Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Nothing. Ignore anyone who tells you Apple Silicon Macs need less RAM, it's a dumb myth that was spread early on and has been annoyingly persistent.
Maybe, but my M1-Pro MBP with 16GB doesn't exhibit more yellow memory pressure while editing photos in Lightroom than my old MBP did with 32GB. And thats with 40MP files now compared to 24MP files back then.

The actual deal with M# Unified Memory is that the new Macs are more efficient with RAM and SSD swap to where they behave the same as pre-M# Macs with more dedicated RAM. That being said, if you are doing really data intensive work where you need all the performance you can get, there is no replacement for more memory.
That said, there's also a lot of people who blindly followed advice to get lots of RAM, and often they didn't actually need it. So, the key question is: do you know for sure you need 40GB?
Most people overestimate how much RAM they need. What they can do is open Activity Monitor, switch to Memory, and observe the Memory Pressure color while doing regular stuff like browsing. some people have a bunch of tabs open, and doing your most intensive work. If MP is only occasionally yellow then they will be fine with 24GB RAM; however, if their MP is regularly yellow then they need to stump up for more RAM.
 
Last edited:
do you know for sure you need 40GB?
That's a good point. For 30 yrs I've been hearing 'more ram is better'. It's hard to know what I need.
That said, I'm using my 2017 iMac to clean up (w/ Silverfast) 50 yr old time degraded slides and as I mark the undetected spots and my fan starts running faster & faster (where in normal use it rarely can be heard) and the marking process takes longer and longer, I've started wondering about upgrading to an M4 mini.
In normal use my i7 does just fine. I pulled up the memory pressure (which until last nite I had never heard of it) and it seems ok, my 40GB is 22/17 apps/cache. So maybe with the 16GB for a basic M4 mini with it's faster processors and benchmark speed I'd never notice less RAM?
 
Maybe, but my M1-Pro MBP with 16GB doesn't exhibit more yellow memory pressure while editing photos in Lightroom than my old MBP did with 32GB. And thats with 40MP files now compared to 24MP files back then.
Thanks for this comment. Since I just discovered mem pressure and the graph is near the bottom and green, I was not aware that it would turn yellow. I'm going to start editing some slide images shortly and it will be interesting to see what the graph does.
 
Maybe, but my M1-Pro MBP with 16GB doesn't exhibit more yellow memory pressure while editing photos in Lightroom than my old MBP did with 32GB. And thats with 40MP files now compared to 24MP files back then.

The actual deal with M# Unified Memory is that the new Macs are more efficient with RAM and SSD swap to where they behave the same as pre-M# Macs with more dedicated RAM.
Sorry, but from an engineering point of view there simply aren't any significant RAM efficiencies exclusive to Apple Silicon. macOS and its apps are compiled from the same source code, and therefore use exactly the same logic in how they allocate and use memory. Compiled executable code is about the same size (arm64 tends to be slightly smaller, but not by enough to matter). Data is the same size, both on disk and in RAM.

Swapping is not more "efficient" on Arm Macs. It is faster, but that's only because the SSDs got faster, on average. This makes light to medium swapping less noticeable, but swapping is still very bad for performance. If you want to avoid it, when doing the same tasks you'll still need about the same amount of RAM.

On your LR observations - what would make them far more meaningful is finding out whether you'd also see yellow memory pressure on a 16GB Intel Mac. I bet that would be the case. Sophisticated 'pro' apps designed to smoothly handle lots of data often try to avoid driving your computer into swapping, so they find proactive ways to self-reduce their use of memory on computers with less RAM.

An example of the kind of thing that's possible: managing undo data. Photo editing apps often have 'infinite' undo stacks, and every entry in the stack can consume a decent chunk of RAM since it might have a bunch of pixel data in it. As a programmer you can put logic into your app to write old undo stack entries to a scratch file on disk whenever free RAM starts to get too low, then bring them back into RAM if they're ever needed again (which they only are if the user presses cmd-Z a bunch of times).
 
That's a good point. For 30 yrs I've been hearing 'more ram is better'. It's hard to know what I need.
That said, I'm using my 2017 iMac to clean up (w/ Silverfast) 50 yr old time degraded slides and as I mark the undetected spots and my fan starts running faster & faster (where in normal use it rarely can be heard) and the marking process takes longer and longer, I've started wondering about upgrading to an M4 mini.
In normal use my i7 does just fine. I pulled up the memory pressure (which until last nite I had never heard of it) and it seems ok, my 40GB is 22/17 apps/cache. So maybe with the 16GB for a basic M4 mini with it's faster processors and benchmark speed I'd never notice less RAM?
It's possible.

One thing to try first, if you're otherwise still happy with the iMac, is to clean some dust out. I had a 2019 i9 iMac, and after only a couple years there was a ton of lint packed into the vents in its 'chin'. This caused it to start running its fans pretty hard under even light load. This cleared up immediately after I pulled the dustballs out with a vacuum cleaner.

If you take that style of iMac apart and look carefully at how air flows through, there's a finer mesh screen inside the 'chin' about level with the bottom of the LCD. Most of the chin is just empty air space, so dust packs up below the mesh screen. Fortunately you don't need to take it apart to get the dust out of there (in fact, taking it apart wouldn't even help you).

The difficulty of regular users figuring out how much RAM is enough for them is very real! In your place, if I wanted to upgrade to a M4 Mini, I'd tale advantage of Apple's 2-week no questions asked return policy. Make sure you have everything set up to do some intense testing as soon as you get it, and if you see evidence of swapping, return it and use the credit to buy one with more RAM.
 
One thing to try first, if you're otherwise still happy with the iMac, is to clean some dust out.
This is definitely something I was thinking about but I thought I'd have to open it up and I didn't want to do that just for dust. I'll give the vacuum a try. Thanks!

I have in the back of my mind to convert the i7 to a 27" monitor as described in a thread elsewhere in this forum. What would hold me back is my age, late 70's, and the difficulty I now have with manipulating small things with my hands plus eyesight that is going downhill quickly. And that is when I'd be looking for dust bunnies. But I'll try the vacuum. It's certainly likely that after 7 yrs it is full of crud, but I figured I'd have to open it up so I have been postponing that.

I'd tale advantage of Apple's 2-week no questions asked return policy. Make sure you have everything set up to do some intense testing as soon as you get it, and if you see evidence of swapping, return it and use the credit to buy one with more RAM.
OK, this occurred to me but I didn't consider turning it in and upgrading the RAM. Great ideas you've given me, thanks for this help.
 
clean some dust out
That worked quite well. Now, I can actually see the screen if I squint w/ a bright light. And I can see beyond the screen in some places, which appear to have even more dust. The fan is not coming on as much as before.
 
  • Like
Reactions: mr_roboto
Well, to unpack a little, GB/s is gigabytes per second - how fast can data move between two locations? That high-end AMD GPU for 2019 Mac Pros I mentioned can access data in its local memory at 512 GB/s, but if the GPU has to read or write anything in main memory, it goes through a 16 GB/s PCIe connection - 32x slower.

Point being, that PCIe communications bottleneck means whatever memory is on the GPU side isn't terribly useful to the CPU, and vice versa. And since the GPU typically has significantly less memory than the CPU, often whatever's currently stored in GPU memory is just a copy of something in CPU memory, so that the system can upload different data to the GPU as-needed. (Sometimes things can be moved so that they're exclusively stored in GPU memory, of course, and that's where you'd get the effect of extending the total memory capacity of the system.)

What are the sizes of these files? Not necessarily the on-disk size, but the pixel dimensions and color depth?
Nice of you to answer and aplogies for taking half a year to respond. Apple refused to allow me to upgrade Mac, and I lost the will to keep trying to figure things out. Thanks for useful info - I think I understand it!

Size? Ranges from ±12000 x 8000 px, 550MB, 16 bit (average) to 106,299 x 46,684 px 8 bit, 17GB file (biggest) that was slow of course on 2018 Intel, but I could use all features needed in Photoshop. On the M3 Pro Max 64GB / 1TB, Gradient Tool (which worked in real time on Intel) is so slow it’s unusable; takes 4-6 seconds to respond to mouse, then jumps somewhere unpredictable. Takes a good 4 mins to save the file to SSD too. Oddly (to me anyway!) memory pressure green, so then why so slow? Maybe because although memory use is only 53 of 64GB, it’s swapping 1.94GB to disk all the time? And continues to do so even when memory usage drops to 42GB (then Memory Pressure is yellow).

Right now the file is open, I’m not doing anything (except type this), memory usage is 46GB and it’s using 7.53GB of swap. Even with all the hungry apps shut down (no Photoshop, Lightroom, InDesign, PPT open), it still using 3.3GB swap for a 16 browser tabs, Slack, Mail Reminders, CCC and one small Word doc.

I took Apple though workflow with example files before getting this Mac and they swore blind thats 64GB M3 Max was over-specced and even 48GB would be 3-4 times faster than the old Intel and would not start swapping. Ha. If only.
 
The swap-file drops to almost zero after some time (not immediately). If this is not the case, it means that some apps is not releasing the swap file. This can have a downward effect on performance if other programmes need access to the swap file. Only a reboot resets the swap file to zero. It should be noted that the way the swap file is used depends very much on the application in question.
 
Nice of you to answer and aplogies for taking half a year to respond. Apple refused to allow me to upgrade Mac, and I lost the will to keep trying to figure things out. Thanks for useful info - I think I understand it!

Size? Ranges from ±12000 x 8000 px, 550MB, 16 bit (average) to 106,299 x 46,684 px 8 bit, 17GB file (biggest) that was slow of course on 2018 Intel, but I could use all features needed in Photoshop. On the M3 Pro Max 64GB / 1TB, Gradient Tool (which worked in real time on Intel) is so slow it’s unusable; takes 4-6 seconds to respond to mouse, then jumps somewhere unpredictable. Takes a good 4 mins to save the file to SSD too. Oddly (to me anyway!) memory pressure green, so then why so slow? Maybe because although memory use is only 53 of 64GB, it’s swapping 1.94GB to disk all the time? And continues to do so even when memory usage drops to 42GB (then Memory Pressure is yellow).

Right now the file is open, I’m not doing anything (except type this), memory usage is 46GB and it’s using 7.53GB of swap. Even with all the hungry apps shut down (no Photoshop, Lightroom, InDesign, PPT open), it still using 3.3GB swap for a 16 browser tabs, Slack, Mail Reminders, CCC and one small Word doc.

I took Apple though workflow with example files before getting this Mac and they swore blind thats 64GB M3 Max was over-specced and even 48GB would be 3-4 times faster than the old Intel and would not start swapping. Ha. If only.

This sounds very odd to me. There is no reason why the same file should occupy more RAM on Apple Silicon Mac compared to the Intel Mac, unless the software is different. Do you observe this behavior across different software suites?

You mention that the memory pressure is green when working with the file, that would indicate that your problem is elsewhere, most likely with how the software has been implemented. It is possible that at least some of the code used for the ARM version is different and does not handle large working sets efficiently.

Seeing non-zero swap is perfectly normal after the RAM-intensive activity has seized. The swap size you see refers to the memory reserved for these purposes, it does not mean that the system is actively swapping. Memory management is a very complex topic and does not always follow your intuition.
 
TLDR: The Endiannes don’t dictate the amount of memory used.

Some of us are old enough to remember that Intel did this as well - sharing RAM for application and graphic cards. Granted they were not as efficient and had to be configured statically, but the same principle.
If you had 16 and you configured your graphic cards to use up 4, you essentially only had 12 to use with applications.

Unified Memory works similarly, albeit dynamically assigned. This in theory allows you to have more RAM allocated for application or GPU as needed.

The amount of memory needed for your application will be the same, especially now that Mac apps are Universal anyway.
 
TLDR: The Endiannes don’t dictate the amount of memory used.

Some of us are old enough to remember that Intel did this as well - sharing RAM for application and graphic cards. Granted they were not as efficient and had to be configured statically, but the same principle.
If you had 16 and you configured your graphic cards to use up 4, you essentially only had 12 to use with applications.

Unified Memory works similarly, albeit dynamically assigned. This in theory allows you to have more RAM allocated for application or GPU as needed.

The amount of memory needed for your application will be the same, especially now that Mac apps are Universal anyway.

With x86 and integrated graphics, there has always been that hard partitioning of "system" and "video" RAM, which contributes to the reputation of integrated graphics being slow and clunky. The bigger difference isn't in how x86 and Apple Silicon handle GPU and CPU allocations, but in how the two platforms handle data that are being worked on by the CPU and GPU simultaneously. In the x86 model, data being processed by both the CPU and GPU simultaneously gets copied into RAM twice (the CPU and iGPU partitions) then the two versions are compiled together before completing the operations in question. With UMA, both the CPU and GPU cores can work on the same piece of data simultaneously, which eliminates the need to reconcile differences prior to completing the operations in question. This is also why M-series graphics performance is approaching and in some cases eclipsing dedicated GPUs from Nvidia and AMD.
 
With x86 and integrated graphics, there has always been that hard partitioning of "system" and "video" RAM, which contributes to the reputation of integrated graphics being slow and clunky. The bigger difference isn't in how x86 and Apple Silicon handle GPU and CPU allocations, but in how the two platforms handle data that are being worked on by the CPU and GPU simultaneously. In the x86 model, data being processed by both the CPU and GPU simultaneously gets copied into RAM twice (the CPU and iGPU partitions) then the two versions are compiled together before completing the operations in question. With UMA, both the CPU and GPU cores can work on the same piece of data simultaneously, which eliminates the need to reconcile differences prior to completing the operations in question. This is also why M-series graphics performance is approaching and in some cases eclipsing dedicated GPUs from Nvidia and AMD.

Not really. Intel had UMA as early as 2012. And Intel iGPUs have access to the same memory as the CPU, just like Apple Silicon. The system/video RAM split was mostly a driver-imposed distinction. And it’s not like Apple UMA or Metal prevent memory copies entirely. Most games are written with the dGPU model in mind and will create unnecessary copies anyway. Not to mention that for best performance textures need to be stored in “GPU-only” allocations so that their layout can be optimized.

And finally, the reason why M-series GPU are fast is because they are fast. Old Intel GPUs would also have been fast if they were large, had huge caches, and plenty of high-bandwidth RAM. But they didn’t. They were budget devices, optimized for cost and not performance.
 
Something I don’t see mentioned often is the impact the number of processor cores has on RAM usage. MacOS is going to cache data in RAM for every existing core doing work. That cache exists to speed up things like application switching, or re-running jobs. The more cores you have, the more cache the OS keeps. In my experience, when user app memory bumps up against the various caches, MacOS is slow to flush out the cache. It’ll first start compressing memory, and even go to swap before it will push out old cache.

To me this was most noticeable going from 2 to 6 CPU cores. With 2 cores, 16GB was more than enough RAM for my work load. With 6 cores, RAM became the bottleneck, and I had to go to 32GB RAM with a discrete GPU. That helped and was sufficient.

With 10 or more cores, I find 48GB is “more than enough”. I tend to use a lot of apps and switch a lot, so RAM swap or even memory compression produces a noticeable slowdown in my workflow.

If OP increased their core count in their upgrade, to me that could explain why 64GB might not be enough.
 
Something I don’t see mentioned often is the impact the number of processor cores has on RAM usage. MacOS is going to cache data in RAM for every existing core doing work. That cache exists to speed up things like application switching, or re-running jobs. The more cores you have, the more cache the OS keeps. In my experience, when user app memory bumps up against the various caches, MacOS is slow to flush out the cache. It’ll first start compressing memory, and even go to swap before it will push out old cache.

To me this was most noticeable going from 2 to 6 CPU cores. With 2 cores, 16GB was more than enough RAM for my work load. With 6 cores, RAM became the bottleneck, and I had to go to 32GB RAM with a discrete GPU. That helped and was sufficient.

With 10 or more cores, I find 48GB is “more than enough”. I tend to use a lot of apps and switch a lot, so RAM swap or even memory compression produces a noticeable slowdown in my workflow.

If OP increased their core count in their upgrade, to me that could explain why 64GB might not be enough.
Core caching is different than file caching. They have nothing to do with each other. More cores doesn’t change the amount of RAM used except that you can do more work in a unit of time with more cores.
 
  • Like
Reactions: Ruftzooi and leman
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.