Linux Ate My RAM, What Do I Do?

[@Shuna322]

You know what's fucking funny? I actually had an Ubuntu 22.04 release that was shipped with a Gnome Extension that had a memory leak that was fixed in a future release. But since the extension was shipped with the system, I was unable to update it, and it wasn't updated with the OS updates, like, ever. I had to remove the extension from system files and install the official release from the developers. But sure, when I complained, 99% of the people called me illiterate and linked me to this page.

[@DataToTheZero]

That's why I love that Brodie, took the time to ask the correct follow ups! I see ram problem. Okay, do you have actual perfomance issues?

[@Trenjeska]

Had the same with Pop_os 22.04 (based on that same ubuntu version you mentioned) It seems fixed now, but I still need to keep an eye on the swap

[@Dratchev241]

i had an issue with a waybar module at some point it would just leak and within about 2 mins id go from my normal 2.8G used to maxed at 15.9G and system crash. was a bitch to track down cause it could go days no issues then something would set it off. and about every time it would go off I would be in the middle of something else (not pc related) and look up to see ram use pegged and locked up system. after about 2 months of that I finally caught it, yeeted the module out of the waybar config and no problem since. Never bothered posting to the waybar/module people about it cause i know it would result in the "nothing wrong mate, linuxatemyram-url"

[@scj643]

Most likely the apptray icons extensions

[@comradepeter87]

Bruh this might be something that's screwed with my Pop!_OS installation too. On a 16G system, it shouldn't be normal for it to casually occupy 10G (and another ~5G for cache) when only Firefox tabs are open. Could you drop the name of the GNOME extension?

[@WooShell]

I paid for that RAM, I'm damn well going to use all of it.

[@808MELOW]

LOW IQ USER , Thats why linux is not for normie

[@lucascamelo3079]

I paid it to not see it going up 25% when on my home screen.

[@vaisakh_km]

TLDR: (for those who commented before watching) It's disk caching, where files are cached on to ram for faster FS access, it's doesn't affect application ram requirement..

[@chillgamervids]

thanks, i'm 5min in and was just about to comment a TLDR, then a I saw you >D i'm here for the drama, not superfluous explanations that are available in much terser form like docs

[@Simulati0n]

I don’t quite understand this settlement behind the scenes of the OS. Specs are already high, parts are moving so fast how much faster can it be by leaving things cached? When gaming I want to see my current game/software use the space. I do not want to see my system telegraphing my possible apps to open.

[@TheExileFox]

The funny thing is that Windows does similar things.

[@Simulati0n]

@@TheExileFox exactly, which I don’t get and this is a fresh install. The philosophy of no ram use is wasted ram is dated. Ram is an overhead of what your system can manage, if it’s simply using it before you launching things imo it’s bloated. Don’t try to predict my use case to speed things up.

[@ericjbowman1708]

@@Simulati0n You don't want to bottleneck your system on disk i/o at 512b|4kb blocksize when RAM is 4mb. Since RAM is faster than disk anyway, wouldn't you rather your shaders and whatnot get stored in cache if they're evicted from the VRAM?

[@SolaceCafe]

Meanwhile I have 64GB of ram and just, ignore my ram being used entirely because it's never gone above 40

[@fomxgorl]

yep. also have the same and haven't had to worry about RAM usage on my personal system in a long time now. will still have to care about it as a dev though as not everyone is going to have 64 gigs of RAM

[@johanngambolputty5351]

@@fomxgorl this also seems to be a good way to maybe get lazy and write hungrier than necessary software, probably how a lot of bloat happens

[@sysbofh]

@@johanngambolputty5351 Up to a point. I think it's important to have the fastest system possible to development - but a slow one to testing and validating. Your code should be efficient - but your development rig should be fast. The testing rig, though... should be REALLY slow.

[@kaz49]

Lucky 🙃

[@averdadeeumaso4003]

Lol I got the same

[@piotrkasperski382]

Unused ram is wasted ram!

[@prakhars962]

welcome to windows

[@NepTuneLuis]

@@prakhars962 and i am micosoft. time to spy and add artificial you to you

[@startjim123456]

Unused ram neesd to stay unused until its need windows does not need ms needs it to spy and then there is the thing it wil be released and usable by programs of the user... It never releases is it only takes more

[@nezu_cc]

@@prakhars962 Windows does the exact same thing, it just reports "free" as free + cache so users don't get confused.

[@MrAlanCristhian]

Also, unused hard drive space is wasted hard drive.

[@darksteelmind]

This man is incredible, I wish he existed outside of my imagination

[@Hyperboid]

yeah me too... wait could this mean...

[@bologna3048]

Maybe one day he will be a real boy

[@deltamico]

No, it doesnt. Take for example god. More than 1 person is imaging him too

[@comosaycomosah]

​@@deltamicoI smell a fallacy....so if God dont exist and believers are....imagining a creator.... are Satanist also doing the same imagination with satan? And all that considered have you consulted the Oracle in templeos about this?

[@faxen123]

protip: download more ram

[@AlanJames1987]

I looked for this comment. I wonder if I can make a drinking game based on how many times it's mentioned.

[@nsr-ints]

I mean, technically you *could* mount a networked drive as swap, but why would you do that?

[@AlanJames1987]

@@nsr-ints that's not a way to "download more ram". My grandma in 2005 wanted to click a button and download more ram, because the box was magic.

[@GirlWhoCriedAardvark]

How much ram? Think of the biggest number you possibly can. Then double it.

[@nsr-ints]

@@AlanJames1987 le rm -fr /* script:

[@lapissea1190]

I think it is genuinely quite important to show what is using ram? Like in terms of actual programs or drivers or what ever else. There are programs like zbrush or JVM that change their behavior based on how much available system memory there is (I don't mean unused but trivially acquirable) and it can be very frustrating when you just know that you have 3gb of ram free and don't know what is using the rest where you genuinely need the number to be lower.

[@azurarobo]

you have top/htop/btop etc for that

[@marloelefant7500]

My laptop has 16GB RAM, but it only takes a few Firefox tabs and an IDE open for it to go out of memory. For years, 8 GB was all I needed, now I consistently go OOM with the double amount. I don't want to believe this.

[@4ngeldus739]

That seems kind of odd. I have 6gb of RAM and have no issue with having Firefox and VScode open. VScode in itself not just an IDE but an IDE built on electron which basically makes it a whole other browser running on my system. I'll admit I'm not much of a programmer and I'm sure there are some extensions you can run that will be more taxing. I'm sure as well you can run into the issue of having too large of a working directory. Sometimes that makes VScode crash (or at least I've heard it does, I'm too much of a noob to experience that)

[@4ngeldus739]

Also I'm on Ubuntu running the snap version of Firefox that comes on Ubuntu Mate, Ubuntu Mate also uses less ram than gnome or kde by default

[@marloelefant7500]

@@4ngeldus739 I am usually using the JetBrains' IDEs which take up a few GB of RAM.

[@attilavs2]

On Linux i personally don't really see that, i have 16GB on my laptop and i have to try to fill it up. I can have a dozen browser tabs + a game + steam in the background + other random stuff and still be fine

[@eDoc2020]

If you're going OOM then you could probably make use of more swap. When on my older laptop I will often notice excessive slowdowns due to swap usage with only a few tabs but that's with 4GBs and heavy 2024 websites.

[@autarchprinceps]

The KDE process plasmashell has several memory leak issues that have been open for years at times. Sometimes it might be Nvidia's fault, or at least partially, but it can be as easy as that you are rotating through wallpapers, because you want to avoid OLED burn in. Between RAM & swap I have legitimately seen it alone use more than 100 GB. I have seen some truly empty Linux installs, they rarely use even a 100 MB, but in the mind of the ordinary person the other open source software running is part of it. I mean non of us actually say GNU/Linux either. So it may not be just disk cache sadly

[@tonnylins]

ooh, nice

[@Triro]

You cant get much worse than windows, casually using 40% of my 64gigs if I dare leave it running for more than 2 days...

[@GrannuQwerty]

Why you have on for more than two day?

[@trajectoryunown]

@@GrannuQwerty Why not?

[@video_cumsumer]

I remember downloading one file off the internet for like 6 days

[@GrannuQwerty]

Cuz they can just buy something like a pi

[@blazewardog]

​@@GrannuQwertyHow well does a pi do at local AI text gen? Or playing BG3? Computer always on isn't a use case that a Pi is extremely good at. It's good to at being a cheap low powered computer.

[@enderzombi102]

oh no, my ram!!! what will i feed my protogen now?

[@flamingscar5263]

In emergency situations protogens can be fed SSD storage instead

[@atemoc]

Your protogen will simply suspend itself to RAM, and it will automatically hibernate (suspend to disk) after 15 minutes of being suspended to RAM in case battery power is too low. You will still have to power it back on and off before upgrading the RAM.

[@DigBipper188]

what if protogen is in your system he *is* the disk cache he eats all the RAM but gives it back upon demand from the kernel because he doesn't wanna be detected. sneaky prooter is sneaky.

[@UsernameDoesntCare]

This poor child.

[@callyral]

the penguin eated it all :3

[@evanotoole787]

People had the same non-issue with Windows Vista, but there was no “Windows ate my RAM” article to explain what’s going on. People thought all their RAM was being taken up by system resources lol

[@Cuperino]

I had so many Plasma session crashes this month due to running out of RAM… for context I have 64 GB. The culprit was an Intel driver add-on that didn’t make the jump from 22.04 to 24.04. This resulted in a daemon continuously trying to make use of it, leaking memory. I use ZFS and BTRFS, and often VMs, all of which contribute to higher RAM utilization, especially ZFS. I haven’t gotten a single crash since I found the responsible daemon, but my is that an annoying thing to encounter!

[@nezu_cc]

And then there is zfs, zfs cache works independently from the Linux FS cache and zfs is not so eager to give up its cache. That's why it's great on dedicated storage boxes where the only thing you want in RAM is cache, but it can cause trouble for systems with a lot of applications, especially ones that don't start at boot and/or fluctuate a lot in usage over time.

[@insu_na]

you can force ZFS to drop some arc by setting `zfs_arc_max` to a smaller value, tho it's important to note that if that value was 0 before, it can't be set to 0 again until the next reboot, so you'll have to specify the desired size limit manually

[@JodyBruchon]

I have zero interest in ZFS.

[@insu_na]

@@JodyBruchon nobody asked?

[@AlexMax2742]

Overcommit is a misfeature, and I would much rather a process die screaming than have my entire machine stuck the throes of OOM-killer thrashing which brings the entire system to its knees. The fact that it's a default setting on so many distros is actually insane.

[@Onihikage]

It really is. I installed EarlyOOM for a reason. Absolutely unacceptable at any time in the last 10 years for the whole system to start thrashing when it's out of memory instead of gracefully crashing the biggest program (or _something_ sensible) and returning the system to a usable state as quickly as possible.

[@JonBrase]

Avoiding overcommit won't stop thrashing, in fact, it will generally cause it to happen earlier. Nor does the OOM-killer have anything to do with thrashing. If you don't have swap, or if the programs using large amounts of RAM aren't using most of it heavily, the system won't thrash and will go straight to shooting processes in the head to free up memory. If you do have swap, and the amount of frequently used program memory is significantly larger than RAM, then the system will thrash itself to death for a fair while before it finally runs out of swap and starts shooting programs in the head. When you start shooting programs in the head, whether the OOM killer is in use just determines what programs die first. Without the OOM killer, the kernel just returns a failure for any allocation it can't fulfill, and given that 99% of programs don't even try to handle allocation failures, the effect is that a totally random process gets shot in the head (because there's no guarantee that the first request to fail comes from a program that's an actual memory hog). With the OOM-killer active, the kernel tries to identify an actual memory hog to shoot in the head (a non-critical process that's using a ton of memory). Whether it properly identifies non-critical processes is sometimes a bit of an issue, but the same processes could just as easily be killed by the allocation failure lottery. And all of this will happen with or without overcommit. All that overcommit does is determine whether the thrashing/allocation failures/OMM-killing happen when too much memory is allocated, or when the allocated memory is actually used. Let's say you have a program that allocates a terabyte of RAM up front, and then generates 10 GB of data, stores it in memory, writes it to disk, and frees 10 GB of memory, repeating that until it's generated a full TB of data and freed all the RAM it allocated. If you have a system that has 512 GB of RAM and 1 TB of swap, then on a system without overcommit, the first copy of the program will immediately swap half a terabyte of empty program memory to disk as soon as it's launched. This will likely bog the system down for a bit, but once it's done with that, the system won't really be thrashing. Launching a second copy of the program will either cause that copy to receive an immediate allocation failure when it tries to allocate its terabyte, or will cause the first copy to be killed by the OOM killer. As the surviving copy works, it will periodically have to reload bits of program memory from swap as half of its allocation got immediately swapped out. This will slow it down somewhat, but because we're proposing a program that really only touches any given bit of memory once, it won't really thrash. If the system is doing overcommit (with a sufficiently high overcommit ratio set), then both processes' allocations will immediately succeed. Since each is only working on 10 GB at a time, and frees memory that it's done with, the system will never show any more than 20 GB used at any given time, and absolutely no swapping will occur (despite more total memory being allocated than RAM+swap actually available). Of course, in reality, most programs use their memory more efficiently than this example, but allocated:actually-used ratios of 2:1 or 3:1 seem to be decently common, and most programs memory usage patterns are more complex and prone to thrashing on swap than this example. So without overcommit, a real workload running on a system with the RAM/swap configuration listed above might go from fine to thrashing to OOM-killing by the time that an equivalent system with overcommit even started to touch swap

[@JPs-q1o]

Agreed 💯

[@juanmacias5922]

Web browsers are the ones who eat my ram lol

[@daniell8387]

Unexplained ram usage should always be investigated

[@BrodieRobertson]

Often times its explained if you look at the correct number

[@daniell8387]

@BrodieRobertson true enough. Not saying panic everytime but awareness is important. Better to investigate something benign than miss something dangerous.

[@dipi71]

A nice bit of »echo 3 | sudo tee /proc/sys/vm/drop_caches«, maybe preceded and followed by a overcautious »sync«, can provide so much clarity. Cheers!

[@LostieTrekieTechie]

Okay but at some point 4GB of ram became insufficient for running a laptop with a bunch of browser tabs (Firefox crashing)

[@JPs-q1o]

​​​ "it's harmless" "it improves performance" What a load of crap. Brody, why don't you do a little experiment: - get 2 identical machines, the less RAM the sooner you'll see the phenomenon (8GB will work) - boot them from optical media into a live session, the slower the optical media access the more pronounced the phenomenon will be (USB2-based optical drive recommended) - use the desktop of your choice with any gecko-based browser of your choice for all your KZbin viewing (with new tab per day) - on the first machine: no installing to storage, no swap on storage (you _could_ create a swapspace but it'll only delay/dampen the inevitable), no closing the browser - on the second machine still no installing to storage or setting up swap but reboot it to live session from optical media and setup back to same desktop+browser every 5 days - observe the discrepancy between the first machine (i.e. where RAM is "used" and the second machine (where RAM is "unused") - when uptime on the first machine exceeds 15d and the desktop becomes nigh unusable tell me again how "it's harmless" and "it improves performance" This affects applications because files are cached on a most recently used basis while applications require system libraries to load/run. These libraries are not locked because an application requiring them is loaded but are instead kept cached only if/when their functions are called which updates their "recent use", this means they unfortunately get punted from the cache in favor of other "recent files" even though the system libraries are much more critical for performance. This results in "thrashing" and is why you witnessed that 1st system in the experiment go to slag for no good reason (i.e. you had enough RAM that the desktop and apps should not have experienced such crippling performance reduction). This has heen an issue for as long as I can remember so why am I putting this up as a KZbin comment on some guy's channel instead of filing a bugreport? Because I've been assuming for 15 years that "any day now" some smart person is going to fix this...but alas...nope. Also the kernel people are arrogant and rude and while you're not.

[@andrewi76]

I use Linux for commercial embedded software development and the first thing I do on a fresh install is set swappiness to zero. I don’t think a machine with 32 or 64gb of ram should ever swap application memory to disk if there are files in the cache that could be purged. I would often get stuttering applications with a couple gigabytes in swap while several times that amount of memory was used as disk cache. I think this is the case that annoys people

[@jefferyrlc]

If you use zram swap, I'd actually recommend turning UP your swappiness.

[@Megalomaniakaal]

Sadly, you need to have a swap, one at least as large as your RAM capacity to be able to hibernate the system.

[@JordanPlayz158]

​@@Megalomaniakaalpeople use hibernate, I have only ever went as far as sleep but I mainly use desktop, maybe hibernate is mostly used by laptop users?

[@andrewi76]

@ sure, I leave the swap on, but generally I’m happy to just sleep rather than hibernate… I just don’t like unresponsive apps because gigabytes of compiler output has exploded the disk cache

[@Megalomaniakaal]

@@JordanPlayz158 Yeah, probably more common amongst laptop users. But I recon also HTPC/media center users might make use of it, for an example.

[@Lordroran]

The problem of users not understanding this could be migitated if this portion of the RAM gets treated better in programs like top/htop. I consider this an UX problem there. It could even be argued that the disk-cached portions could be hidden per default and only shown per user request. It's not made better by htop's default on Ubuntu that disk-cache is yellow, suggesting a warning.

[@Hartie95]

100% agree, if it was shown differently, people would probably not come to the conclusion that the memory is full

[@ababcb3005]

To add more confusion to the mix, tools like the system monitor built into Ubuntu might not even show all the processes. My machine is currently using around 70 GB of RAM in total (have a bunch of VMs running), but the top 10 most memory-consuming processes that it lists only take up maybe 5 GB in total, with the rest only summing up to maybe 2 GB at most.

[@turculaurentiu91]

MacOS also uses ram for disk cache, but it's properly reported so nobody freaks out about it.

[@chocolateimage]

13:02 you can use "-mh" instead for a *bit* of a more human readable output

[@PanophobicCuber]

Or just "-h", in which case it selects the best units for you.

[@exp2745]

I am a firm believer that the number of people constantly monitoring and freaking out about memory usage and the people who feel compelled to run neofetch every time they open the Terminal is basically a circle. Memory is there to be used and keeping stuff cached makes a system snappier and more reliable. You paid for 256GBs, so use all of it.

[@immibis]

Linux finds it hard to use all of my 256GBs even with caching. A slightly fun story: I got a 10Gbps Internet connection at a hacker camp where they were always telling everyone to use more bandwidth so they can justify having such fast internet. I managed to download something at 4.5Gbps and then realised I was downloading onto a single SATA HDD, so went to check how my disk cache was doing. Yup, downloading 4Gbps into memory and at that rate, I'd have 4 minutes before I'd even notice any memory pressure. 256GB is nuts. Then had to take the computer apart for troubleshooting later, accidentally booted it with only one RAM stick, comment from someone else "what do you mean you ONLY installed 64 gigabytes?"

[@juanreina4554]

What if you don't have much memory and just want to just run things and be resource safe?

[@UmbreonShapeshifter]

Then leave it, the kernel will handle it

[@exp2745]

@@juanreina4554 Genuine question, did you ever encounter memory used for caching not being accessible when needed by your software? Page Frame Reclamation on Linux is renowned for being rock solid and makes it so you'll never have a downside from caching, in fact, especially when you are resource starved you should want as much caching as possible at every stage to get every last bit of performance from your system. And again, the second an application properly needs that memory, it gets reclaimed reliably.

[@electronnuclear]

Probably bcz those ppl don't know what cache memory is. I mean we have windows CACHING ram for running bloatware in the background. So they assume there's something sketchy running in the background.

[@The8bitbeard]

Some people just have some kind of weird anxiety about how the resources in their PC are getting used. They don't use their computer. They watch resource monitor software. They don't play games, they run benchmarks. They could have 64 GB of RAM or 128 GB of RAM and they're in a panic because some browser with a number of tabs open is using over 1 GB of that. It's like purchasing a whole warehouse and insisting that all your stuff only occupy a tiny 3'x3' corner of that space.

[@le9038]

If there is no more avaliable ram, what will my Chromium use now???

[@brlin]

Storage cache can be immediately released upon demand so your Chromium will still have RAM to consume.

[@eDoc2020]

Swap.

[@AccSwtch50]

Available ≠ Free

[@notCAMD]

Kent's Tech World showed that you can relatively normally use a Windows 11 PC with 4 GB of RAM even though at 8 GB it uses more than 4 GB with the same tasks

[@Arcidi225]

Maybe it's because of garbage collection? If you have more available ram, in theory it could be less frequent

[@marloelefant7500]

After my work Windows 10 PC has run a few weeks, I can close all open applications and it's still using 20 GB of RAM, even though there's no application in the task manager that has allocated remote as much. At that point, I frequently run out of memory when doing basically anything.

[@ilya_mzp]

Linux also has write cache, which is by default a certain percentage of your whole RAM amount. After installing a lot of RAM on my system, I started to notice periodic lags. Turns out the write cache became so big (in terms of absolute size) that the system disk couldn't flush it fast enough. So write operations would add up in cache and then block the process while it flushes. This is also the reason why dd writes an image to a USB drive at an insane speed and then gets stuck for a long time.

[@ilya_mzp]

It's configured with vm.dirty_ratio and vm.dirty_background_ration sysctl properties. Also vm.dirty_writeback_centisecs and vm.dirty_expire_centisecs control the amount of time before caches get flushed to disk.

[@thedave1771]

That… Doesn’t make sense. Assuming it starts writing and keeps going, and assuming you’re writing data faster than the disk can keep up, the cache size is irrelevant, the only difference is when you get blocked, in that a large cache can push that down the line. The write cache just gives you more runway, so that if you happen to stop writing data you get unblocked faster.

[@ilya_mzp]

@@thedave1771 it blocks the process until the whole cache is flushed, so if the cache is 6GBs, it'll take a while. And any other processes couldn't read, as a cache flush was in progress. Sure, it'll try to write the data in the background after you exceed dirty_background_ratio or the background timeout, but it'll fully block the process if you reach the dirty_ratio. After changing the defaults, the random stutters disappeared. It doesn't make sense, but that's Linux for you. The whole i/o stack is kinda screwed.

[@thedave1771]

@@ilya_mzp Is the process flushing to disk? Or actually... If it's an external drive, the OS might not return until the write operation is flushed to disk, because users tend to expect that once a save "finishes" they can pull the drive and go. Never looked into this on Linux, I don't think I've used an external drive ever other than the initial OS install. Either way, the cache isn't going to slow you down in terms of absolute clock time, it will either be the same (if the application is constantly writing faster than the drive) or faster (if the application is writing in bursts and then pauses, like a download that is being decoded in chunks). But it may shift the apparent wait time to the very end. Unless the cache just sucks, but I can't imagine that that is the case in general.

[@ilya_mzp]

@@thedave1771 google "Low write performance on Linux servers with large RAM"

[@insu_na]

there is at least one legitimate usecase I found for dropping the filecache at runtime: if you want to start a vm with a GPU passed through in PCIe passthrough mode. if the memory you have on your host is not contiguous then allocating that memory to the VM takes *forever* last time I did that without clearing the cache beforehand it took almost 2 hours to boot a VM with 96GB of RAM

[@bodgemaster7946]

This question is why typical GUI system monitors bunch in cache with free memory and why `free` has a column for available memory.

[@celdaemon]

Because the cache will be released if there's memory pressure, it's technically "free" for use by other programs if needed

[@nezu_cc]

Because the GUIs on the Windows and Mac side do the same. So they are lying to the users, just like every other OS because that's what users expect. Windows also has a bunch of cache that's almost always full, it just doesn't show it to you (well it does, just barely). free just shows the real state of things directly as reported by the kernel.

[@bodgemaster7946]

@@nezu_cc I wouldn’t call it "lying" so much as looking at it as a matter of definition (like what Brodie said in the video). For most use cases, this memory might as well be treated as unused and is free to be taken.

[@thedave1771]

I’d argue it’s mislabeled. Better labels would be “available” or “unallocated” for the sum of cache+free, and “wasted” for the subset that is free.

[@fahadahaf]

@@nezu_cc Where are you getting the "lying" on macOS? MacOS separates out the memory stats into Free, Used, and "Files Cached" in Activity Monitor, "Used" is further broken down into "Wired", "Compressed", and "App". Are you sure you're not just imagining things to feel better about using Linux?

[@arcane_spark]

My biggest issue is that when I run out of ram programs are not getting killed Running into that way to often that my system freezes up because I ran out of ram

[@thedave1771]

Why would programs get killed? That sounds horrible if one program with a leak trashes your system. If you’re seeing crashes it sounds like a misconfigured page file 🤷‍♂️

[@Aera223]

​@@thedave1771by oom ki||ing the program leaking memory, it may help keep the system stable

[@Onihikage]

@@thedave1771 One program with a leak ALREADY trashes your system when it runs out of memory and starts thrashing to the point of unresponsiveness until the user eventually power cycles it. Tools like EarlyOOM simply kill whatever process is using the most memory when available memory drops below a certain threshold (depending on total memory size). In the cases of a memory leak, it's very likely that the process with the leak is also the one using the most memory and the user would choose to kill it anyway if they were aware of the problem. In the end, the most important thing is that the system stays responsive, even if it means an important program crashes. People who work for a living are not going to tolerate their system becoming completely unresponsive for 30 minutes / two hours / overnight / however long it ends up taking for the default OOM killer to stop wringing its hands and actually kill something. I cannot think of a single case where it's happened to me that I would have preferred a frozen system over a program crash. Not one.

[@recurser.]

you can try installing earlyoom

[@arcane_spark]

@@recurser. thanks that did the trick

[@toxiccan175]

It’s called caching.

[@AhmedNasser-tj2fb]

The burialgoods gnome has invaded the Linux lands.

[@VictorSouza02]

My only real issue with RAM on Linux is Firefox + KZbin... Easily eat 6-7gb ram and youtube tabs starts lagging after some time, and i do not have this issue with Brave.

[@MikeSharpeWriter]

Sadly, this is also an issue with Windows version of Firefox and KZbin!

[@Dratchev241]

@@MikeSharpeWriter sadly it is just an issue with firefox and firefox forks. FF is just horrible with resources.

[@VictorSouza02]

​@@MikeSharpeWriter Yea just saw on Reddit... Windows users also having the same issue :/ When I was on Windows this year, if I'm not mistaken, in version 126 of Firefox there was a giant memory leak that could consume up to 40GB if I watched a lot of KZbin shorts in sequence. It seems that Firefox has been having problems with KZbin.

[@Mooooov0815]

@@VictorSouza02Firefox seems to be constantly leaking memory. I had Firefox eating up 20+ GB of RAM with only a few tabs open multiple times now. I really want to like Firefox, but they’re making it pretty hard sometimes

[@bekiraltindal9053]

Strange. Firefox works much smoother on my 10+ year old machine than Chromium. Firefox seems more resource friendly.

[@Frank_Erz]

My only RAM problem is that Firefox tends to use so much RAM that sometimes I can't launch certain games

[@lagomoof]

How many tabs do you have open? If it's a lot, it's not all Firefox's fault. Consider a tab unloader extension, or using bookmarks.

[@Lampe2020]

I tried installing various Linux distros on my Micro$oft Surface 3 (which my web programming teacher had lying around and recently gave to me) but it never worked, because it always used ~3.8GB of the available ~4GB, right after boot. And then, depending on the distro sooner or later, it would hang. The various Androids I tried worked well enough to be installed but not daily-driveable and Linux Mint held up for half an hour before freezing. But Fedora and Ubuntu with various desktops always would freeze no more than two minutes after starting or when I entered my username in the installer.

[@JodyBruchon]

To quote The Fifth Element, *"GIMME DA CACHE!"*

[@bigpod]

fun fact windows caches even more aggressivly and actually just reserves memory at start it thinks you wont need and releases fi needed for apps

[@utahnl]

Windows is also very reluctant to release disk cache and reserved memory, often forcing active applications to go into the pagefile.

[@Mooooov0815]

The issue is that windows is pretty shit in freeing RAM as memory pressure increases. I am dual booting right now and my Linux Desktop feels substantially slower in any regular task than the windows install that is using 2.5x the memory.

[@thedave1771]

@@utahnlis it? I haven’t seen that at all. I’ve got 64GB and babysit it a bit, I haven’t seen writes into the pagefile. You might be being confused by pagefile allocated vs pages written, when an application requests 1GB of memory from Windows, Windows will carve out space somewhere, but that might be pagefile-backed as long as it is empty, to avoid evicting anything from the cache. As soon as you actually store anything it’ll use up cache space. This will cause your pagefile to grow if you’ve misconfigured it, Windows won’t allocate virtual memory until it has space to store it somewhere (RAM or disk), but the actual cache release is the moment an application uses the memory, not when the application requests it. It may also proactively write pages to disk (while still keeping them in RAM) so that if that RAM is needed it’s available instantly, but I think it either doesn’t do this on SSDs, or was dialled back significantly for systems with a ton of RAM, or both. Actually I bet it’s something like when it has less than 512MB or something, that takes into account the current rate of allocations, or whatever.

[@harpskid]

You can still track the uncached memory and i do care if that numbers big. There isnt anything wrong with trying to use software that uses less memory, and it allows me to have lots and lots and lots of background programs that i actually want to use that memory. Im not a ram usage minimalist, i just want it used for things i care about.

[@H3cJP]

windows is even more confusing, there is: in use(compressed) commited cached paged pool non-paged pool available

[@moarjank]

Linux also has these - under different names. I don't remember how to get these values, but it's in /proc and /dev

[@fireztonez-teamepixcraft3993]

I have never seen this article before, but definitely really interesting and useful to know. I do remember when I first tried linux, being a bit confused about why so much of my memory was in used, but one of my friend explain to me the reason back in the day. Definitely good to know, especially when setting up VM, you should always assumed that the memory assigned will end up being fully used. I totally agree with the; unused RAM is wasted RAM!

[@SIMULATAN]

Hipedy hopedy, your RAM is now my property! ~ that darn penguin, probably

[@ryebread095]

I avoid posts about ram usage on help forums because the vast majority of people don't like hearing that it's not a problem. Not worth the headache for the small number of actual ram usage issues.

[@HumbleBountyHunter]

About the network caching, Firefox since v59, has done things a little differently. > The Race Cache With Network (RCWN) feature races the cache with the network when the cache is slow. So if reading from the disk is slow, we will send a network request, and return the channel from the network, even though we have the entry in the cache. RCWN Heuristics If the request: - is not using appcache - is smaller than 256 kb - doesn't have alt-data Then: - if the cache is slow: we race immediately - otherwise: set a timer for 3 * avgCacheTime

[@IuliusPsicofactum]

It is not your linux, it is your browser.

[@JGrffn]

Hi, my system with 64GB of RAM is ACTUALLY using too much RAM. I'm using EndeavourOS on a 2020 Razer Blade 15 Base model. It's a laptop with an Intel CPU and an Nvidia 1660Ti, using Hybrid mode, and I have it connected to an Alienware AW3418DW monitor and an Acer 1440p 75hz normal monitor (among other stuff) via a CalDigit TS4. I'm using KDE. I have a bunch of issues where I suspect the Nvidia GPU and maybe the dock are adding some fun stutters (my refresh rate visibly goes down whenever things like Dolphin are focused), hiccups (randomly the system just stops updating for a second or two, and this keeps happening somewhat frequently once it starts happening, until it suddenly stops doing happening entirely), slowdowns (youtube videos become extremely sluggish and slow to respond to inputs), etc. Those I can [somewhat] live with for now since I'm planning on switching to a laptop with an AMD CPU with integrated graphics, and a desktop with a 9800X3D and a 7900XTX, and because I understand that both Nvidia GPUs as well ass discrete/integrated graphics switching on laptops is sketchy, but my actual issue - I believe - stems from my Acer monitor. The monitor has always had issues with sleep: when it stops receiving a signal it tries to go to sleep but keeps waking up. This happened back when I used Windows, this happens even when everything is turned off, but now that my laptop is using it, the laptop keeps receiving the plugged in/unplugged signal from it, which means it keeps notifying me every time the monitor turns on and off by itself. The issue, I believe, stems from SOMETHING in kwin_wayland freaking out due to this on/off signaling, which in turn causes a memory leak on kwin_wayland. The result? If I forget to manually turn off the secondary monitor whenever I step away, once I come back my laptop will be consuming hoards of RAM and will continue to do so until my system runs out of RAM and something crashes. I usually manually kill kwin_wayland before it eats up all my RAM, which either kills all my apps and restarts the UI, or puts my computer to sleep in a weird state that I have to do a mayan rain dance to get back from, or it just crashes the laptop entirely and I have to reboot it. It's not fun. I THINK the issue has been reported already, since I've seen some logs on my machine that other users have posted as potential issues on the KDE bug tracker, but I'm still too unfamiliar with these things to actually know for sure that it's the same thing been reported, nor would I know how to actually report this properly if I had to do it myself. It started happening a few months ago, so I'm kinda hoping this has been noticed and it's being worked on..... Still a better experience than Windows :v

[@aerospherology2001]

KDE can have memory leaks on Nvidia GPUs

[@xanderplayz3446]

Opinion: Swap is great if there’s a lot of data in RAM that doesn’t change often and isn’t accessed often, such as chunks of a Minecraft server, functions related to a feature of a program that you don’t use, or state of a peripheral that’s not being used.

[@stevethepocket]

Mint's system monitor actually treats cache as completely separate from actual memory usage, reports them as separate numbers, and only shades in the actual used memory in its little pie chart.

[@timothyt.82]

44 seconds is a new record for me.

[@Linuxdirk]

I WANT Linux to use my RAM. That is what RAM is for - using it. Cache all the things!

[@Simulati0n]

Horrible take, I want whatever I use to take ram. What if I didn’t open any of those apps that it has cache due to behavior patterns. I shouldn’t see 12GB ram use before launching any game. It’s almost like a point where 32GB is no longer good enough

[@AnEagle]

​​@@Simulati0nit can cache it at a low priority, and just release it whenever a high priority process asks for ram

[@Simulati0n]

@@AnEagle my point still stands even on a fresh install of cachyos 11-12GB is use by default. When running PoE2 that’s another 10GB use. At this rate I’m already pushing 70% usage.

[@svgaming234]

@@Simulati0n and? in case any applications need the extra the ram will be freed for them. did you even watch the video?

[@moussaadem7933]

​​​Nope, you can practically assume the 11GB are unsued. they have no harmful effect whatsoever. they only make the system fast. your game is allowed to use that 11GB whenever it needs it, the system isn't going to prevent an app from using RAM in favor of cache. Also, guess what ? while your game is running, if there's RAM left, the system will store things like shaders and skins because it notices that the game is accessing these files often. Your game will be extremely slow without caching, same for the rest of your system. What's the point of preventing the system from using RAM to speed things up when no APP needs the RAM and knowing the system will immediately give it back whenever an app needs it ?

[@SilkCrown]

I first came across that post when I was trying to figure out why my system was freezing and I suspected it might have been an issue with a program using up too much RAM.

[@tato-chip7612]

turning down to swappiness makes more sense for HDDs actually. nowadays with SSDs the performance is much better even with a lot of swapping

[@RealHellession]

I don't think people worrying about high RAM usage are unreasonable. For me personally I never had issues since switching to Linux, but before I decided to do so and upgrade my PC, I was often struggling with a scenario where my (Windows) system would start seriously stuttering and when I checked Task Manager, it would report near maxed out RAM. This was happening often on my 16 GB system, while I was game deving with Unity Engine, Visual Studio and Google Chrome with KZbin playing on other screen all on at the same time. Ever since I would check and make sure my RAM usage reported wasn't too high. High memory usage essentially meant "okay so I can't open more shit willy nilly, let's just be mindful of that". With that base assumption, if I was suddenly presented with 99% RAM usage, I'd be worried as well - not just because "big number bad" as you imply in the beginning. I think I looked this up at some point too, though during one of my first times using Linux ever for some VM or VPS. The misunderstanding largely originates because it wasn't as obvious to me that Windows did the same thing. It's good to learn about disk cache on RAM... just that I don't believe people fearing about (implied) high RAM usage is unreasonable. It only becomes unreasonable when they think disk caching is bad

[@klightspeed]

There was also a case with one of the recent kernels with BTRFS where being low on free memory (even if a good chunk of the used memory was clean cache) would cause the BTRFS filesystem worker to be not nice and cause stuttering due to using 100% CPU on a random core at a high priority. Dropping the clean caches would de-trigger its non-nice behaviour.

[@markozagar]

An interesting example for me was ZFS, which can use significant memory for ARC (a kind of cache, but not one that can be dropped whenever). Unless you know where to look, that memory just seems "gone", and clearing buffers/cache won't get it back. It shouldn't be a problem under default options, but it becomes really noticable if you enable e.g. deduplication. Luckily, that's not a default option even if some distro might use ZFS by default.

[@botondkalocsai5322]

I surmise that the situation is a little bit more complicated from my experience. When I cold boot into console, my base system shows around 180 MiB used from free -m. After I launch my desktop environment, the used memory becomes around 230 MiB. ps --forest -e -o comm,rss tells me that Xorg itself uses around 96 MiB (+dwm use 7.3 MiB +slstatus use 2.25 MiB ). The numbers does not seem to add up since, from free -m I only see around 50 MiB usage increment after startx . This means that the base system allocates a lot of memory, that it relinquishes to the desktop environment.

[@andrewwojtas8486]

0:07 the only distro I had an issue in regards to ram usage that seemed unusual was on fedora ashai, it has a lot more ram useage than normal

[@TestAccount666]

Very informative video, thank you! :) I recently started having issues with my install though :/ Sometimes, everything starts to be slow and buggy as hell. Want to open the terminal? Take a coffee and it will be done at the second one. Want to actually write something in the terminal? Time for your third and fourth coffee. Oh, don't you dare move that terminal btw! It just randomly starts and fixes itself eventually. During the issue, RAM-, Disk-, CPU- and GPU-Usage is completely fine...

[@prashantkumarsrivastava1977]

There used to be ways to load the entire kernel in ram , ramdisk for blazing performance. It was bragging points among Linux users.

[@stupidburp]

I set up a Mac video editing station to do this about 25 years ago. It took some tricks on startup but was possible at the time and definitely improved performance.

[@Kotz_en]

I think it's pretty funny that none of these people would be complaining if both Free and Available memory were reported as just "Free". Let Linux cache all the things whilst reporting this memory as "Free" so the whining stops.

[@akahenke]

Thank you so much I always wondered why some production servers were running out of ram whenever badly written applications on 32 bit systems were creating excessive amounts of files that caused the 4GB of ram to get used up.

[@littlek3000]

Having 96GB of ram, and after launching a few games, I regularly see my cache use as much as 20-40GB, depending on the games I launch. Like Helldivers 2 will essentially load every map it can into cache and after a few mission cache usage will jump to 30+GB. Same with Rocket League and COD, or seemingly any game that has maps that load/unload.

[@linuxguy1199]

Recently added 384GB of RAM to my fileserver to get that oh so se(x)y ZFS RAM caching on all my linux ISOs.

[@orbatos]

The way "free" raml is calculated and displayed is a big causal issue here, but often enough a bigger problem is how the information is displayed across different desktops and operating systems. Windows, which most Linux users are coming from, only displays "real" memory usage in summaries, the opposite of tools like top.

[@mattelder1971]

For quite some time, TrueNAS Scale had an issue with its disk cache where it wouldn't free memory that was in use for the cache if, for example, you wanted to start up a VM that needed some of that memory. It seems to be fixed now, but at one point (not too long ago) it was a real problem.

[@NiffirgkcaJ]

Oh my goodness! Am I the only one having this issue? My system always crashes after idling for a few hours! For some reason, the swap is always being filled and not getting cleaned up automatically after closing applications! I always need to log out or reboot, because log out never cleans the entire swap memory. 🥺

[@MohammedKazim]

Pro tip: Use zram instead of swap or zswap, and set the max actual size to your RAM minus 1GB, or max theoretical size to 3 times your RAM (you can go more if you cap the actual size), and swappiness close to 200. Now your system can cache in your RAM and fast, and you can use a lot more RAM since it will be compressed. I used this to compile large programs with full LTO.

[@seancondon5572]

14:00 wow. Just noticed you have 32GB of RAM. The only systems I have with that little (or less) are Raspberry Pis, laptops/chromebooks, and 2 decommissioned intel rigs from 2013 and 2015.

[@seancondon5572]

The decom'd intel rigs are technically still fully usable. They just don't have NVMe and run DDR3 and PCIe 3.0. Not really worth keeping active.

[@BoDiddly]

I had a Memory problem with Chrome once. It wasn't quite a "memory leak", it was that when you left tabs open, the ads on the websites would get cached in the background, so leaving the browser open with lots of tabs, would slowly eat up memory over time. It appears that Chrome fixed this by shutting down tabs that aren't in use for a certain amount of time.

[@ApeironTsuka]

In my particular case, it seems to be an actual leak somewhere I can't pin down as it seems to takes multiple weeks to creep up. 50% of RAM in use right now (not buff/cache/shared) with Xorg being 15% of that. The rest? I have no idea. The next-highest is ~5%. I can have zero applications open and still have 20gb of RAM in use. I can log out of my DE, dropping back to lightdm greeter, log back in, and all that RAM is now free, so I suspect it's Xorg/amdgpu related. Maybe. It's not TTM pages not getting freed (already checked), at least. Works as an excuse to finally apply that pending minor dbus update and reboot, I suppose, ha.

[@anonymouscommentator]

i kind of have the opposite problem. i have 32 GB of ram and after booting up, my system takes not even 2GB. at the same time, applications have a slow start. i tried looking into loading applications automatically into cache but i didnt find anything

[@Roxor128]

When it comes to disk cache, I want a second level of it that persists between boots. Main storage on spinning rust, L2 cache on flash, and L1 cache in RAM. The capacity of a HDD and the speed of an SSD for what gets regular use.

[@michaelwright2986]

It's not just a Linux thing--I first learned this in connexion with Windows--but maybe people especially worry about it with Linux, because one theme is the search for the Lightweight Desktop Environment, where lightness of weight is measured in RAM usage. Which I am sure is grossly oversimplified, unless you're wanting to run Linux on an actual 2009 laptop with 2 MB (or 2 MiB).

[@GeorgeGzirishvili]

What's even more annoying than people constantly complaining about disk cache using their RAM is self-proclaimed experts claiming that your system is all fine when you got a memory leak and you're asking legitimate questions.

[@hacxman1]

i truely love where we got with ram consumption. i"ve seen several times how firefox ate 55G of memory managing to bring system with 64G of ram to crawl. its so bad even with swap, more if you have one and forget to pay attention to it. i"ve ended up with system thrashing swap for hours at a time so many times. i cant even... the faster the swap is the worse problem it creates when you run out of it. there must be something very wrong with kernel oom today. i used to run computers low on ram with no swap at all, because oom was so fast it identified offending process and killed it within miloseconds, but past years (at least 7) it doesnt do its job properly anymore

[@wrathofainz]

I've actually been having an issue lately. Where when my computer wakes up From sleep itll start consuming more and more Ram until it either reaches a certain amount and resets down to minimum and stops or crashes the p. C.

[@randyriegel8553]

RAM is made to be used! A small company I was consulting for said the SQL Server was always at 90% memory. They were freaked out and I said it's fine. SQL Server is made to take as much memory as possible for caching. The server had 64GB of memory... they didn't believe me so I left. Couple weeks later they call me and said they upgraded to 192GB and still has high memory usage. I tried to explain it to them but they still didn't believe me. So I gave up and didn't go back.

[@Kneedragon1962]

Thank you, very informative. I would add one thing. There are several monitors or tells about system status, which report the memory use numbers differently. This can be and is very confusing. So, let's say I have Mint + Mate and I have a conky on the desktop, which reports a number of things, including memory use. I also have the Mate System Monitor ~ which reports memory use. I have terminal aps, like tops ~ which report memory use. All of these can and do report wildly differing figures. I can fire up Virtualbox and start a Linux guest. I can ask the guest what it's memory use is, and then I can check on the host, how much memory that instance of VBox is using. And sometimes these figure line up fairly well and sometimes they don't. Windows guests are a lot worse. Final complaint ~ I have 32GB of physical RAM, and the rule of thumb I was taught back in the day, was the swap size should roughly equal the amount of system RAM. Ok, so try this stuff out ~ let's set the swapfile size to 16GB and start doing stuff to load up RAM. Up to about 4 GB of swap in use, not a lot changes, but by the time you get to 6GB (six, not sixteen) the machine slows to a complete crawl and effectively locks up. the 'fix' is to pull the plug and count to ten and start again. Why does a swap with less than a quarter of my true physical RAM result in a locked up machine? Adding a bigger swap again, makes no functional difference. Alright ~ next question ~ What is getting loaded into the swap that obviously should have stayed in physical RAM? How can I change that so it swaps out stuff I'm not likely to need or use, (like old web pages in a browser running in a virtualbox) and keeps things like the Linux kernel in the host entirely in system RAM? I hear all the time about the Windows sheduler trying (failing) to load all the parts of a game into cores on the same CCX, the one that has the 3d cache ~ surely there is something like that in server world / rack-space to not just change 'swapiness' but to control what gets swapped and what stays in physical RAM? Like - What is important to keep in physical RAM and what is not? And while making requests, can we get all the dozen or more things that report RAM use, to agree on what it means to be used or not? Because the way things are now, six different monitor / task-manager type programs will give you six different answers to how much RAM you're using. Conky gives an answer, different to tops, which is different to neofetch, which is different to fastfetch, which is different to Mate System Monitor... There are a LOT of tools in Linux to report or monitor RAM use, and every single one of them gives you a different answer.

[@TeknoMage13]

Every time without fail I have had worse performance when I get low on RAM. Not directly related because I'm primarily a windows user with a Linux box that I play with. On my main windows machine I have 64 GB of RAM and I need to reboot every few days to free up RAM and restore system performance. This has always been the case in my experience over the last 30 years of using computers. I've heard people claim that it doesn't impact performance to have it allocated and it isn't actually used... if that was the case then performance wouldn't improve after a reboot and it does. Once again, I realize this is talking about Linux and I don't currently use it enough to know if it impacts performance in that scenario or not.

[@CloudCuckooKing]

Oh, thanks to you, Brodie, I know I'm eating up 20 gigs of my RAM by doing nothing at all! Wow, I might never have found out! I'll make sure to go and fix this right away! j/k j/k I set this thing up to build Gentoo on it I have 128 GB

[@MarvinRB3]

using good utils helps users understand this. e.g. htop does a good job of highlighting the difference between used RAM and cache

[@TheDireLynx]

got curious and checked my own system. i knew that caching was a thing and would eat up a lot, but holy cow I didn't expect this much. 12.6 GiB used, 18.7 GiB cached, 691MiB free.

[@CKOD]

Me looking at my trueNAS install, wondering why the disk cache is only 159GB where there is still ~7gb free.

[@johanngambolputty5351]

Come to think of it, is there a way to view how the kernel/different kernel modules is using ram?

[@averagetechnologyenojyer]

Browsers too consume a lot of resources

[@Matrox473]

Swap is also OP when you rent cheap VPS. You often get like 2G of ram but you can make a fat swap. Since VPS run often on really fast ssd, it leave the system usable for most task.

[@NiffirgkcaJ]

Oh my goodness! I thought I am the only one having this issue! My system always crashes after idling for a few hours! For some reason, the swap is always being filled and not getting cleaned up automatically after closing applications! I always need to log out or reboot, because log out never cleans the entire swap memory. 🥺

[@fogcat5]

this question comes up in interviews from time to time and I can tell them all about memory allocation in linux and how the page cache works. doesn't explain how much ram chrome uses though ;)

[@buddybleeyes]

Reminds me of a friend who used to panic whenever her cpu temp would hit 85 Degrees C on an older i7 on a cheap cooler.... it was not thermal throttling, and it was a reasonable temperature for running games at for her system. She would have the monitoring up for every game and then start to freak out when it crept up to 82, and full flip as it would sit at 85 🤣🤣🤣 edit: yes we told her to turn off temp readings, yes we told her it was perfectly fine at that temp as long as it wasn't higher than that.

[@johnpekkala6941]

This is a problem I have had like forever with my Linux game dev machines. They run Unity, VScode/Rider ect with no problem at first but after I have been working for a while the entire machine will suddenly lock up and the hard drive goes like GRRRRRRRRRRRR constantly on the swap/virtual memory because the OS is for some reason gradually filling the entire RAM with crap. I have between 12 and 16 GB at least on these computers so should be enough as they run just fine freshly rebooted even with all my working programs loaded like the game engine, code editor ect. When this thing happens I cant move the mouse or anything. Not even the clock updates, the computer is completely frozen churning away on the virtual memory and I have to kill the power to the machine to get out of it. It seems to me like Linux in general have BIG issues with memory leaks. This could hopefully be a solution Im gonna try to see if i can finally find a fix to this VERY annoying problem. I rather have some performance penalty then loosing both work and time by being forced to pull the plug on the computer over and over because it locks up completley all the time as the RAM gets full.

[@MaxLothaire]

I used to have performance issues with ZFS on a linux desktop. ZFS cache seems to be seen as used memory by the kernel, and ZFS did not free its cache fast enouth, causing some games to slow down. I solved that by reducing the cache size. ( and by switching back to ext4 after an OS re-installation)

[@DavitTheCore]

How is everyone saying that RAM is being eaten by Linux, when my system barely went to 12GB of usage finally when compiling PPU modules in RPCS3, and normally I don't even pass 1GB of usage. I do not have any swappiness changed, and I am running GNOME. GNOME, the famous RAM eater. I'm not sure if 1-2GB of RAM usage is gonna hurt anyone.

[@LilithNephilim666]

I didn't know about that website and now I remember all the times I had to explain disk caching to friends when I could have just send them that website

[@xymaryai8283]

when its overusing my system Swap too, and causing thrashing issues even with a low swappiness, i think it might be something wrong. i only have 14GB of usable physical memory, a single game shouldn't be using 20GB. the game itself reports 8GB, and system at idle reports 2-4GB. what is the other 8GB doing? i thought it was spillover VRAM, but that maxes out at 6 GB. if its disk caching, why is it causing my game to crash?

[@13thravenpurple94]

This video was superb! I really appreciate your effort! 🙌

[@chaos.corner]

Swap means idle stuff can be swapped out and the RAM used for even more disk cache so it's worth configuring even if you have plenty of RAM. The benefits do get a bit slim eventually though.

[@CGwatcher]

Funnily enough, I still use the drop_caches approach sometimes, to force WSL to release as much memory as possible when I know I won't need it for a while.

[@JessicaFEREM]

anything under 100c is perfectly fine for a computer to reach. especially modern computers that are almost always pinned to that temp to begin with. if it's reaching over 100c, then you are causing damage. ideally you wanna stick to under 90c.

[@derekday4832]

When I saw the big RAM numbers I was surprised. My first thought wasn't "that's a problem - must fix it." It was "I'm a Linux noob. The system is probably fine. It's probably my understanding of it that needs work. Either way I need to learn more so I don't try to fix something that ain't broke."

[@rahulvadhyar2378]

Fedora KDE with 8GB of ram is partially a disaster for me. Things are slow, sometimes OOM killer kills process etc. Plus I want to compile code that can be compiled on a more minimal system like hyprland, but not on KDE. Sometimes it’s just truly bloatware.

[@Warrentheo]

Just always be crazy like me, and max out the motherboard ram of every computer you buy... Been running 64GB on both my laptop and server for about 7-8 years now...

[@CKOD]

Untill DDR5 was a thing, I was in the same camp. If I ever ran low, I truly knew it was time for an upgrade. But with DDR5 being fussy about 1 or 2 DIMM per channel, and how much AMD platforms benefit from RAM bandwidth, I'd say stick to maxing it out at 1 DIMM per channel as a generic guide.

[@PokettoMusic]

my problem is 32gb of ram isnt enough anymore, I'm constantly running out of available memory

[@w4gap]

Windows users have been lamenting this exact thing for decades. ;-) I used to be an advanced professional services consultant for Brocade Communications and some users always complained about the very small amount of memory shown as available when they were monitoring our FibreChannel switches which were built on Montavista embedded Linux and our "FOS - FabricOS" actually was taking all of the memory to manage internally actually making the performance of managing/processing of Fibre Channel traffic much better than if managed by the OS....