Thanks so much for this post. I had been waiting/hoping for a comment like this.
This definitely leads me to the $999 ($899 education) MBA.
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.
8GB RAM is a disappointment?
- Thread starter MacBH928
- Start date
- Sort by reaction score
This definitely leads me to the $999 ($899 education) MBA.
I was just looking at a GitHub project meant for ARM and supporting macOS. By default, compiling the project required more than 8GB of RAM, otherwise the code simply would not compile. Content creation, part of the M1 marketing, is another good use case.
IMO a lot of the flak comes from the fact that the design of these is imbalanced.
It’s like selling a Tesla that can do 0-60 in 3.5 seconds and then putting the same tires on it that one would find on a Civic, and also fixing the tires permanently to the chassis. The car is capable of so much, but then hampered by something that seems superficial. Sure, if that car is sold at the price of a Civic, and marketed to a Civic audience, it’s awesome for them. But it’s just as usable for a Tesla audience if all they did was either make the tires upgradable or sell a Z-rated option.
So there’s a whole group of people that see a Tesla (or Corvette, whatever) and then look at the tires and go “wtf”.
Apple trained all of us over the years to recognize good design. And from a product design standpoint, this is weird design because the overall package is imbalanced.
IMO a lot of the flak comes from the fact that the design of these is imbalanced.
It’s like selling a Tesla that can do 0-60 in 3.5 seconds and then putting the same tires on it that one would find on a Civic, and also fixing the tires permanently to the chassis. The car is capable of so much, but then hampered by something that seems superficial. Sure, if that car is sold at the price of a Civic, and marketed to a Civic audience, it’s awesome for them. But it’s just as usable for a Tesla audience if all they did was either make the tires upgradable or sell a Z-rated option.
So there’s a whole group of people that see a Tesla (or Corvette, whatever) and then look at the tires and go “wtf”.
Apple trained all of us over the years to recognize good design. And from a product design standpoint, this is weird design because the overall package is imbalanced.
I'm not sure I follow your analogy here...
Are you saying a Tesla has the wheels you'd see on a Civic and holy cow some how it miraculously drives just as well as if it had the wheels you'd normally expect...
or
Are you saying that the wheels are the wrong size and thus even though it's a Tesla, with the wrong wheels then it might as well be a civic?
This does not make sense at all. We have already learned what this can do with 8gb ram from various videos and benchmarks as well as real world reviews from users.
If you think 16gb is not enough for your use case, then there will be devices next year. But there will be many many users who will be fine with a base MBA/MBP/Mini. Nothing in their approach shouts imbalanced.
I am sure there are many cases out there in the wild, but I have not seen a specific task where a 16 inch 64GB Ram MBP does better than the base M1 config (that does not involve multiple ports). And once we find that task, I'd bet it will be lousy, noisy fan chugging it.
Now do I want a 64gb ram souped up laptop, or a laptop that might barely lose out on a specific Pro task but is quiet and has 15+ hours of battery life?
Last edited:
16GB is fine for my use case, which is light 3D content creation away from a desk, since more serious work would be done on a workstation.
But my point is that these machines can perform computational tasks as well as $2500-$3000 MBPs, so they’re getting the attention of audiences that use a higher-end system. Those audiences would get slightly improved performance but with a much thinner & lighter laptop, insanely better thermal emissions, and insanely better battery life, for half the price.
Except they can’t get 32GB of RAM, which they need for virtual machines, or working with 6k video, or doing simulation work, or compiling moderate to large codebases as my GitHub example demonstrates (Embree for aarch64).
So in the end, they can’t get their machine at half price because of what seems like a superficial limitation. And I find that frustration very justified.
But my point is that these machines can perform computational tasks as well as $2500-$3000 MBPs, so they’re getting the attention of audiences that use a higher-end system. Those audiences would get slightly improved performance but with a much thinner & lighter laptop, insanely better thermal emissions, and insanely better battery life, for half the price.
Except they can’t get 32GB of RAM, which they need for virtual machines, or working with 6k video, or doing simulation work, or compiling moderate to large codebases as my GitHub example demonstrates (Embree for aarch64).
So in the end, they can’t get their machine at half price because of what seems like a superficial limitation. And I find that frustration very justified.
There must be sth wrong with your Mac. How many tasks showed in your activity monitor? Usually Chrome could cause that issue but not the Safari.
Been running loads of simultaneous apps on (more than I actually do in actual use case) and computer is always hovering around 6gb consumed. Played 3D games as well and it never pegged the memory.
Was this the case all the way up until OS 9? I just don't remember people adjusting their RAM per application, neither on Windows 95. I thought all operating systems manage RAM just as MacOS does today.
He's hilarous.
People are doing 8K raw render on the M1 8GB and he's complaining about not having 64GB needed for safari and mails.
Yes, it was with Mac OS back then. I did it myself in System 7 to Mac OS 9.
Sorry for the bad quality. Couldn‘t find any better.
EDIT: Inserted picture.
Last edited:
Last edited:
Not necessarily. It's possible this is what @MisterAndrew sees - and misconstrues the RAM used as RAM needed. macOS will make use of all the RAM it can, even though it really doesn't need it. This tends to reinforce folks belief they made the "right" choice - though it's the equivalent of giving a teenage daughter a bigger bedroom; she'll strew her stuff across the floor of any size room, giving the impression of needing more space than is truly needed.
Yet we will see people stomp up and down saying they need 64Gb to even open the lid.
I am an EE for a a lot of time , but I do like my SW side projects , this claim of yours is a first I hear :
" By default, compiling the project required more than 8GB of RAM, otherwise the code simply would not compile"
Can you give a concrete example ? I am intersted in it from a technical side of thins , as I can't understand why would the code won't be able to compile , its not like we need the entire code in the DRAM......
I agree, a majority would be perfectly fine with 8gb, I myself would probably fine as well, but I prefer to always max out the non upgradable ram, only because I keep my machine for a long time, and I don't use them for work, other than replying to a few emails. For my workstations, it's a different matter, they need much more ram, as well as dedicated graphics.
It all comes down to personal preference, and not necessarily what is really needed.
"Seems" being a key term here. They may be saving money by not designing the M1 to accept more than 16GB. Seems likely to me that if they could have just as easily put more in, they would have offered that possibility for the Mini at least.
Except...
With the two-port MBP and MBA, the 16GB max matches the Intel systems they replaced. The Intel MBA and two-port MBP13 are no longer part of the current product line. They're defunct, even if leftovers may be around at third party retailers.
Not so the Intel mini. It's still available. Thus it seems the M1 mini is a new branch in the mini family. Fewer ports, lesser configuration options.
IMHO it is likely that Apple will replace the higher tier four-port MBP13 and MBP16 at a later date - with similar configuration options as they posses now - and likely will do the same with the legacy Intel mini
My suggestion is don't make assessments of why one tier is the way it is when we've only seen the first stage of what's clearly going to be a multistage transition.
Right. But I still suspect that if they could have just as easily offered an M1 Mini with more than 16GB, they would have.
Perhaps one reason why the M1 laptops do not have 4 ports is not due to a limitation in the M1 chip but in the design of the computers. They changed some things on the inside but maybe they weren’t able to fit access to 2 more ports in there.
But that’s a totally uneducated guess and speculation.
I suspect the chips / SOC for the higher tier systems aren't due to be ready yet. Wanting to release an M1 option for desktop / lower budget folks they established a lower-tier model of the mini.
The prototype mono soc designs likely were set a year ago. Hard to see changes happening
Yeah if the system has lots of available RAM it will use it. It sees candy and can't help itself. It's not a bad thing though. Even though it showed 14 GB of RAM being used, 50 GB was still free. My MacBook Pro with 16 GB RAM is using about 7 GB with the same apps open.
More RAM is better. If you are doing things like scientific tasks or virtualization, you need lots of RAM. Therefore it's ideal to have a system you can add more RAM to as needed. For the M1 Macs that's not possible, so you should max it out at the time of order.
I looked at the embree-aarch64 repo. Where exactly does it say you need more than 16gb to compile? They tell you to use clang or reduce NJOBS on gcc or you might get out-of-memory error on 16gb ram. That is not the same as „does not compile on 16gb“
The thing Gruber and David Smith are talking about has nothing to do with improving RAM usage efficiency. It's an improvement to the performance of reference counting, the method by which Objective-C, Swift, and the rest of Apple's platform track how many active references there are to each object. Once the number of references reaches zero, that means the object can be deallocated and its memory reclaimed for other purposes.
Since it's very common for code to increment and decrement the reference counter for each object, speeding these operations up by 4.6x is a big deal. But that performance improvement doesn't alter the life cycle of the object at all, so there is no change to memory usage.