Of course, the term disk originates from the shape of a flat, circular storage medium (hard or floppy disk)... We're never going to escape commonly used terms whose origins are long gone, especially in high tech fields.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
The Mac mini 2024 actually has socketed SSD
- Thread starter Gnattu
- Start date
- Sort by reaction score
But the number of pins on Apple's NAND chips may just be the number of pins you would get if you asked an SSD maker to produce a controllerless version of a standard SSD (in which case the interface that's exposed to the outside would have a number of pins equal to the number of channels that normally go into the controller internally).
I.e., the question at hand isn't whether there's some technical advantage to having the controller be on the processor. The question is whether there are standard methods to produce "controllerless SSDs" (or, if you prefer "unmanaged flash") and, if so, whether there is some technical advantage to doing this using a proprietary approach (and one that is thus difficult for aftermarket providers like OWC to replicate) over a standard approach.
Last edited:
That's incorrect. In fact, the first MacBook with Intel was made so that replacing the HD was extra easy. All you had to do was remove the battery and remove a few screws. The HD slid out easily and could be quickly replaced if it was damaged or you wanted an upgrade.
The battery and RAM were also easily replaceable in most of these models, although Apple started making it less common as time went on
That lasted until at least 2014, if not 2015.
It was a great way to save money: Buy a lower-end amount of HD and RAM and upgrade it yourself. It's obvious why Apple has worked against it, but we shouldn't be making excuses for them.
All such NAND option, or "single package BGA ssds" available on the market have significantly larger package size and those cannot be used in iPhones due to its size alone:
Even in Macs such option still has its own problems to have the controller embedded in the NAND packages. Apple can scale the internal storage from 256GB to 8TB simply by adding more or using larger NANDs, but having controllers being embedded into the same package is not as trivial and how much storage you can have in a single package is more limited, and once you want to split it out to make the packaging easier then why not using external controllers in the first place?
Also, having embedded nvme controllers in the SoC can directly using the system RAM as the SSD DRAM while traditional approach you have to either make an on-module DRAM cache too boost performance or use the HMB DRAMless technology but sacrificing performance. Apple's implementation is definitely not the fastest SSD on the market, but it is still faster than other options with similar module form factor size.
Apple also has special requirements to its SSD controllers to implement their platform security properly, this includes the on-the-fly full disk encryption. Since the very first Apple Silicon Mac, the full disk encryption is no longer optional and your data partition will be encrypted even when the FileVault is off. On Apple Silicon Macs, all FileVault does is to encrypt you machine key additionally with your user key. The standard encryption ensures others cannot read your data simply by removing your NAND, and the FileVault ensures your data cannot be read even they managed to get the whole machine but does not know your user key. Yes there are standard self-encryption SSDs over the market and the ones for enterprise also have serious security features. But those still comes with its own controller and still faces the problems listed above, and if Apple has the in-house encryption system originally developed for iPhone and being tested by gov agents like FBI already, just use that is not only easier to reduce potential security flaws, but also a more economical choice.
To make it simple, the ability to use the same hardware and software for iPhone, iPad and Mac is a big enough advantage and a strong enough reason for Apple to use their in-house design. This reduces costs on their side. People may argue this is anti-consumer, but the available "consumer-friendly" way to do the same thing will just give us worse designed product so there is really no silver-bullet.
In my own opinion, the problem was never how "proprietary" Apple developed their storage system, it is how the storage is ridiculously priced.
Easy to replace a mid 2014 MacBook Pro battery? Holy macarony, that`s the most optimistic comment I have ever read. Anywhere.
This is all about the downsides of having the controller in the same package as the NAND. None of it addresses the question I raised, which is whether there is a non-proprietary way of having the controller off-package.All such NAND option, or "single package BGA ssds" available on the market have significantly larger package size and those cannot be used in iPhones due to its size alone:
Even in Macs such option still has its own problems to have the controller embedded in the NAND packages. Apple can scale the internal storage from 256GB to 8TB simply by adding more or using larger NANDs, but having controllers being embedded into the same package is not as trivial and how much storage you can have in a single package is more limited, and once you want to split it out to make the packaging easier then why not using external controllers in the first place?
Also, having embedded nvme controllers in the SoC can directly using the system RAM as the SSD DRAM while traditional approach you have to either make an on-module DRAM cache too boost performance or use the HMB DRAMless technology but sacrificing performance. Apple's implementation is definitely not the fastest SSD on the market, but it is still faster than other options with similar module form factor size.
Apple also has special requirements to its SSD controllers to implement their platform security properly, this includes the on-the-fly full disk encryption. Since the very first Apple Silicon Mac, the full disk encryption is no longer optional and your data partition will be encrypted even when the FileVault is off. On Apple Silicon Macs, all FileVault does is to encrypt you machine key additionally with your user key. The standard encryption ensures others cannot read your data simply by removing your NAND, and the FileVault ensures your data cannot be read even they managed to get the whole machine but does not know your user key. Yes there are standard self-encryption SSDs over the market and the ones for enterprise also have serious security features. But those still comes with its own controller and still faces the problems listed above, and if Apple has the in-house encryption system originally developed for iPhone and being tested by gov agents like FBI already, just use that is not only easier to reduce potential security flaws, but also a more economical choice.
To make it simple, the ability to use the same hardware and software for iPhone, iPad and Mac is a big enough advantage and a strong enough reason for Apple to use their in-house design. This reduces costs on their side. People may argue this is anti-consumer, but the available "consumer-friendly" way to do the same thing will just give us worse designed product so there is really no silver-bullet.
In my own opinion, the problem was never how "proprietary" Apple developed their storage system, it is how the storage is ridiculously priced.
The two are tied together. The proprietary nature of the storage, which in turn means aftermarket upgrades aren't available, becomes a problem precisely because of the OEM price.
Off package and embedded into the SoC like Apple does and provides all the same features and having similar performance as we currently have -> No, there is no such "open" technology that fits all the needs, and Apple just makes their own.
And to this statement: no it is not. Just having a controller-less version will require significantly more pins and even requires even higher quality PCB for better signal integrity. Apple's NAND saves data pins a lot due to their design of the interface and the wire protocol, not due to just removing the controller from the module.
I answered your question by telling you the downsides of having the controller in the same package as the NAND because I thought you would know that NANDs after controllers saves pins, but it seems like the confusion started there.
If so, then unmanaged flash devices (other than those made for Apple) would typically have more pins than managed flash devices. If we were to examine the former, is that what we would find?
Depends on what kind of "unmanaged interface" such flash exposes. The very simple SPI flash can has as small as 4 pins and is still "unmanaged". Keep in mind that for a transitional nvme SSD, a controller can manage multiple NAND packages and each requires its own data pins, but the controller only has to connect to the host with a finite number of pin count over the defined width PCIe bus.
You wrote a bit more than that. Selective removal of words changes the lot entirely.
You are correct! I looked up some of the older Apple manuals, they explicitly state that the HDD is user-replaceable. I must have been thinking about some of the later models, where the HDD was easy to access but was not considered to be user-replaceable by Apple. Of course, all of that was rendered moot by the retina laptops with custom storage modules.
Yeah, my 2008 MBP came with an access door that gave direct acces to the battery and HDD. And while that access panel wasn't present on my 2011 MBP and 2014 MBPs, their drives could be replaced without needing to remove the battery first.
The only one I upgraded was the 2011 model, since it came with an HDD, which I later upgraded it to an SSD after they became more commonly available. My 2014 MBP came with a 1 TB SSD, and that was big enough for my needs.
I also upgraded the SSD on my 2019 27" iMac from 512 GB to 2 TB, even though that's technically not user-replaceable. I supplied the SSD, and had the upgrade done by the Apple-authorized service center at my university's bookstore, which they said would enable me to keep the warranty. They charged me a very reasonable $100 for the work, which requires removing the device's glass panel (something I didn't want to touch).
Last edited:
The SSD in the Mac mini is upgradable. On a 256 GB Mac mini, iFixit swapped in a 512 GB SSD from a different machine, and after running Configurator, it worked just fine.
That's interesting because I thought this was something you couldn't do on the Studio—that the Studio only allows a lateral swap, i.e., replacment using a NAND chip of the same size. But those were preliminary results; I don't know if further investigation of the Studio has shown otherwise.
The Studio works the same exact way, it's just more complicated because it has two NAND modules. If it's a single module config (4-NAND, for 2TB or less), you can just swap that over to any other Mac Studio (of the same generation, M1 Studio modules are not compatible with M2 and vice versa) to the correct port, and it will work just fine. In the case of the 8-NAND, two-module configs (4TB and 8TB), you must keep those modules paired and install them into the same slots on the other Mac Studio, and it'll work just fine. The issues arise if you combine modules pulled from machines with different configurations. For example, say you have two M1 Mac Studios, each with a 2TB config (single 2TB NAND module in each). You can't just take the 2TB module out of one and put it in the second slot of the other and expect a working 4TB result, as even though that could be a valid config, each set of NANDs is already programmed to be in a single 2TB config, which cannot be altered. Complicating the Mac Studio situation further is that the actual slot positions are swapped between the M1/M2 Max and Ultra models, so if transferring between those systems, the modules need to be installed opposite of how they were in the other.
So, TL;DR, you CAN swap different capacity storage configs between Mac Studios without any issue, so long as you adhere to the valid/programmed config and install the modules in the correct slots.
Now, in the case of the M4 Mini, because it only has one module type without any variations (single module only with 2 NANDs), there is no chance for error when swapping, hence why any attempt to do so has been successful.
Last edited:
Is there a possibility that this wouldn't work for, say, 256 GB <--> 2 TB because, I dunno, one is single-sided and the other is double-sided?
No, because all the valid configs for the M4 Mini utilize both chips, and if there were a valid single-NAND config for it, that would work just fine as well. Again, because there are not variations between the modules and because there is only one slot, there is no chance for error when swapping. If the module worked fine in one M4 Mini, it is guaranteed to work just fine in ALL M4 Minis.
Note that the M4 Pro model Mac mini has support for 8 NANDs on a single module, so the same would apply there as well. But of course you cannot swap modules between a regular M4 Mini and an M4 Pro model.
The whole reason for so many issues swapping in the Mac Studio was literally just because they used two separate modules, which lead people to falsely believe they could just be arbitrarily combined. With only one slot, there is no chance for such errors.
All Mac Mini modules are double-sided
In the case of the M4 Pro, do we know yet how the NAND numbers correspond to the total capacity? I'd assume 1TB and 2TB still uses 4 NANDs, with 4TB and 8TB uses 8 NANDs?
While unlikely, this difference may introduce incompatibility between 4 and 8 NAND boards. At least concerning the soldering stage, with capacitors missing.
While unlikely, this difference may introduce incompatibility between 4 and 8 NAND boards. At least concerning the soldering stage, with capacitors missing.
From what I can tell, I think all the boards are the same, they just either have all 8 NANDs populated or only 4 depending on config. I doubt anything would change in that regard, so it would be safe to assume that 2TB and lower will use 4 NANDs, and 4TB/8TB will use all 8 as usual.
You are most likely correct in that the passive components will also be missing for the second set of 4 NANDs on modules originally configured with only 4 NANDs.