely keep up with the 68030-40Mhz. The emulator performed even worse on the PPC 603.
  - astrange 5y ago
    That could be, I think it was explained to me by an former engineer on Mac OS 8 or so. But there was still 68K asm in there up until Carbon.

adolph 5y ago

Gosh I had to think back a moment to remember 68k to PPC. I wonder if that transition could be considered “botched” in that it happened at instead of going directly to x86. Outside hindsight I recall it was considered a questionable choice at the time.

GekkePrutser 5y ago
It wasn't really though. Motorola was going nowhere. As to why not x86, that's another story but Intel has gone down the wrong path several times. Like with the Pentium 4.
zozbot234 5y ago
Not really, PPC was a reasonable choice for a high-performance architecture back then, and arguably a better fit for former 68k coders than x86. And a move was necessary because the 68k was becoming a dead platform by then.
doctor_eval 5y ago
My recollection is that the PPC at launch was much faster than x86. Jobs talked about the road map a bit, and there was a lot of press about it too, but the road map didn’t pan out, and their partners dropped the ball. And many other companies made the transition to x86 (Data General was one I worked with) and subsequently died.
- atdrummond 5y ago
  The biggest problem was Motorola/Freescale saw that their future was in embedded, not the (relative) high performance required for personal computing. So the chip provider for Apple's mobile lineup no longer was producing performant chips. Unfortunately, IBM's implementation of PPC led a dual life as a server/workstation and desktop chip meant that getting it's power consumption, heat profile and performance optimized for a mobile device was an extremely difficult proposition. It would be interesting to see where we'd be if IBM had ensured Apple they could deliver a G5 laptop and had done so at a price and spec competitive with, or superior to, Intel.
  - GekkePrutser 5y ago
    > The biggest problem was Motorola/Freescale saw that their future was in embedded, not the (relative) high performance required for personal computing. Ironically that's also the niche where PowerPC ended up when Apple dumped them :)

ghaff 5y ago

The striking thing about this time is that it is essentially transparent to most people. Probably more third party programs have been broken by security changes that Apple has made over the past five years than have been broken by the M1 transition. Yes, there are performance implications but M1 is sufficiently fast and most important performance sensitive programs are being quickly ported that it doesn't matter that much. (And, for most people, ultimate performance is mostly not a big deal on laptops these days.) Transitive's tech, combined with processors that can afford some inefficiency, is pretty much magic to anyone who remembers what ISA transitions used to look like. (As a hardware product manager, I lived through a couple of them including a Motorola 88K to x86 one on Unix.)

quitit 5y ago
I feel like more software died during the 32->64 bit transition than any other time in recent history. Lifting the rug is also par for the course for Apple, even artificially so with their latest crackdown on apps that don’t run well.
- astrange 5y ago
  That mostly seemed to be video games (at least their developers complained the most), but I just had to wonder why they couldn't make their games 64-bit-ready with 15 years of warning. Btw, 32-bit Windows binaries still run through WINE, just not 32-bit Mac binaries.
  - stormbrew 5y ago
    Games just don’t have a long enough tail to make this kind of effort worth it. They aren’t like business software where you try to keep people on the hook for years. There’s little reason to keep updating a game past its first few post release patches. So you can assume most of their code bases have been untouched for years, were probably written by people who are no longer around, and haven’t been compilable by modern toolchains for a long time. The tech debt is just insurmountable most of the time.
  - jedberg 5y ago