3 ms·
> and you can get 4GB more memory than the prior maxiumum which frankly seems a small jump FTA: He speculates the M3 variant Apple is testing is the base-level
by Someone 3y ago
> and you can get 4GB more memory than the prior maxiumum which frankly seems a small jump
FTA: He speculates the M3 variant Apple is testing is the base-level M3 Pro the company plans to release sometime next year.
Speculation, yes, but coming from Bloomberg, I think it’s more a bit of a controlled leak. Also, if they go from 5nm to 3nm, area-wise, that is comfortably over twice the number of transistors (5² vs 3² is 2.777…). I think a ≈doubling of memory would be possible.
Even if it isn’t, 20% in a generation is nothing to sniff at. The times of huge jumps every few months are long behind us.
- ggm 3y agoThe memory isn't on the die, its soldered to the board. It's normal dimms isn't it?
- foldr 3y agoIt's not on the die, but there's no DIMM in the normal sense of the word: https://www.macworld.com/wp-content/uploads/2023/01/iFixit-M2-MacBook-Air-M2-teardown-accelerometer.jpg?quality=50&strip=all https://www.macworld.com/wp-content/uploads/2023/01/iFixit-M... (That's an M2 Macbook Air.)
- ggm 3y agoDoes it track the nm density of the core? I would think it's sourced as an independent path at whatever functional density works for the vendor at that time meeting spec. which is bus speed limited, not on-die. sorry if I mis-use terms of art, It looks to me like its unlikely to be specced to match the apparent design density of VLSI used to make the CPU.
- foldr 3y agoSure, I don't think there's any connection between the two silicon processes.