The M3 Studio can have 512GB of RAM, the M4 Studio maxes out at 128GB, and the M5 laptop 32GB. I assume the eventual M5 Studio will allow more, but can anyone explain why newer generation chips are getting paired with lower max memory? There must be some marketing/sales reason I don't understand.
The M3 version of the Studio can have either an M3 Ultra chip with up to 512GB RAM, or an M4 Max with up to 128. And these aren’t different generations of the product, they were announced at the same time: https://www.apple.com/newsroom/2025/03/apple-unveils-new-mac...
My guess is they’re doing this to solve some yield issue. M4 is a more difficult process with worse yield, so harder to make an Ultra version today. M3 has better yield so it’s tolerable to make an Ultra.
Probably either a tick/tok pattern forming, or a “previous gen gets an ultra version” pattern forming.
I see, thank you. I'd have thought the memory size was (mostly) independent of the chip "level" (Ultra, Max, whatever), but my understanding of memory controllers is really limited.
The cynic in me thinks Apple is diverting/prioritizing M4 Mac with 512GB yields for their own internal use first, building out their Apple AI Private Cloud Compute servers
I was so impressed with the MacBook M3 Pro that my daughter had that I switched from Windows desktop and laptops to the Apple ecosystem. And I'm a veteran Microsoft user from the '90s!
Currently using a MacBook Air M4 and a Mac Mini M4.
Welcome to Macintosh (I've seen CPUs go from 68k -> PPC -> Intel -> Silicon). The hardware architecture just keeps improving (but can't say the same for macOS).
>so impressed with the MacBook
The sound on these Apple Silicon laptops is absolutely impressive, making an immediate & lasting impression upon first-timers.
----
The basic configuration M4 (mini) is a very capable system, with an incredible price... it's just slightly slower than an M2Pro, for half the price.
I have an M3 (air) which is ignificantly less-capable than either (the mistake Apple made was default 8GB configuration).
3 nanometers is generally a marketing number to indicate relative fineness of the process. But it has real relevance to the physical structures that are manufactured by the process.
Transistors have many distinct features. There's the current source and drain, with the gate between them. There's a number of different ways to build these structures.
As a lousy analogy, a light switch has lots of parts: wire terminals, contact points to make a tight connection to a conducting blade that can be moved with a fancy handle, etc.
Maybe you want to print your switch with your home 3D printer. A really fancy one that can print lines that are 3nm in width on a good day.
I have an M2Pro (mini), an M3 (air), and an M4 (mini).
The above basic-configuration models are listed from fastest to slowest, oldest to newest, most-expensive to least...
The M3 Studio can have 512GB of RAM, the M4 Studio maxes out at 128GB, and the M5 laptop 32GB. I assume the eventual M5 Studio will allow more, but can anyone explain why newer generation chips are getting paired with lower max memory? There must be some marketing/sales reason I don't understand.
The M3 version of the Studio can have either an M3 Ultra chip with up to 512GB RAM, or an M4 Max with up to 128. And these aren’t different generations of the product, they were announced at the same time: https://www.apple.com/newsroom/2025/03/apple-unveils-new-mac...
My guess is they’re doing this to solve some yield issue. M4 is a more difficult process with worse yield, so harder to make an Ultra version today. M3 has better yield so it’s tolerable to make an Ultra.
Probably either a tick/tok pattern forming, or a “previous gen gets an ultra version” pattern forming.
I see, thank you. I'd have thought the memory size was (mostly) independent of the chip "level" (Ultra, Max, whatever), but my understanding of memory controllers is really limited.
The cynic in me thinks Apple is diverting/prioritizing M4 Mac with 512GB yields for their own internal use first, building out their Apple AI Private Cloud Compute servers
I was so impressed with the MacBook M3 Pro that my daughter had that I switched from Windows desktop and laptops to the Apple ecosystem. And I'm a veteran Microsoft user from the '90s!
Currently using a MacBook Air M4 and a Mac Mini M4.
Welcome to Macintosh (I've seen CPUs go from 68k -> PPC -> Intel -> Silicon). The hardware architecture just keeps improving (but can't say the same for macOS).
>so impressed with the MacBook
The sound on these Apple Silicon laptops is absolutely impressive, making an immediate & lasting impression upon first-timers.
----
The basic configuration M4 (mini) is a very capable system, with an incredible price... it's just slightly slower than an M2Pro, for half the price.
I have an M3 (air) which is ignificantly less-capable than either (the mistake Apple made was default 8GB configuration).
I love how they claim 6x CPU / GPU performance increases, where the numbers clearly say ~2x.
Regardless, they are still great chips.
I don’t love that.
I also like how they call it a "3 nm process".. bruh DNA is 2 nanometers. Are you saying your transistors are like a tight fit on a DNA strand?
3 nanometers is generally a marketing number to indicate relative fineness of the process. But it has real relevance to the physical structures that are manufactured by the process.
Transistors have many distinct features. There's the current source and drain, with the gate between them. There's a number of different ways to build these structures.
As a lousy analogy, a light switch has lots of parts: wire terminals, contact points to make a tight connection to a conducting blade that can be moved with a fancy handle, etc.
Maybe you want to print your switch with your home 3D printer. A really fancy one that can print lines that are 3nm in width on a good day.