No. User replaceable RAM is never going to be a thing again on Apple products.
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.
M1x with 'secondary' RAM possible?
- Thread starter Mr Screech
- Start date
- Sort by reaction score
It does go against the heart of the idea of mobility products as apple see it.
The space slotted RAM requires can always be traded off for longer battery life (larger battery), more beautiful more portable - lighter, thinner forms (reclaim the space entirely), better performance per watt per $. Also, ease of support by taking the randomness of third-party items out (just swap out the device instead). I don't know how the genius bar would work practically speaking without known configurations.
But yes, it doesn't strictly go against the idea of SOC. For non-mobile "professional" items that plug into a wall and where thinness and lightness don't matter as much and absolute performance is a goal, like a Mac Pro, maybe.
It does when you are Apple and the idea is A: make as much profit on the original sale as possible and B: optimize the components and their interconnectivity for maximum performance and efficiency.
I very much doubt that Apple makes more money from their current soldered-on RAM than they did couple of years ago with socketed RAM. Apples RAM upgrade pricing have always been consistent but the cost of the RAM they use went up significantly. And even in the earlier days the number of people who did aftermarket RAM upgrades was negligible.
My point is: I don’t think that profit is a deciding factor here. Functionality is.
I watch the local used iMac market and, in general, a lot of people sell Macs with the base amount of RAM but with good processors. They found that their iMac got slow because of the Fusion drive and lack of RAM and upgrade to the next iMac with the base amount of RAM. The stores buy these up, add RAM and may change the Fusion drive to an SSD and add a hefty markup.
My point: Apple doesn't care about functionality or expandability. Socket RAM loses them money because there are cheaper alternatives 3rd party. No one is buying an 8GB laptop with the intention of buying 8GB more from Apple later. or buying a 128GB SSD and it to a 512GB from Apple later. The goal is to replace that computer when more memory is needed.
There is no way Apple could get away with a $200 upcharge from 8GB to 16GB of RAM if there were an alternative, especially if the option is there for someone to defer that purchase to when that spec RAM became more affordable
Bingo!
Instead we'd have the OWC's of the world selling kits and lots of forum posts about great sources of great values for RAM upgrades and barefeats doing testing of the options, etc
There's no way, at all, that at least part of "locking it all down" by Apple isn't about --- money.
It's about money
Every with Apple now seems to be, first and foremost, about money
IIRC, LPDDR4X can only be soldered. Since M1 was designed for low-power products, I can understand why Apple, and others, use it in these types of products. From what I can tell, Apple's upgrade pricing isn't much higher than competitors.
Apple tends to make data-driven decisions. Most users won't ever open up the machine, much less install upgrades. If moving to soldered components really cost Apple a ton of hardware sales, they would have reversed course. It would seem that most users are comfortable accepting the trade-offs.
It's like any other product. You'll have some enthusiasts and tinkerers who want to mod them, but most users will simply use the product as-is.
Apple tends to make data-driven decisions. Most users won't ever open up the machine, much less install upgrades. If moving to soldered components really cost Apple a ton of hardware sales, they would have reversed course. It would seem that most users are comfortable accepting the trade-offs.
It's like any other product. You'll have some enthusiasts and tinkerers who want to mod them, but most users will simply use the product as-is.
There's so much more friction that keeps people in the ecosystem than simply "soldering down components" (or not).
Its' exceptionally hard, if not impossible, to isolate that variable and understand the relationship to sales.
Slotted RAM uses volume ( i.e. "space" ) more effectively than soldering RAM packages does. SO-DIMM slots allow the RAM DIMMs to overlaps, so you get a more RAM packages in a fixed volume. What it doesn't do is three things.
First, minimize z-height. Even using SO-DIMMs is taller than soldering the RAM packages to the board. It "thinnest" is a very high priority then "space" is not. If you declare war on one of the three dimensions , then pragmatically by proxy you are declare war on space. Thin laptops typically also constrain themselves to similar dimensions to the keyboard + trackpad. If minimize z height and put a upper cap on x and y then "space" is going to be capped where can't trade loss of z for more x and/or y.
Second, minimize power . a RAM socket typically is placed farther away from the CPU/GPU package. Even more so when the capacity goes up. Also more transitions ( e.g, RAM package , DIMM board , DIMM connect, logic board , CPU/GPU. versus RAM package , logic board , CPU/GPU. ).
Third, control vendor validation complexity. If Apple picks their own narrow subset of RAM package vendor (or vendors if don't want to single source) then don't have to worry about variances allows for by the RAM slot standards. They know the RAM subsystem characteristics because they implement the collection of packages.
Apple gets space back by either using "denser" (and more expensive) RAM packages or tossing other stuff (i.e., legacy storage form factors , diversity of sockets , etc. ) . As long as Moore's Las and RAM die stacking inside of RAM packages makes progress the denser RAM package will provide a path over the future. Short term Apple will tend to trail on max RAM capacity.
there are still variations in BTO options. "Genius" bar where they engage in no substantive diagnostics... kind of an oxymoron. "Something wrong with your box .... well here is another one. That will be $xxxx.yy dollars". $80 component failure and have to buy $1,000 board replacement. That is "simple" for Apple ( "more money" ) , but not quite as simple for the customer.
Support shouldn't be about making Apple support task easiest and making customers happy ( well, at least while still under the initial purchase warranty) . It should be about getting to correct diagnosis and solving the problem.
Soldering on the RAM is at least as much a 'win' for Apple as it is for customers ( incrementally better battery life. lighter laptop. etc. )
and yet the iMac 24" is just as about as thin as a iPad Pro. So aggressively thin, they squeeze the Ethernet port out into the power brick. If Apple applies hyper aggressive thinness to the larger screen iMac there isn't going to be a substantive base of "desktop" Macs left for a non "laptop" oriented SoC to go into. The Mini has almost always bow-waved on the laptop components. Probably still will when the MBP 16" gets a substantively more powerful SoC.
Apple's application of laptop constraints to desktops is the real substantive threat here for slotted RAM. It that because a dogma across the whole product line, then slotted RAM is probably on the way out. It will be cheaper for Apple just to do laptop oriented SoC and just "strech" ( e.g., chiplet/tile , multichip module ) those into modifications for the desktop. Avoiding larger triple (and quad ) digit GB RAM capacity gives them excuse to avoid having to implement robust ECC coverage also. (again cheaper) for a long time until the RAM density gets much more higher.
The modus operandi from the iOS/iPad product lines has been to use the minimum number of SoCs to cover the maximum number of products. A quite large amount of SoC and SoC component reuse. The Mac sales are 70+ % laptops. Which means Apple mostly needs laptop SoCs to have a viable Mac product line. The farther away from laptops you go the smaller the impact on the line up in terms of unit volume and revenues.
Soldering on the RAM lines up very tightly with Apple's openly stated primary objective for "Apple Silicon" ; performance/watt. Lower the power consumed and that produces a better ratio result. RAM max capacity cap versus power savings ... as long as perf/watt is goal #1, the former is likely to loose that trade off.
@deconstruct60
Good post - you've mostly convinced me on the models where space is a premium.
What do you think they'll do on machines where space isn't (or at least shouldn't be) at a premium?
It's really frustrating to think about massively expensive workstations that you have to decide ALL the specs on at the time of purchase and if anything substantial in ones use case changes, the only option is "a totally new machine"
Good post - you've mostly convinced me on the models where space is a premium.
What do you think they'll do on machines where space isn't (or at least shouldn't be) at a premium?
It's really frustrating to think about massively expensive workstations that you have to decide ALL the specs on at the time of purchase and if anything substantial in ones use case changes, the only option is "a totally new machine"
Apple very much cares about functionality. Apple does not cares about post-purchase upgradeability. It's a non-goal for their product, plain and simple.
I agree that no one buys an 8GB laptop from Apple with the intention of being 8GB more from Apple later, they directly buy a 16GB laptop from Apple. The amount of costumers that upgrade their laptops is very very small. Sure, tech savvy folks do, but they do not represent the typical user. An average user does not want to tinker with their machines, they have a professional do it. The average person also gets rid of their iPhone when the battery life goes down even though there is a perfectly fine and much more affordable option of getting battery service.
In fact, the only reason why Apple products sell like hot cake despite non-upgradeable components is because almost nobody cares about upgradeable components in practice. Companies wouldn't do it if customers were so against this practice. For most customers, upgradeability adds very little value, but it does come with a cost (e.g. want upgradeable memory? Forget about excellent battery life).
They did get away with it, for years. And they still do, because they still sell machines with SO-DIMMs, and customers happily pay Apple's upgrade price. Sure, some people would buy a base config and put their own RAM sticks in there to save a couple of bucks, but its not representative of average user behavior.
The 27 inch iMac is an exception. You can buy it with 8 GB of RAM and expand it later to 128 GB.
Functionality would also mean providing legacy ports that people request (SD HDMI etc) as well as providing an upgrade path for the future by allowing people to add memory/storage as their needs grow. They compromise functionality with the quest for ultra thin aesthetics and either profits up front or profits later when the OS requires more RAM or the customers needs grow and the computer cannot be upgraded.
I've had macs for 20 years and known people using them for longer than that. Unless it was for a business where it was being written off I have never known someone to buy a fully maxed mac from Apple, especially when Apple made it so easy to slide in a new drive or ram like the early intels.
Which legacy port(s) should Apple include? How many ports should a modern machine have? Part of the reason why Apple doesn't do this is to speed adoption of new standards. I got the original iMac back in the day. USB was relatively new and there were few available peripherals that supported it. Apple OTOH, killed several legacy ports (Mac serial, SCSI, ADB, etc) in the iMac in favor of USB. Sure enough, in a very short time, USB peripheral adoption soared. Sure, if you wanted to use an older peripheral, you had to grumble, then suck it up and buy an adapter. Same deal when Apple adopted TB3/USB-C.
If socketed RAM was a dealbreaker for a large number of users, then Apple would likely go back to using it. Perhaps maybe, just maybe, the majority of Apple customers don't find it to be a necessity. If you like to tinker with your hardware and want to swap commodity components in/out, then you should probably start buying from someone other than Apple. That's been the case for a while now.
I would add, as a PSA, that if you want this sort of modular design, then you should consider voting with your wallet and ordering a Framework laptop.
If socketed RAM was a dealbreaker for a large number of users, then Apple would likely go back to using it. Perhaps maybe, just maybe, the majority of Apple customers don't find it to be a necessity. If you like to tinker with your hardware and want to swap commodity components in/out, then you should probably start buying from someone other than Apple. That's been the case for a while now.
I would add, as a PSA, that if you want this sort of modular design, then you should consider voting with your wallet and ordering a Framework laptop.
Last edited:
I could see upgradable RAM being nixed in the short term, but the Right to Repair movement is gaining major traction lately, with both the EU and FTC aiming to ramp up enforcement, and Microsoft just announced this week a more proactive and open minded stance on the issue. So that’s definitely something to keep your eye on. Electronic waste is a growing a problem, and if Apple wants to continue keeping up appearances as a environmentally conscious company, they’re going to have to make some changes in the coming years.
There is no universal definition of what constitutes functionality and your personal take on it is not representative of it all. In my world, for example, a universal USB-C port offers much more functionality than a fixed-function one, because that makes the machine more flexible. Sure, you might need to use adapters to do certain things but it still beats fixed-function ports that cannot do these things at all. And regarding upgrade path... I reckon that most users prefer having better battery life than upgradeable RAM.
Of course only few buy fully maxed out Macs, because only few need configs like that. Please stop equating the amount of RAM and/or storage with functionality or performance. The majority of users out there are perfectly fine with a standard 8/16GB + 256/512GB config and when their machine runs out of steam after couple of years they usually prefer to do a full upgrade (with new CPU/GPU/tech) instead of just getting more RAM.
An initiative that prescribes which components must be used would tantamount to a technical censure and has nothing to do with the Right to Repair. Such an initiative would entirely kill all kind of technological innovation. Both the companies and the users need to have the choice what kind of products they want to build and use, and what kind of tradeoffs they are interested in. Standardized replaceable components are not drawback free, they prevent tight integration, are less power efficient and require more space. Plus the decisions what has to be upgradeable will be entirely arbitrary.
Besides, it is not obvious at all that, say, upgradeable RAM with it's additional components, slots, boards and connector traces, is more environmentally friendly that the tightly coupled space and material-saving approach Apple utilizes with their silicon. Electronic waste will not be prevented by flooding the system with replaceable components, it will be prevented by meticulous recycling, component level repair and reuse of refurbished electronic components (btw. Apple is already doing that one and they got sued because apparently users don't like receiving refurbished components in their repaired computers).
In other words, you won't solve these problems by putting a break in the technological progress and relying on outdated non-scalable solutions, you have to push the innovation to implement better repair and refurbishment technology. As in: make the companies responsible for the electronic waste, outlaw shipping off non-recycled waste to developing countries illegal, commit companies to component-level refurbishing and allow the use of refurbished parts in repaired or even new devices. All while of course opening up access to documentation, certification and part access for independent repair shops as well as improving customer protection laws. You have to reward companies that make high quality, sturdy and functional products, instead of pressing everything into the same level of mediocrity.
Of course, if you want to be environmentally friendly, cost of parts and repairs will increase, but that is not something that an average buyer will or can understand. In this regard I feel that much of the discussion on Right to Repair is deceivingly missing it's mark. I have a feeling that most of the mainstream movements boils down to making long-term ownership cheaper, at the cost of innovation and choice, and frankly, I do not understand why Right to Repair discussions so frequently single out electronic devices when the issue is much much acute with household appliances and various industrial equipment.
Mac Mini:
Even if Apple kept the current 'classic' Mini enclosure the chances for a larger die, "M1X" , version would still likely be soldered on RAM. If Apple is going to push the next MBP 16" up into the Intel low end Mac Pro performance zone , then it is extremely likely they'll just take the "win" and stuff that SoC into the Mini also.
Some potential Mini buyers have grumbled for years for a better GPU in a the Mini. If there is better than 6600/3050 performance then many take it as VRAM which would be soldered down also.
If Apple could squeeze in the 2 die ( and eight RAM package ) solution into the Mini then they wouldn't have to "backslide" on RAM capacity. ( and would have way more cores and way more power GPU than Mini folks even asked for. Again if the perormance leap is large a substanitve number of folks will "excuse" the lack of RAM modularity). Apple could through internal volume at some replaceable NAND storage (e.g. iMac Pro , Mac Pro). SSD has a much higher likelihood of failing than the RAM; especially if exposed to a high workload for years.
The problem is that Apple may be on one of those "shouldn't be" paths for the Mini.
High-End Mac Mini Said to Feature Thinner Design With 'Plexiglass' Top, Magnetic Power Port
Last week, Bloomberg reported that Apple is working on a high-end version of the Mac mini with additional ports and featuring the same Apple silicon...
www.macrumors.com
I am hoping that this isn't the 'high end" Mini, but a revised "low" end (the one on the M1, M2, M3 , etc sequence). A plastic top for better Wi-Fi does exactly nothing for Mini deployed in a racked computational cluster. Even a small 1-2 node system stacked on a desk. Two tone color doesn't do much in a server context either. If Apple can only "afford" one Mini chassis and it tracks the "bigger die" , 'M1X', into this "even thinner" case then Mini is being turned into a "headless" laptop chassis .
The only plausible way out for the Mini is 1) keeping the current case and 2) if the 2 die solution ( much more bigger die usage ) enables DIMMs in some way. Either one of those doesn't come through then the chances are extremely low.
Large screen iMac ( 27-32" )
If Apple puts a design constraint on the iMac that it can only be about as thick as the XDR display ( 1.1" , 2.7 cm ) then it will be back in the "shouldn't be path". Design constraints on the iMac 27" have previously impacted user upgrades of RAM. Apple design objectives wants to hide the cooling vents and RAM door. The iMac was suppose to be a "magical" enclosed shell where can't see the pesky things like vents or access door. So those are hidden from casual view behind the pedestal arm. When the iMac Pro needed a big air flow exit vent, that squeezed out the RAM access door. So Apple has already sold a high ( and low 21.5") iMac with no easy access to the RAM DIMMs. There will be very vocal minority that grumble about it, but a very large fraction will just buy the system if it is good.
The 24" iMac thins out the pedestal arm also. So both the airflow inlets and exits are regulated to the bottom on the case ( out of casual view). If Apple thins out the pedestal arm of the larger iMac in a similar way then again will be "painted into a corner" of where to hide the input/exit airflow vents ( and will cap thermal dissipation capacity ).
If Apple pushes the large screen iMac down to about the same thickness as the MBP 16" system then the "Z height " constraints are basically going to be about the same.
Like the Mini, Apple is probably going to be very tempted to throw the bigger die , M1X , into at least the $1,800-2,100 versions. Once it is the same SoC from the MBP 16", that will drag along the same RAM constraints. Apple probably will not pass up the chance to ramp up the economies of scale on selling more units of this "M1x" die. If they can put it into 4 different Macs variants that will drive the component unit volume up and the costs down.
Probably the only thing that might save the larger screen iMac from being super thinned out is if Apple is going to try to also put the even bigger die(s) in it. The "jade2C" leaks suggests that there could be two full "M1X" dies coupled together into a single multiple chip module. That is dual edge sword. It would be a straightforward way of going from 10C-32G to 20C-64G. Performance and Performance/watt wise that would be huge jump for the iMac. ( Although Apple could bring back a "iMac Pro" label for this larger screen version). However, if they simply just "reuse" the same memory controller subsystem from the "full Jade" ( M1X : 8p-2E-32G ) then probably also deeply coupling this SoC also to the design objectives of the MBP 16". (hence soldered LPDDR4/LPDDR5 ).
The only way to slotted RAM is to get to a different memory controller subsystem. If Apple is just reusing the same one from MBP 14-16" class laptops all the way across Mini and iMac then it is going to be the same RAM implementation constraints. There is a better chance that Apple puts a LPDDR5 along side LPDDDR4 ability into the M1X than they implement something that the high end laptops do not need at all.
the path to slotted RAM is if there is a different memory controller implementation that is independent of the laptop designs. That could come if the two die multiple set up isn't two completely identical dies. They don't really need a second set of thunderbolt controllers, secure enclave , usb port controller , camera processors , and a few other minor things. But if the CPU/GPU cores on one die has one set of latencies to RAM and a substantively different set of latencies to RAM on another die then that would be problematical. They could "hammer" (brute force ) that to work but decent chance that complexity isn't worth it.
One path would be completely decouple from the laptops die and have one die for CPU cores and slotted memory and one die for GPU cores and soldered memory. Each one with their own L3/System cache and shared/"flat" access to both. Different latencies for the "other kind" of core/memory on the "other" die, but as long as homogenous cores/memory that would go much more smoothly.
Either way, they need a high bandwidth , low latency inter-die communication channel so that wouldn't necessarily be any more slower (or faster) with this split.
If "Performance/Watt" is the absolutely number one goal for the dual die SoC though, using the laptop memory controllers will probably be the course of implementation. Lowering the per die TDP allows them to be placed closer together in a multiple chip module which in turn lowers the power. With 3D packaging tech the "interdie" connection switch can be placed below them and the switch and "M1X" dies bonded together for lowest power consumption connection (and also keeps latencies relatively very low) .
Mac Pro
This is the only one where the case/chassis probably won't cause problems. There are leaks of a "half size" mac Pro. Half size would mean loose maybe half of the PCI-e slots but probably wouldn't be a case so thin in one dimension that pragmatically "had to" solder down the RAM packages. As long as the standard PCI-e cards are in the mix for case volume constraints it won't shrink into a "painted into a corner" problem.
I think the need to provision PCI-e slots is going to supersede priority for need for slotted RAM here. So they do need a variant die here. Drop the 3-4 Thunderbolt controllers for perhaps two PCI-e v4 x8 controllers. Swap some of the USB controller space for a x2 PCI-e v4 (or x4 )for provisioning a discrete SATA controller( or a M.2 connection). Basically need to provision more than just one internal storage drive that user can optionally add.
So end up with a die set up like the following ( T -- TB controllers present , P -- PCI controllers present )
[T][P]
[P][T]
They would pragmatically need two different dies in order to implement the connection of each that the "four corners" where all four of these meet. Again 3D package technology the interdie switch is underneath the four dies in that "four corner' area. If the "interdie" connection subcomplex is in the lower right corner of a [T] die then can rotate that die 180 degrees and put it in the lower right of that "4 cluster" above. For a iMac could rotate 90 degrees and get
[T]
[T]
but if didn't want more than 3-4 TB controllers on the iMac then something like
[T][P]
would work for the two-die solution. That's what possibly opens the door for two different RAM controllers . I wouldn't 'bet the farm' on that happening, but it is possible (with a substantial amount of extra work applied to the overall implementations. )
When Apple gets to a 4 die cluster the "redundant" stuff gets a bit excessive. So dropping the 3rd ad 4th camera processor makes a decent amount of sense. Especially in the Mac Pro product zone and they can't provision any PCI-e slots at all. ( Apple already tried that with the MP 2013 and got a supertanker worth of negative feedback on that. They will get some new negative feedback for soldering down the RAM, but likely not as much as that "no PCI0-e slots" move (and the "One and only one internal drive" choice). Over the intermediate term Apple can continue to 'ban' 3rd party GPU cards , but there are a very large number of other kinds of PCI-e cards and even larger number of optional internal drives that a "Mac Pro" should accommodate.
The Mac Pro is highly likely going to get a medium to large Apple GPU implementation. That kind of a GPU needs a highly stable and "know" RAM working storage. The most effective path to that is to solder it down. ( soldered on almost all dGPU cards. Same constraint here since pushing into that same performance space. ). The number of active memory controllers has to be always high and have to spread the workload out over all of them in a "best effort" way. The iGPU is going to bring new contraints. As long as Apple doesn't allow any 3rd party GPU drives ... the Apple GPU is a requirement.
Unified but non homogenous memory is more possible over the longer term. For a 1st generation implementation that is a bit risky. Intel has done this previously . Their 2nd generation Xeon Phi ( Knights Landing ) had HBM memory that could use as cache or memory. The upcoming 4tih Gen Xeon SP ( Sapphire Ridge) processors will eventually have an option of with a layer of HBM memory and Optane memory mixed in with DDR5. The non-homogenous RAM , but "Unified ", issue is complicated but tractable. But not while minimize power utilization and implementation complexity. Intel does these but they also soak up loads more power doing so. Intel also needs to bubble up changes into the OS (and sometimes apps) to handle this also.
Apple may take a stab at it on the 2nd or 3rd generation "Mac Pro class" SoC, but the first generation is likely a stretch to cover that. It is also unlikely that the Mac Pro all by its lonesome is going to drive major adjustments in the macOS kernel and library stack. Even more so on the first generation implementation. ( NOTE: still zero movement on 3rd party GPU drivers. )
Pretty good chance that the "half sized" Mac Pro has soldered on RAM. Even more so if Apple 'refreshes" te larger Intel Mac Pro on W-3300 foundation. ( Apple can 'kick the can' several more years into the future. )
Likely to be mitigated as an issue because also more likely that Apple isn't going to give buyers much of a choice. The base RAM for a workstation will start out at "large" RAM capacity setting and probably just 1-2 other "even larger " options.
If the "M1X" starts off with "lowest RAM configuration of 16GB then a two die solution would probably start off with a 32GB base RAM configuration. The four die solution at 64GB. If the deskstops had LPDDR5 like densities then perhaps 32GB -> 64GB -> 128GB base progressions. If the vast majority of users have "overbought" on their initial purchase then even moderate increases in use cases demands will be already covered.
Similar impact on CPU and GPU cores. Need 128 GPU cores... well CPU and RAM are coming along for the ride. Need 20+ CPU cores ... GPU and RAM are coming along for the ride. Once highly coupled if buy for one then also get the others.
The folks that 'hunt' for the barebones buy-in to a Mac Pro back in the 2006-2010 era probably aren't going to find a configuration in any new system here. To a lower extent also with the iMac. There will be a drop off of those folks buying into Macs, but if Apple pulls more users into MBP 14-16" , Mini , and iMac it may not matter to them. Mac Pro sales dropping off 20% and the others going up by 3%. ( 3% * 25 > 1%* 20% ) . Pretty sure Apple will take that trade off all day long.
As long as Apple won't enable all 3rd party GPU drivers , the super modularity focused folks who used to buy Mac Pro aren't going to buy the system even if there was DIMMs or not. For lots of those folks the GPU is going to be the show stopper. The non socketed CPU is going to be another for what is "left over" after the that first group bolts.