> moving from 200 mm to 300 mm requires a total redesign and for some chips the redesign is not worth it. There are other differences between 200 and 300 mm than just the size of the wafer.
That is very interesting, I would not have thought so and I'd like to know more, can you elaborate?
Naively I would have expected that you would "just" make new 'master masks' (or whatever they're properly called) which have more copies of the exact same design - i.e. something that takes some money to fabricate the new tooling but not any significant design choices.
The entire 300mm process is different on a microscopic level. For instance generally 200mm fabs don't need to use copper at all because the benefit of that is the dual-damascene process which isn't seen on the bigger tech nodes. Wafer stresses on 300mm wafers is dealt with a lot differently too so that is factored into the manufacturing processes as well.
The major transition from 200mm -> 300mm will be when people like Broadcom use newer nodes for their chips as they go from iPhone 13->14->15->16...and eventually that will require 300mm processing.
They likely have changed the process to better optimize for the bigger size. This might mean the layout of individual transistor structures may have to be different due to chemistry (different chemical processes may be used to accommodate the requirement of even distribution over larger area), or things like different number or order of various doping and metalization layers, etc. Most designs are packed to maximize efficiency, so even if a small part of the chip needs to be modified, it can require cascading redesign.
That is very interesting, I would not have thought so and I'd like to know more, can you elaborate?
Naively I would have expected that you would "just" make new 'master masks' (or whatever they're properly called) which have more copies of the exact same design - i.e. something that takes some money to fabricate the new tooling but not any significant design choices.