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.
Why is 16GB on Big Sur like 8GB on Windows 10 or 4GB on Linux?
- Thread starter mi7chy
- Start date
- Sort by reaction score
So because macOS has a different RAM governing strategy, things “seem” like too much to you. Come on.
Is this really the path your taking?
Complaining that macOS takes more memory than other operating systems without stating actual performance issues you‘re seeing?
What is the real world impact of what your stating.
I had a 16 inch MacBook Pro with 16GB of memory, and my 13 inch MacBook Pro (Apple Silicon) with 16GBs runs circles around the Intel model. While my CTO order was being built, I used my wife’s Macbook Air with 8GB (also Apple Silicon) and generally speaking it also ran better than my retired 16 inch Intel MacBook Pro.
Other than screenshots, with comparisons of how things look, what is the real world implication that you are driving towards?
I think it’s worth saying that a lot of these conversations are extremely loaded for a lot of reasons.
Because the memory is shared between all the resources, and it’s on package it seems very clear that Apple has been fairly aggressive to use memory where needed, and page out to increase “user experience”. I’m not saying that means 8GB is the new 16GB. But I am saying that the shared memory, and the system being so close together does seem to have a lot of benefits we haven’t seen before from a true SOC. It’s not magic, it’s the results of not having to transfer pages of memory to get something executed. How far does it go? REALLY hard to say, especially since Apple doesn’t talk about these things.
It’s really not magic, it’s seizing the opportunity of everything being close to what’s being processed. The additional benefit of the thermal headroom also helps with the “feeling” that 8GB on Apple Silicon is like 16GB on Intel.
Not saying it’s right, or correct. M1 won’t cure cancer. But it could very well be the closest thing we’ve seen to penicillin in the last 20 years.
Right, I’m not saying there’s no improvements and optimizations here but anyone saying 8GB on M1 is the same as 16GB on Intel is absolutely wrong. And yes, people have made this claim.
People have made the claim.
I’ll tell you what I’ve observed, though I can’t share screenshots it is an observation.
macOS on Apple Silicon is more likely to page things to SSD. There isn’t a noticeable performance hit when this happens, but I can tell you that at one point I was actively using effectively 20GB of memory. Without knowing on how things are implemented, it’s hard to say why there is no perception of this happening. But I can tell you that I’ve done it, with an oversized Windows VM running, lots of Mac apps running, while on Teams and there wasn’t any real performance issues.
I don't know about you, but this sounds like good news to me. Now all of your RAM is used, and none goes to waste.
That is what is happening actually. Applications in Windows run faster.
I switched from Windows —> Linux in the past to gain more performance. And then switched to Mac OS X as I was tired of maintaining Linux myself for personal use.
Snow Leopard was faster than Windows but ever since then, OS X has been bloated and more “iOS like”, up to the point that Windows even became faster.
Used by what? RAM should only be used when RAM is needed by something. I am not sure what is happening with OS X these days, but it is not particular fast anymore, like how it used to be in the past.
I’ve only ever seen OP complain about M1 macs on these forums ever since they came out. I honestly don’t understand why they even own a Mac if all they do is complain about how Linux and AMD is so much better.
My GF has Big Sur on my old 2013 MBP with 4 GB RAM, it runs fine.
Why? RAM doesn't use more energy nor any additional processing power when filled with useful data.
If you can tell with a high degree of certainty that you may need some pages in the RAM soon, you can just leave them there (if used recently) or preload them to avoid the expensive page faults later. There are basically no downsides to this approach, and it reduces latency.
Yes, it makes things harder to reason about by self-taught computer science connoiseurs who read a book on von Neumann architecture and thought that that was it, no progress to be done here.
High RAM usage points to sloppy programming in general. If software A does something with 4 times more RAM than a similar software (which is even more flexible as in the case of Linux), something is wrong.
And I really do not see the "performance gains" with this high RAM usage as it seems that I get better performance out of Windows on my Macs nowadays with the same software.
Last edited:
Buddy, you don't have to worry about my "level". I've been writing system software for decades (with focus on low-level optimizations), I have a degree in CS, I work and have an academic record in a CS-related field, not to mention that I teach this stuff at a major university. You instead are all over the place, grabbing random bits of facts and fiction here and there and twisting them to your liking, without showing much understanding for technicalities. I'm happy for you that you managed to catch a thing or two about working with Linux, but that doesn't make you an expert in computers.
This myth has been discussed extensively and laid to rest multiple times. Why are you still going on about it? No, 8GB on M1 obviously does not equal 16GB on other platforms. The mere suggestions laughable. At the same time, yes, there is ample evidence that M1 Macs retain their high performance and responsiveness even in memory-starved scenarios, unlike their x86 counterparts, which means that an average user can probably get much more mileage out of a specific RAM configurations on Apple Silicon.
And none of these things have anything to do with average system RAM utilization. MacOS kernel has a very different approach to memory management than Linux or Windows, and it will obviously result in different figures. Not to mention the system framework differences (Cocoa is a high level of abstraction, and it does come with a significant memory overhead). It doesn't matter what your RAM counters say unless you are trying to track a specific issue. Relying on these counters to make points about system efficiency is a pointless exercise, as these are not useful facts. "Memory used" is not a statement about performance. It's not like you get more work done or save electricity by occupying less RAM or anything.
Congratulations! You have discovered that disabling all remotely useful system services will make the system use less resources! Now you know why your Linux distro is so efficient with RAM — because it doesn't do anything beyond the bare minimum. It's easy to be fast and efficient if you only ship a bare-bones OS with as little utility as possible.
MacOS is an opinionated system. It comes with ton of utility, whether you want it or not. If that is not to your liking, stick to Linux. Tryin to tweak macOS services will inevitably lead to issues since the system was not designed to be tweaked in that fashion.
Last edited:
If consultant A delivers me a piece of software that does the same thing with 1 GB of RAM while consultant B delivers the same software but it requires 8 GB of RAM while not being even faster, I will go with consultant A his solution.
I don't care about what they teach in computer science.
You know what? I agree with you! But that's where the problem is: OS-level RAM usage statistics tell you very little about how much RAM is actually required. If you compare different macOS systems that only differ in the amount of installed RAM, you will find out that RAM usage is nearly always on the higher end. But each of these systems will perform well. On macOS, RAM utilization scales with the amount of available RAM.
Refusing to learn from people who are literally experts in these things better does not necessarily speak to your benefit.
That’s not what is happening.
Consultant A gives you software that uses 1GB of RAM and nothing more.
Consultant B has also given you software requiring 1GB of RAM. Except this time when it sees there’s unused ram available it starts to preload data it will need to reduce latency. If something comes along that needs it. It will dump its unnecessary stuff.
Exactly what I was talking about.
High RAM usage points to sloppy programming in general. Yes, usually. Except when the high-memory-footprint algorithm or data structures are vastly more performant than the alternative. But it's mostly true.
We want our processes' memory footprints to be as small as possible in order to fit as many of them as possible into physical RAM. A high-memory-footprint process negatively affects other processes and memory-starves them. Processes COMPETE for RAM.
OSes on the other hand don't compete for RAM. They manage it. They own it. They own all the processes and every last bit of memory on the machine. The OS itself has a memory footprint of its own (the kernel, drivers, processes integral to the system and user experience) and let's be honest, macOS is not the tiniest OS around, but there is no one to memory starve except the OS. The OS doesn't owe anyone anything. Its only responsibility is to provide fast memory to processes (in layman's terms), and how it does that is none of the user's (or processes') business.
OSes use different pagers with different strategies, there's virtual memory, there's memory compression, there are swapfiles, page caches and more. There is literally no "RAM usage" metric anymore that would be comparable between different OSes with different memory management.
macOS will show you some values in Activity Monitor, but those values are to be interpreted by professionals, in the context of macOS, and ideally in the context of "macOS running on a X GB RAM machine".
My MacBook has 16 GB of RAM and its memory usage is currently over 8 gigabytes.
Does it mean that my GF's older 4 GB MacBook would die in the same situation? Not at all, I only have Safari open with three tabs, Sublime Text and Skype, nothing else. The rest is macOS knowing we're on a 16 GB machine used as a software development workhorse and managing the memory accordingly.
Bro, your Linux is too bloated, mine only uses around 73 MB RAM ?
Great summary! To add to this, macOS has some unique features that makes reasoning about RAM usage even more difficult. For example, purgeable memory, which allows an application to allocate large chunks of RAM for caching purposes, with the caveat that this memory can be freed at any time by the OS. If an application uses this feature aggressively, it may seems like it's gobbling up all the available RAM, but that RAM is still immediately available for any new application. For example, right now, about 30% of all allocated RAM on my system is marked as purgeable.
