Cost. You get more bandwidth by having more interconnect which takes up die space which means you get fewer per wafer and lower your yield. TSMCs N3 process is very expensive and is rumored to be dealing with yield issues as well.
Consumers aren’t going to spend more money for memory bandwidth. Nobody cares; the target market for the Pro is not people who want every last 1% of performance.
The 25% less bandwidth is related to the 25% fewer performance core count (8 to 6), in favor of efficiency cores (4 to 6).
Odds are Apple looked at real world usage of the Pro chips and found that performance cores were rarely fully utilized while efficiency cores were, so it made sense to shift the design. And with fewer performance cores, you need less memory bandwidths
This whole thread is full of poeple who couldn’t even explain why memory bandwidth is important (hint: doubling bandwidth will haze zero impact on performance unless it was the bottleneck).
Curious what financial benefit is there to making it have 25% less memory bandwidth?
Is it to drive consumers to a more expensive product?
I get they have 25% less bandwidth. But why?
Cost. You get more bandwidth by having more interconnect which takes up die space which means you get fewer per wafer and lower your yield. TSMCs N3 process is very expensive and is rumored to be dealing with yield issues as well.
Consumers aren’t going to spend more money for memory bandwidth. Nobody cares; the target market for the Pro is not people who want every last 1% of performance.
The 25% less bandwidth is related to the 25% fewer performance core count (8 to 6), in favor of efficiency cores (4 to 6).
Odds are Apple looked at real world usage of the Pro chips and found that performance cores were rarely fully utilized while efficiency cores were, so it made sense to shift the design. And with fewer performance cores, you need less memory bandwidths
This whole thread is full of poeple who couldn’t even explain why memory bandwidth is important (hint: doubling bandwidth will haze zero impact on performance unless it was the bottleneck).
Especially when you consider that the average DDR5 in PCs runs about 70GB/s, Apple has heavily over provisioned bandwidth for the CPU anyway.
I’d be interested in seeing if their new GPU is constrained by memory bandwidth though.
All I want is a 64gb 14” M3 and I have to get the Max for that. So no issues here but the 3k hole in my wallet with trade in.