M1 Owners: Is 16GB the same as Intel or has it better memory management?

[armoured]

does it means that a MactIntel with 16GB RAM +GPU 8GB RAM has more available RAM than a MacM1 with "only" 16RAM for both CPU and GPU?

The 'ram is ram' point applies strictly only to system ram. If ram is on a dedicated GPU, that's distinct and used differently.

So the answer (in my view) is that the 16gb + 8gb GPU should have a bit more system ram available compared to the straight 16gb / no dGPU. But the key thing to keep in mind is it absolutely is NOT a one-for-one substitution - because for the most part the system can't use dGPU ram as a direct substitute for system ram. (And as leman put it above, because they have different uses, there will often actually be duplication of the two types and copying/transfer between the two). That might mean an 'extra' [some number] of system memory available/free, but it will always be less than the '8gb' amount of dGPU ram (how much will depend a lot on what you're doing - perhaps a big delta for gaming, not very much for general computing).

In other words: You can't just add different types of ram together, even if they're both a type of ram. 8x + 8y does not just add directly to equal 16x. 16 + 8 ram/dGPUram here does not just add up to 24.

Similarly your system can use the SSD as a type of memory (call it logical memory if you want, or swap memory), but it doesn't mean that you can say your system has 18gb ram + 128 gb ssd = 146 gb 'ram.'

 

[PLH]

> M1 Owners: Is 16GB the same as Intel or has it better memory management?

I'm seeing better, even thought the memory warnings come up sooner, they look more like false alarms than my i9 macbook pro where if I see that warning, running applications are likely to fail. For the same work, the M1 is taking up less RAM, I think.

 

[zevrix]

THIS is the kind of success story that keeps me optimistic.

I am going to try loading all my apps onto the new M1 that usually ran a bit above 32+ GB and see how it does under 16GB.

I wonder what was your experience eventually if I may ask? I think you didn't follow up in the end... I'm in a similar situation.

 

[NJRonbo]

I wonder what was your experience eventually if I may ask? I think you didn't follow up in the end... I'm in a similar situation.

I ended up returning the M1 Mini. It could not handle the load that I was putting on my i9 16" MBP with 64GB

There were huge slowdowns with CPU overloading at bootup.

Mind you, the way I use my computer is pretty unique. I have two dozen programs running at startup. I kind of knew that the M1 Mini would not be able to handle it but I had to find out for myself.

 

[pshufd]

My comparison is a 2015 MacBook Pro 15 with 16 GB of RAM and dGPU with 2 GB RAM and a 16 GB M1 Mac mini. I would say that the two behave similarly on RAM. RAM use is usually 10-12 GB and the rest as file cache. There is sometimes some SWAP but it's under 100 MB about 95% of the time. There is no trashing. Sometimes I restart Firefox which drops RAM use several GB and then it creeps back up. I would feel more comfortable with 20 or 24 GB but those aren't options. If I had a spare Mac that doesn't run really hot, then I'd consider using two of them for 32 GB of RAM - the stuff I do on Mac doesn't require a strong CPU.

 

[JMacHack]

I think the typical answer is something like this:

RAM is RAM. If something uses a gig it will use a gig. There’s no if ands or buts about it.

But M1 Macs seems to be more “graceful” when maxing out the RAM you have. They don’t always fall flat on their face like other computers. Attested to by owners and reviewers.

This argument has gone around in circles for awhile now, and my understanding may be wrong.

 

[leman]

I think the typical answer is something like this:

RAM is RAM. If something uses a gig it will use a gig. There’s no if ands or buts about it.

But M1 Macs seems to be more “graceful” when maxing out the RAM you have. They don’t always fall flat on their face like other computers. Attested to by owners and reviewers.

This argument has gone around in circles for awhile now, and my understanding may be wrong.

I think your explanation is spot on. If your workflow requires a lot of active RAM (e.g. a single application with very large data sets, multiple simultaneously active VM's or large databases etc.) then it doesn't matter much whether we are talking about an Intel or an M1 based Mac — you will need a lot of RAM and the 16GB M1 ceiling might be not enough for you. But if you are multitasking with multiple demanding applications which do not have to be active at the same time (i.e. only the foreground app is doing actual work, the rest is idling), you might get away with having less RAM in your M1 machine as Apple Silicon seems to be better at juggling physical RAM between processes that need it. Empirical tests show that M1 remains smooth and responsive in memory-starved multitasking scenarios where Intel machines will bog down. The crucial requirement is that the active applications still fit into the physical RAM.

 

[NJRonbo]

I think the typical answer is something like this:

RAM is RAM. If something uses a gig it will use a gig. There’s no if ands or buts about it.

But M1 Macs seems to be more “graceful” when maxing out the RAM you have. They don’t always fall flat on their face like other computers. Attested to by owners and reviewers.

This argument has gone around in circles for awhile now, and my understanding may be wrong.

That has been the Holy Grail of questions to be answered. Apple is certainly being secretive about it and depending on whom you talk to, you can never get the same answer concerning RAM on the M1 machines.

I think what you just posted above puts it out there best: RAM IS RAM but the silicone Macs and their improved memory management seem to be more "graceful" with stretching its capabilities.

 

[bobcomer]

I think the typical answer is something like this:

RAM is RAM. If something uses a gig it will use a gig. There’s no if ands or buts about it.

But M1 Macs seems to be more “graceful” when maxing out the RAM you have. They don’t always fall flat on their face like other computers. Attested to by owners and reviewers.

This argument has gone around in circles for awhile now, and my understanding may be wrong.

Well said! Bytes are bytes. And yes, MacOS handles tight RAM situations better than Windows, but that only gets you so far -- it doesn't make up for RAM you don't have...

 

[southerndoc]

For the vast majority of users, 16GB is fine. I frequently have all Microsoft Office apps, Google Chrome (a memory hog at times), Adobe Acrobat (non-AS as of now), and a few other work-related apps open at the same time and have had zero memory issues. Usually hovers around 12GB. I don't think I've ever seen it use any more. Like previous posters have mentioned, cache size is usually pretty low.

 

[Apple Knowledge Navigator]

I ended up returning the M1 Mini. It could not handle the load that I was putting on my i9 16" MBP with 64GB

You’re surprised?

 

[zevrix]

I ended up returning the M1 Mini. It could not handle the load that I was putting on my i9 16" MBP with 64GB

There were huge slowdowns with CPU overloading at bootup.

Mind you, the way I use my computer is pretty unique. I have two dozen programs running at startup. I kind of knew that the M1 Mini would not be able to handle it but I had to find out for myself.

thanks that's good to know. i suspect it would be the same for me.

 

[mi7chy]

My impression is MBA M1 has slow storage I/O so it relies heavily on RAM disk cache which puts memory pressure on apps. So, 8GB on M1 is more like 4GB that's why it's always on sale since no one wants it and 16GB more like 10GB.

 

[pshufd]

My impression is MBA M1 has slow storage I/O so it relies heavily on RAM disk cache which puts memory pressure on apps. So, 8GB on M1 is more like 4GB that's why it's always on sale since no one wants it and 16GB more like 10GB.

Any modern operating system does disk cache with free RAM.

My Windows system has 16.5 GB in use and 29.9 GB cached. It will get up to about 60 GB cached after running for a while.

 

[NJRonbo]

thanks that's good to know. i suspect it would be the same for me.

Very strong rumors that the Mini Pro will be introduced in two months alongside the two new MacBook Pros.

I have also been reading that a few people saw “hang-ups” with the current M1 Mini which may be inherent in the device itself rather than the RAM. The jury is still out...

 

[pshufd]

Very strong rumors that the Mini Pro will be introduced in two months alongside the two new MacBook Pros.

I have also been reading that a few people saw “hang-ups” with the current M1 Mini which may be inherent in the device itself rather than the RAM. The jury is still out...

I think that most people would be happy to have the ports and expansion options of the 2018 Intel Mac mini. And I think that Apple prefers to put in something where customers don't lose functionality in replacement models.

 

[thingstoponder]

does it means that a MactIntel with 16GB RAM +GPU 8GB RAM has more available RAM than a MacM1 with "only" 16RAM for both CPU and GPU?

Depends on the Mac. Laptops still use the iGPU. Desktops don’t.

 

[thingstoponder]

Very strong rumors that the Mini Pro will be introduced in two months alongside the two new MacBook Pros.

I have also been reading that a few people saw “hang-ups” with the current M1 Mini which may be inherent in the device itself rather than the RAM. The jury is still out...

What rumors are those? All we’ve heard from credible sources is from Bloomberg that the device is in the works.

 

[pshufd]

What rumors are those? All we’ve heard from credible sources is from Bloomberg that the device is in the works.

You can buy Intel/AMD desktops with more computing capabilities than M1 macs. All you need is a big enough workload to demonstrate that.

My production stuff is run on a big Windows desktop. It wouldn't run at all on the M1 mini because the M1 mini doesn't support 4x4k but it wouldn't support my workload and the other stuff that I do on the M1 anyways. You could get a 5950X or ThreadRipper which would easily outclass an M1 in CPU. You could, of course, get a number of GPUs that would outclass the M1. The M1 runs about the performance of a GTX 1050 ti which is a GPU from 2016.

The Dell XPS 15 offers the GTX 1650 Ti which is also an old video card but offers twice the performance of the M1. The Dell XPS 17 offers the RTX 2060 which offers three times the performance of the M1. And you can go nuts with the Dell Precision 7750 with the available RTX 2060 which is almost five times the performance.

 

[Joelist]

What Apple Silicon (NOT ARM) has is DIFFERENT memory management. Because of the huge caches and HIGH ILP (due in part to the 8 ALUs per core) it has different priorities for what stays in RAM, what order things can swap in and out and a host of other elements.

 

[dogslobber]

Well here's an example of memory in use on an 8GB MBA running a 1.5GB Linux instance. It's remarkable how low it manages to keep swap.

 

[driven01]

I'm seeing something similar. My primary work computer (2019 MBP i9 32GB RAM) used for Xcode, video editing, a bunch of Safari and Chrome windows, Windows VM, Linux VM, etc. In this configuration I typically have moderate memory pressure, and often 2GB allocated to the SWAP file. Activity monitor reports 27GB used on a consistent basis.

For s_its and giggles, I borrowed my wife's M2 Air (16GB RAM) and ran the same work load for a few days on it. Besides everything feeling notably faster, I also noted that my RAM usage never topped 14GB, and the swap file never got above 0.

So I'm all for this bytes == bytes thing when it comes to memory. HOWEVER: Something is different. Either the universal binaries don't allocate as much on the Apple Silicon architecture, or the memory compression works far more efficiently. All I can say is that if I tried the same thing on the Intel with 16GB I'd be having a bad day on the performance side of things.

 

[NJRonbo]

I'm seeing something similar. My primary work computer (2019 MBP i9 32GB RAM) used for Xcode, video editing, a bunch of Safari and Chrome windows, Windows VM, Linux VM, etc. In this configuration I typically have moderate memory pressure, and often 2GB allocated to the SWAP file. Activity monitor reports 27GB used on a consistent basis.

For s_its and giggles, I borrowed my wife's M2 Air (16GB RAM) and ran the same work load for a few days on it. Besides everything feeling notably faster, I also noted that my RAM usage never topped 14GB, and the swap file never got above 0.

So I'm all for this bytes == bytes thing when it comes to memory. HOWEVER: Something is different. Either the universal binaries don't allocate as much on the Apple Silicon architecture, or the memory compression works far more efficiently. All I can say is that if I tried the same thing on the Intel with 16GB I'd be having a bad day on the performance side of things.

Just bought an M1 Studio

I put 64GB memory on it because I can push towards that usage when I was on an Intel Mac.

The 64GB was overkill. I'm only pushing 30GB of memory tops.

It does seem to be a completely different animal on silicon. Memory optimization is incredible. Of course, Apple is never going to tell you to go with less memory as they want to sell you on upgrading.

 

[pshufd]

Just bought an M1 Studio

I put 64GB memory on it because I can push towards that usage when I was on an Intel Mac.

The 64GB was overkill. I'm only pushing 30GB of memory tops.

It does seem to be a completely different animal on silicon. Memory optimization is incredible. Of course, Apple is never going to tell you to go with less memory as they want to sell you on upgrading.

I have the base Studio and regretted not getting 64, but, in actual use, I don't need it.

 

[arefbe]

Better. I had a 2019 21" iMac with 32GB of RAM, and it was constantly swapping. I now have a 16GB M1 iMac, which I've had since last August. And with the same workload, it has not swapped once.