3 ms·

Quite possibly - I was a huge Apple fan who's now using a PC because I was fed up with the lack of viable options for me.

rbanffy 6y ago
Apple will certainly offer an ARM-based MacPro, but I'm assuming it'll be a very different beast - current one maxes out at 1.5TB of RAM and it doesn't seem likely anyone will integrate that much memory on a chip anytime soon ;-) Memory bandwidth is one key feature impacting M1's performance. When Apple builds an ARM-based MacPro, we can expect something with at the very least 5 DDR5 channels per socket. It's clear, from this, the M1 is a laptop/AIO/compact-desktop chip.
- sliken 6y ago
  The M1 already has 8 LPDDR4x channels per socket, running at 4266MHz.
  - rbanffy 6y ago
    My bad. I was looking at the specs. It's 300GBps, which is roughly 5x DDR5 IIRC. So yes, at the very least 8 DDR4 channels, or one per core, but I'd expect more from a workstation-class board. Now, speaking of the board, all those memory channels will be funny.
    - sliken 6y ago
      Funny? 8 channels is the standard AMD Epyc socket. Most threadrippers (AMD's workstation chip) are 4 channel, but there is a variant that's 8 channel.
      - rbanffy 6y ago
        I would expect more, so that cores don't get memory starved. The M1 has 4 fast cores and 4 slow ones. If we imagine an M2 with 8 fast cores, I would expect it to need 16 channels to have the same performance. That's a lot.
        sliken 6y ago
        Dunno, the M1 CPU package is tiny, thin, power efficient, etc. It's got 4 memory chips inside the package. I don't see any particular reason why a slightly larger package could have 4 memory chips on one side, and 4 chips on the other to double the memory bandwidth and memory size. However the M1 is already pretty large (16B transistors), upgrading to 8 fast cores is going to significantly increase that. Maybe they will just go to a dual CPU configuration which would double the cores, memory bandwidth, and total ram. Or move the GPU and ML accelerator offchip.
        davidf18 6y ago
        I'm guessing the next "M2" chip will be fabbed on the upcoming TSMC 5nm process, 2nd version. In 2022 there is projected to be the TSMC 4nm process. Thus, there will be room for more transistors. Expect also a density increase and capacity for DRAM chips (and SSD chips).

rbanffy 6y ago

Apple will certainly offer an ARM-based MacPro, but I'm assuming it'll be a very different beast - current one maxes out at 1.5TB of RAM and it doesn't seem likely anyone will integrate that much memory on a chip anytime soon ;-) Memory bandwidth is one key feature impacting M1's performance. When Apple builds an ARM-based MacPro, we can expect something with at the very least 5 DDR5 channels per socket. It's clear, from this, the M1 is a laptop/AIO/compact-desktop chip.

sliken 6y ago
The M1 already has 8 LPDDR4x channels per socket, running at 4266MHz.
- rbanffy 6y ago
  My bad. I was looking at the specs. It's 300GBps, which is roughly 5x DDR5 IIRC. So yes, at the very least 8 DDR4 channels, or one per core, but I'd expect more from a workstation-class board. Now, speaking of the board, all those memory channels will be funny.
  - sliken 6y ago
    Funny? 8 channels is the standard AMD Epyc socket. Most threadrippers (AMD's workstation chip) are 4 channel, but there is a variant that's 8 channel.
    - rbanffy 6y ago
      I would expect more, so that cores don't get memory starved. The M1 has 4 fast cores and 4 slow ones. If we imagine an M2 with 8 fast cores, I would expect it to need 16 channels to have the same performance. That's a lot.
      - sliken 6y ago
        Dunno, the M1 CPU package is tiny, thin, power efficient, etc. It's got 4 memory chips inside the package. I don't see any particular reason why a slightly larger package could have 4 memory chips on one side, and 4 chips on the other to double the memory bandwidth and memory size. However the M1 is already pretty large (16B transistors), upgrading to 8 fast cores is going to significantly increase that. Maybe they will just go to a dual CPU configuration which would double the cores, memory bandwidth, and total ram. Or move the GPU and ML accelerator offchip.
        davidf18 6y ago
        I'm guessing the next "M2" chip will be fabbed on the upcoming TSMC 5nm process, 2nd version. In 2022 there is projected to be the TSMC 4nm process. Thus, there will be room for more transistors. Expect also a density increase and capacity for DRAM chips (and SSD chips).