3 ms·

That's a nice way of saying "Deliberate hardware incompatibility", when those modules look pretty clearly to be MSATA in form-factor.

14u2c 5y ago
Yes, mechanically, but certainly not electrically.
- colejohnson66 5y ago
  I can’t speak for Apple’s design process, but for me, I’ve thought of and seen some personal projects use PCIe connectors because they’re cheap and have dozens of connections. I even saw a teardown on EEVBlog a while back of a piece of test equipment that used PCIe connectors for board-to-board interconnects. For $2 or so in quantity 1, I can order an x16 connector from DigiKey and have 164 pins to connect two modules together. And being an edge connector, I don’t need one for each board (like “DuPont” or DIN 41612 does). If the board won’t be removed, you could even get away with the free tin plating (skipping gold).
musicale 5y ago
> That's a nice way of saying "Deliberate hardware incompatibility" As others have noted, that doesn't necessarily mean "incompatible primarily for the sake of breaking third-party upgrades." It may very well be that there are advantages (such as better or more predictable performance, lower component cost, better power or heat dissipation, or higher reliability) to using raw modules with a custom controller rather than going through an SSD interface. Perhaps (for example) there are advantages to having the SSD's cache in the Mac's uniform memory and controlling when writes are performed. (As an interesting aside, Steve Jobs famously convinced Shugart to make a stripped-down "SA390" floppy drive for the Disk II to save money. With Woz's controller and improved encoding they also managed to squeeze more data onto the floppy disks, so it saved [Apple] money while improving density and read rate - though possibly increasing the likelihood of bad sectors and data loss.)
- adgjlsfhk1 5y ago
  Well we know they don't have significantly better power/heat because those aren't problems for regular SSDs. They may have lower component cost, but they certainly aren't acting like it (they're charging over 2x as much as high quality NVME SSDs). They might have better performance or reliability, but that is just wild speculation with no evidence to back it up.
  - musicale 5y ago
    There is also no evidence to back up the implication that Apple's primary reason was to prevent third party upgrades. Based on what I've heard from people who work (or worked) at Apple, that isn't normally how they roll - usually there's a cost, performance, user experience, security, reliability, or functionality (etc.) reason and breaking third party upgrades, compatibility, or repairs is just a happy (for Apple) side-effect. ;-) Placing the SSD cache in unified RAM would seem like a plausible performance and cost justification to me, not "wild speculation" as you claim. I've seen papers describing how to optimize the use of flash storage that require low-level control of the device as well as modifications to the OS/file system or even the cache and memory system - all things Apple could certainly do. However, as you note, evidence will tell. Regarding whether Apple passes its own savings on to customers, I will note that the Disk II was said to be "obscenely profitable." In general Apple maintains some of the highest margins in the industry. Which is not to say Apple products are a poor value (I generally consider them an excellent value for what you get, and for my own use cases) but it is saying that they are very good at keeping money in their own pocket.

14u2c 5y ago

Yes, mechanically, but certainly not electrically.

colejohnson66 5y ago
I can’t speak for Apple’s design process, but for me, I’ve thought of and seen some personal projects use PCIe connectors because they’re cheap and have dozens of connections. I even saw a teardown on EEVBlog a while back of a piece of test equipment that used PCIe connectors for board-to-board interconnects. For $2 or so in quantity 1, I can order an x16 connector from DigiKey and have 164 pins to connect two modules together. And being an edge connector, I don’t need one for each board (like “DuPont” or DIN 41612 does). If the board won’t be removed, you could even get away with the free tin plating (skipping gold).

musicale 5y ago

> That's a nice way of saying "Deliberate hardware incompatibility" As others have noted, that doesn't necessarily mean "incompatible primarily for the sake of breaking third-party upgrades." It may very well be that there are advantages (such as better or more predictable performance, lower component cost, better power or heat dissipation, or higher reliability) to using raw modules with a custom controller rather than going through an SSD interface. Perhaps (for example) there are advantages to having the SSD's cache in the Mac's uniform memory and controlling when writes are performed. (As an interesting aside, Steve Jobs famously convinced Shugart to make a stripped-down "SA390" floppy drive for the Disk II to save money. With Woz's controller and improved encoding they also managed to squeeze more data onto the floppy disks, so it saved [Apple] money while improving density and read rate - though possibly increasing the likelihood of bad sectors and data loss.)

adgjlsfhk1 5y ago
Well we know they don't have significantly better power/heat because those aren't problems for regular SSDs. They may have lower component cost, but they certainly aren't acting like it (they're charging over 2x as much as high quality NVME SSDs). They might have better performance or reliability, but that is just wild speculation with no evidence to back it up.
- musicale 5y ago
  There is also no evidence to back up the implication that Apple's primary reason was to prevent third party upgrades. Based on what I've heard from people who work (or worked) at Apple, that isn't normally how they roll - usually there's a cost, performance, user experience, security, reliability, or functionality (etc.) reason and breaking third party upgrades, compatibility, or repairs is just a happy (for Apple) side-effect. ;-) Placing the SSD cache in unified RAM would seem like a plausible performance and cost justification to me, not "wild speculation" as you claim. I've seen papers describing how to optimize the use of flash storage that require low-level control of the device as well as modifications to the OS/file system or even the cache and memory system - all things Apple could certainly do. However, as you note, evidence will tell. Regarding whether Apple passes its own savings on to customers, I will note that the Disk II was said to be "obscenely profitable." In general Apple maintains some of the highest margins in the industry. Which is not to say Apple products are a poor value (I generally consider them an excellent value for what you get, and for my own use cases) but it is saying that they are very good at keeping money in their own pocket.