5 ms·
As you get into more and more modern designs, there are more high speed signals and the motherboards get increasingly more complex. Not that it can't be done,
by geerlingguy 1y ago
As you get into more and more modern designs, there are more high speed signals and the motherboards get increasingly more complex.
Not that it can't be done, but the work to reproduce something made at the cutting edge in the 2000s feels like it'd be an order of magnitude harder than 70s/80s designs.
Though I'm always amazed what the retro communities will do to preserve the tech for future generations!
- whartung 1y agoThat’s alright though. SE/30 was Peak Macintosh anyway.
- sneak 1y agoThe 9600/350 was a thing of beauty.
- runjake 1y agoLink: https://en.wikipedia.org/wiki/Power_Macintosh_9600 https://en.wikipedia.org/wiki/Power_Macintosh_9600
- mixmastamyk 1y agoMy memory is that the Power Towers were incredibly hard to service. If correct, that's a shame, because the IIci I used at work was lovely to work with.
- linguae 1y agoI’ve read that the predecessor to the 9600, the 9500, wasn’t the easiest machine to work on. The 9600 had a more convenient pull-down case, which was continued in the designs of the beige G3 tower, the blue and white G3 tower, and the G4 towers.
- crest 1y agoAn unlike the 9500 you could even get at the RAM slots without breaking the damn plastic clips as use tear down the whole machine. Whoever combined the 9500 case and mainboard deserves as special place in engineering hell.
- vondur 1y agoHeck, I'd be happy with a board that had the power/emulation of a 68040 so we can run MacOS 7.6 and some of old apps from back in the day.
- bigfatkitten 1y ago68040 CPUs are easy to find. Motorola was shipping them in ASTRO mobile radio infrastructure equipment well into the 2000s, and a lot of that gear is getting scrapped now.
- rbanffy 1y agoA small ARM board can do the job easily, but that won’t be very close to the actual experience. Playing with the C64 Maxi made me understand how important the physicality of the system is to the experience. It’s nice to have a physical 68000, but that level of fidelity is not really necessary for a user to understand what a Mac was about. A keyboard with a locking Caps Lock key is.
- cosmic_cheese 1y agoA big problem with emulation that has yet to be overcome entirely is the added latency. Any of my newer machines can emulate a 68k Mac at full speed and I can hook up my Apple Extended Keyboard II to help replicate the physical portions of the experience, and yet using a real old Mac feels notably different simply because it doesn’t suffer from the latency papercuts brought by USB and modern operating system. That’s one of the things that jumps out at me when I pull out my old PowerBook G3 and boot it into OS 9 on occasion: it feels incredibly responsive relative to modern counterparts, especially with an SSD removing the wait times that normally came with a spinning disk, despite it being only a tiny fraction as powerful.
- rbanffy 1y agoYou can mitigate the latencies by making the CPU faster, pushing more work to the GPU, using an analog VGA output, or trimming down the OS so that as little as possible preempts the emulator.
- johnklos 1y agoConsidering that we've moved from wire wrapping to being able to design and order multi-layer circuit boards, and we've gone from 74 series and basic PALs to CPLDs and FPGAs that regular people can program, I don't think what tinkerers can do will hit any barriers any time soon. The ability to recreate classic computing is wonderful, both in preservation of history and in making things available to people who hadn't even been born when these machines were new :)
- userbinator 1y agoFortunately for later CPUs, especially on the PC/x86 they are usually based on reference designs, and the amount of documentation available in electronic form much greater. Late 2000s is when they started closing up and being more secretive, and I'd consider that a greater concern.
- phire 1y agoThough, the late 90s feels achievable. The Front Side Bus of the Pentium III maxed out at 133 MHz, single transfer (and was often configured at 100 MHz for lower spec CPUs), and the AMD K6 was even slower. I don't have much PCB design experience, but my understanding is that 133MHz is quite achievable for hobbyists these days. Things very quickly go off the rails after that.