How does RAM usage work?

L1NTHALO · n00b Joined: 27 Aug 2024 Posts: 22

I always want to keep my systems as minimal as possible, that's a huge reason why I switched to gentoo.
On of the most fun parts so far has been checking my idle RAM usage after installing an OS (arch, windows, gentoo etc), the lower the better.

However I have realized some inconsistencies. First of all Gentoo is supposed to be minimal but I often get the same or just slightly higher RAM usage on arch. Sometimes gentoo uses more RAM.

Also on my different gentoo installs memory usage varies. For example kernels that I deemed more minimal than others use the same or more RAM than bigger kernels. And I noticed that systems with more GB of ram use more of it idle.

I used Hyprland, minimal use flags and openrc on all these systems (other than systemd on arch).

Now I know that caching and memory allocation exists but how does it work? For example I noticed that ram usage spikes when starting the system and then gets lower with time (if doing nothing).
Is more RAM getting allocated if you have more system RAM available?
Also do optimizations such as -O3 -lto and other aggressive CFLAGS increase ram usage?

And yes I'm aware that less ram usage =/= better but it's fun to look at fastfetch and see only 200MB ram usage in idle (tty).

eccerr0r · Posted: Sun Oct 20, 2024 9:54 pm Post subject:

compare software versions including kernel, which has a great affect on memory usage especially if security options are enabled.
compare init scripts loading in files into cache that are subsequently unneeded later during boot
compare compile optimizations and configuration flags, usually -O3 uses more text RAM and -Os less. It will not affect runtime/stack/heap usage where glibc/kernel/software versions could.
compare how much ram you have (more memory you have, more page table memory needed)
compare drivers/hardware used (using a Trident card will use less memory than an nVidia).

I don't even care anymore on PCs until I actually run out of RAM, and inter-distribution comparisons are virtually worthless. Most of the time the memory is used to improve my experience (things like udisks/elogind/polkit and automount...saves me a manual step of manually mounting) and that's worth the RAM especially if it's not being used -- again unused ram is wasted ram.
_________________
Intel Core i7 2700K/Radeon R7 250/24GB DDR3/256GB SSD
What am I supposed watching?

pietinger · Posted: Sun Oct 20, 2024 11:27 pm Post subject:

Maybe you are interested in "Tutorial: Linux Memory Management and Containers - Gerlof Langeveld, AT Computing"
https://www.youtube.com/watch?v=ql1axx--8sI
(dont worry, the first part is not about containers; you will also see "atop -B" ;-)

)
_________________
https://wiki.gentoo.org/wiki/User:Pietinger

CaptainBlood · Advocate Joined: 24 Jan 2010 Posts: 3872

eccerr0r · Posted: Mon Oct 21, 2024 5:31 am Post subject: Re: How does RAM usage work?

L1NTHALO · n00b Joined: 27 Aug 2024 Posts: 22

Thanks for the answers!

A lot of really interesting info here.

What would be examples of more minimal wayland compositors? I wanted to experiment with other WM's anyway so why not try something minimal.

Also how would I get started on learning more about these things on the side?
Just started my CS college degree and linux and the whole concept of gentoo (compiling, use flags) as well as the kernel & hardware side of things has really catched my interest.

logrusx · Advocate Joined: 22 Feb 2018 Posts: 2443

flexibeast · Posted: Mon Oct 21, 2024 9:40 am Post subject:

eccerr0r · Posted: Mon Oct 21, 2024 1:19 pm Post subject:

LTO is generally a good thing except when it can't be done - when you simply don't have the memory resources to do the optimization during binary build. After you do the optimization of the binary you'll reap the benefit of the binary from then on out, it won't use more memory again until you need to recompile the package. LTO affects static memory consumption of the software being compiled but does not affect dynamic memory consumption.

Lately I've been wondering about binary sizes and how they have and haven't been changing. Yes most software is using more and more RAM as time goes on and versions go up and up, but how much of it is going towards static memory/text/code versus dynamic memory/data, which I also include heap and stack. I think text sizes increase linearly over time for the most part, but data is growing faster than text IMHO. An old text HTML page in 1990 is a couple kilobytes, now in 2024 it's megabytes with all these jpegs and video which as 2D and 3D (2D images plus time, and soon we'll go to 4D - 3d images plus motion/time) sizes grow nonlinearly.
_________________
Intel Core i7 2700K/Radeon R7 250/24GB DDR3/256GB SSD
What am I supposed watching?