Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Hold up now - does the Compressed Menu column show the physical memory used by compressed memory, OR the size of that memory when uncompressed ? I think its the later because on my machine the compressed memory for one item is bigger than allt he compressed memory down the bottom.


You are not entirely correct in your interpretation. @Ryan P 's correction to my post above was correct (thanks Ryan P). The cache is loading up stuff that the OS thinks will be soon needed. But that memory would otherwise be sitting there charged up but doing nothing (you can't turn off RAM fully). The OS is using spare CPU cycles and disk access to predict whats next and load it into cache. If it is right then you get a 'performance boost' as it doens't have to fetch stuff from disk. But its not really your apps using that memory - its just what the OS thinks is next and it can throw it away whenever. Its simple more efficient than having all zeros there or random bits. You are mixing up swap space and cache (as I did before Ryan P's correction).


NO, your point was that you should have extra RAM even if you are NOT using up all the RAM. As you can see I'm using way over that. Nobody buy nobody is disputing extra RAM is needed in that case. (well.... some people will tell me I should optimisie the code blah blah blah... but those people are idiots who dont' realise its not my code causing the problem 😜).


No the swap does what you think cache does. The cache is a speculative exercise than might save you loading from SSD to RAM time IF the OS guessed right. This being said - have more space in which to guess things may improve the odds of guessing right 🤔



I'm probably more in agreement with you than most on this thread!



I'm not talking from a single time point. Watch it for a while while you load some stuff up - they mostly stay in balance ass I suggested. Cache is physical minus used IF the machine is not too busy doing other stuff (i.e. CPU is at 100% or SSD i/o is maxxed). So it can happen that it won't be used - and it seems to keep a few hundred MB doing nothing most of the time - it doesnt' add up to 100%.

I was going to ignore this but your memory pressure calculation makes no sense.

Edit: Also. Pay attention to the design - the graphic tells you how it makes up physical memory - the littel dent is to show you the thing on the left i.e. physical memory is made up of the stuff in the on the right (App + Wired + Compressed). Its Apple - you know they love such little design features:

View attachment 877992

edit: Okay, after some testing, you are right about Cached Files not being part of Memory Used. But Cached Files doesn't always equal Physical Memory minus Memory Used either. In my iMac, it never does, not even close, but maybe that's because I have 40 GB of RAM. And as a computer engineer, I still think that Cached Files should be included in the Memory Used calculation, because everything I know as an engineer tells me that freeable memory, which is what Cached Files are, is still used memory until it has actually been freed, not free memory. Apparently, the engineers at Apple used to think the same way as me, as you can see by the previous design of the Activity Monitor below. This screenshot is from Mavericks, and as you can see, Cached Files used to be included in the Memory Used calculation and grouped in with App Memory, etc in that pane on the right. Not sure when Apple made the change to take it out of there, but it seems like some confused intern decided that freeable memory was the same as free memory. I guess Apple was trying to make it look like computers had more free memory than they actually did, because maybe some users were concerned that almost all of their physical memory was always being used on machines with lower amounts of memory.

os_x_mavericks_activity_monitor-100388083-orig.png
 
Last edited:
There’s no question memory compression happens. That’s why Catalina can run with 2 GB RAM but on my machine with 24 GB it idles around 8-10 GB used with no apps running.

The more physical RAM useable, the less compression and swapping to the disk you’ll encounter. Ever used Windows Vista/7? It will go right up to the ceiling of your RAM and slow down like crazy. macOS will almost never do that (Unless you have 4GB.)

And- RAM upgrades are important but there are many times when your CPU/GPU will slow you down. Try editing a half hour multi cam 4K video in FCP X with 32GB RAM and you’ll see what I mean...
 
Anyone got a comparison of 32GB vs 64GB when it comes to the battery? Wonder how much of an impact 64GB will have. Certainly more than a CPU upgrade.

I'm still on the fence, but I think 32GB will be enough. 64GB would be "just in case", but I feel that usually I need less than 32GB and whenever I need more than 32GB, I also need more than 64GB, which is why I run those things on servers with CPU/GPU power and RAM that can't be bought on mobile systems.
 
  • Like
Reactions: dead flag blues
Are these purported RAM performance hits while still in the green quantified anywhere, in any tests, or is it a case of just feeling them because they somehow have to be there?

I would also like to see that :)

Edit: I'be highly interested in numbers for "that 64GB will eat through your battery in no time anyway" as well – how much more battery power does 32 or 64GB RAM really use?

According to the tests done by TomsHardware in 2014 (https://www.tomshardware.com/reviews/intel-core-i7-5960x-haswell-e-cpu,3918-13.html), 32GB DDR4 consumed 11.85Watts on average. Current DDR4 is still the same spec, so I can't imagine it being that much more power-efficient. At any rate, even if we assume that power-efficiency has improved by the factor of 2 (which would be an incredible progress), 64GB RAM would continuously pull 10W — add in everything else and you are looking at average power consumption of around 15-20W on idle (5-6 hours of battery life) — in best case. If it is indeed 20W for RAM alone, then its 4 hours on full battery.
 
  • Haha
Reactions: MrTemple
are you sure the power consumed by RAM is constant? or only under load?
I hardly see a mac consuming 10-20W on idle just by RAM alone
 
  • Like
Reactions: MrTemple
are you sure the power consumed by RAM is constant? or only under load?
I hardly see a mac consuming 10-20W on idle just by RAM alone

Since DRAM is volatile storage it needs a constant supply of power. The LPDDRx variants have enhanced power modes that can make them more power-efficient. I don’t know much (or anything to be honest) about semiconductor electronics, but at least this paper suggests that DDR4 background power consumption is a significant factor: https://www.cs.rochester.edu/~ipek/micro15.pdf
 
EDIT: I will read that paper but im my testing.. it doesn't affect battery life in any meaningful way lol

I have tested this (kinda ghetto). I put 64GB of ram in a HP 14 with a dual core Pentium gold. The idle power draw is so low I thought I would test how much 2 sticks draw.

The laptop came with a single 4GB stick. I replaced it with 2x 32GB sticks at Idle.. it drew on average, 250mA more. So basically it doesn't affect battery at idle. Under load I couldn't come to any conclusions.


4417u 64.jpg
 
My whole 64 Gb machine can idle down to about 7 W - obviously, the RAM isn't pulling 10 W on its own... It's also very capable of pulling triple-digit wattages under load...
 
Interesting discussion. It would challenging to quantitatively evaluate the speed improvement due to the numerous confounding factors.

One confounding factor is certainly the style of usage. In the OP’s “minimal” setup he keeps many applications open at once. There is no way he is looking at all those applications at the same time. Meanwhile, I am constantly closing and re-opening applications to effectively manually manage RAM. Perhaps with more RAM I could just keep everything open which would definitely be faster. However, keeping so many things open is a lot to keep track of. Maybe you just don’t have to once you have done this.

As an aside, I think my current usage pattern is related to a fundamental difference between older Windows versions (they have been slowly moving away from this) versus macOS. Coming from many years of Windows usage, I am used to having a taskbar that has one tab for each instance of each application. I would use this to well, manage open windows. The macOS Dock can kind of do this (you can set it to only show open applications) but even when you enable icons for open discrete instances it’s difficult to manage because they are small icons without the full application title.

I have finally succumbed to keeping all of my often used applications on the Dock, looking for the tiny dots, and right clicking to select specific open instances. PITA. But, by having all the applications laid out there you are indeed encouraged to keep them all open all of the time. Having more RAM would probably save me from having to micro-manage open applications. But it won’t help me more easily select specific instances of said applications. IMO you should be able to hover over a dock icon and see open instances. Is there a way to do this?
 
I just saw an idle power of 4.2W on my 64 GB model with the display on (and I've seen power in the 5W range repeatedly). Even assuming that the CPU was doing pretty much nothing at all, and was idling at a few hundred mw (and that no other component was drawing anything significant), half of that had to be the display. That leaves a maximum permanent draw for the RAM of a couple of watts... There's no power reason to avoid high-RAM configurations.
 
I ended up ordering the 2.4GHz i9 with 64GB RAM, 5500M 8GB and 2TB SSD. If for nothing else, I'd at least be able to run two VMs conveniently with that config if I ever need to.
 
I ended up ordering the 2.4GHz i9 with 64GB RAM, 5500M 8GB and 2TB SSD. If for nothing else, I'd at least be able to run two VMs conveniently with that config if I ever need to.
Depends what kind of VMs you mean. I semi-regularly run 10-12 (headless) Debian VMs concurrently on a 6-core i7/64GB 2018 Mac mini without an issue.
 
If you are using a Mac laptop from 2015 or later or using a magic trackpad 2, you can force click on the icon on the dock.


Yea the ergonomics for the trackpad don't work for my wrist. That said, thanks, I realized Ctrl + Down also does this (and I created a more sensible one-handed hotkey): Option + Caps Lock (modified as Escape) for App Expose as well as Option + Tab for Mission Control (and immediately show desktop preview). Also set Option + ` for "Menubar in Context Menu" because I use multiple monitors and can't stand the menu bar positioning in macOS. All these modifications except Caps Lock for Escape were done using BetterTouchTool.

FWIW I also have Option + 1 = left third of screen, Option + 2 = middle third, Option + 3 = right third, Option + Q = left half, Option + W = right half, Option + X = Maximize Window. Option + A = Tab View in safari, Option + S = screenshot, Option + C is set up in Alfred as the hotkey for its Clipboard, Option + V = paste unformatted, Option + Z = open all previously closed tabs in safari.

OK that's a lot and I have more ... but those combined with the standard Cmd + Tab to switch between apps, Cmd + ` to switch within apps, Cmd + T for new tabs, Cmd + W to close tabs, Cmd + Q to close app, and so on: this computer feels like a damn rocket ship to me again.
 
I just measured the draw of 64 GB directly, using Intel Power Gadget - the RAM draw is around 1.25 watts! Even under heavy use, cataloging images in Lightroom and editing in DxO Photo Lab, I couldn't get the RAM to draw more than a couple of watts. Either all notebook DRAM is much more efficient than its desktop counterpart, those 2014 numbers are way off for today's memory, their measurements are off, or Intel Power Gadget is way off...
 
Depends what kind of VMs you mean. I semi-regularly run 10-12 (headless) Debian VMs concurrently on a 6-core i7/64GB 2018 Mac mini without an issue.
Most of the time, Windows or Linux but with 3D graphics. I do Deep Learning research and in addition to the "regular stuff", my research group and our students sometimes work on AI bots for games, which is sometimes impossible for Mac.
 
I just saw an idle power of 4.2W on my 64 GB model with the display on (and I've seen power in the 5W range repeatedly). Even assuming that the CPU was doing pretty much nothing at all, and was idling at a few hundred mw (and that no other component was drawing anything significant), half of that had to be the display. That leaves a maximum permanent draw for the RAM of a couple of watts... There's no power reason to avoid high-RAM configurations.
oh my goodness, i didn't even see DRAM before. I guess that's how you look at power draw :rolleyes: duh lol

so the idle numbers match up and the full load numbers are just shy of 2 watts. yeah.. when Apple said ram hurt battery life.. idk how they came to that conclusion. i mean it does but.. by minutes haha
 
The big power draw on the 16" (besides obvious heavy loads - importing 16,000 images into Lightroom draws a ton of power) is anything that unexpectedly kicks on the Radeon...
 
Cached files in that table is files that are cached in physical memory. Used memory + cached files = your physical
Memory, and another ~6gb of currently idle app memory has been moved to swap.

Also keep in mind that swap isn’t necessarily bad. If the program is idle and another process needs to load a heap of data, using it for file caching (and letting the first app be swapped out) is likely to give you better performance.

For reference, my mini has 64Gb and I’ve seen it swap a few gb here and there.
[automerge]1574100973[/automerge]
I should add: I agree that getting as much memory as you can afford is a good idea. But I also wanted to clarify about the swap vs cached files stuff.


Another reason swap isn't bad: even if you aren't short of memory, modern platforms copy inactive applications to swap before they have to. Why? Because if they are already swapped to disk and a big allocation request comes in, you don't need to WAIT while the memory manager pages out an application. If it's been in-active - it's already paged out, just dump it from memory.
 
I've been comparing a 16gb and a 32gb 2019 16" . Benchmarks don't report any significant difference between the two.
 
I have a 64Gb machine 16in, previously had a 2016 15in 16Gb machine. The performance is far better. 'Memory used' (10 apps open, 10 Safari tabs) is around 16Gb. So I'm guessing there would have been a performance hit for the previous machine managing this (light) workload.

The older machine was pretty poor for Windows 10 via VMWare performance. I haven't installed that yet on this - but expect a corresponding performance boost both in the VMWare OS and MacOS concurrently.
 
The older machine was pretty poor for Windows 10 via VMWare performance. I haven't installed that yet on this - but expect a corresponding performance boost both in the VMWare OS and MacOS concurrently.

The biggest difference I've found (admittedly I don't use Windows VMs that often) is the ability to allocate a pretty decent chunk of physical memory to a VM, so the guest OS is less likely to need to use swap via a virtualised disk interface.
 
OK - this plays much much nicer with a single Windows VM and around 10 Mac apps open. No fans spinning up. Performance on Windows is good. Activity monitor shows 25Gb memory used. I do think if I'd maxed out ram on the previous machine, it would have been better.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.