>>108637757
how much ram were you supposed to have? 64 gb?
cause you've counted the missing memory as "1.2" in your image filename as well as on the thread topic but as "1.4" elsewhere so I'm a little confused

>>108637786

I have 64 GiB in total.

If i boot memtest86+ i can utilize 63.8 GiB

So in total 1.4 GiB is unavailable in linux which is 1.2 GIB more than it should be.

>>108637757
Are you using ram as cache for web browsers? anyway no idea. Something is clearly allocating that ram for hardware or program.

Let me guess you have integrated GPU in your CPU....

if you have dedicated one, just disable the integrated in BIOS

simple as

>>108637757
It's troonix. You're lucky it even boots without throwing fifteen kernel errors

>>108637757
If you have an igpu it can reserve anywhere up to 2GB maybe more. Do dmesg | grep "Memory" and look for a line that says reserved to find out what's allocating it

>>108637867

It doesn't show up in mate-system-monitor and isntead i simply says i only have 62.6 GiB of ram even though i paid for 64 GiB (it should be exactly 64 GiB physically).

But the actual amount of ram that can actually be utilized (without resorting to a something like SWAP) is of course less than 62.6 GiB, Max is around 62300 MiB last time i checked (due to XFCE4, etc).

>>108637894

I bought a CPU without one (13900KF).

GPU: RTX 3090

>>108637903

Seems like it's reserving a lot only to then not actually use it.

[ 0.000000] DMI: Memory slots populated: 2/4
[ 0.157135] Memory: 65373456K/66935632K available (20562K kernel code, 2951K rwdata, 16740K rodata, 4776K init, 4600K bss, 1517888K reserved, 0K cma-reserved)
[ 0.157896] x86/mm: Memory block size: 2048MB

>>108637757
is it a different unit? the missing amount is really close to some 1000 vs 1024 shit.

>>108638022
this was my thought as well

if you guys didnt know OP is a mouthbreather who sources framechasers and random schizos

>>108638022

No.

>>108638022
No. First, that's for storage. Second, 64GB is about 59.6GiB, so that conversion wouldn't even account for this amount.

My uneducated guess is this is likely memory held by the kernel for various tasks. Eg: your SSD doesn't have cache and is system memory mapped for its lookup.

Retard

>>108637757
this is memory mapped space, you can sudo cat /proc/iomem to see how it's been carved out.
hardware often needs to interface with the system through memory mappings
in my case, i have 256GiB physical but get 251.2GiB in userspace, approx 2~3GiB being PCI/MMIO, and 1~2GiB being ACPI/firmware/reserved
>if you run out if it the system will halt so badly you can barely use it at all
make sure you have an oomkiller on!

>>108638342
Why would these use up physical memory? The whole point of virtual memory, apart from separation, is to avoid these sorts of things.

>>108638399
External devices can request ram instead of virtual memory.
For instance smart access memory/rebar gives the gpu access to the entire ram pool instead of only chunks. NVME drives without cache will use your ram as cache and its up to the OS to decide how much. On Windows by default its 64mb, other Linux OS may differ.

tmpfs ramdisk for things like pid's and sockets

tech support thread

>>108637757
https://kernel-internals.org/mm/memory-reservation/

Note: your GPU RAM might be in the list below

Kernel Memory Reservation Mechanisms

Not all physical RAM is yours to use — the kernel, firmware, and hardware carve out regions long before user space ever runs

Why reservations exist

Physical memory is not a flat, uniformly available resource. By the time the kernel reaches start_kernel(), several parties have already laid claim to portions of the physical address space:

Firmware tables — the BIOS, ACPI tables (RSDT, XSDT, DSDT), and SMBIOS data reside in RAM and must not be overwritten

Kernel text and data — .text, .rodata, .data, .bss, and related sections occupy fixed physical addresses set by the bootloader

EFI runtime services — on UEFI systems, boot services memory and the EFI memory map itself need protection

EBDA and legacy BIOS region — the Extended BIOS Data Area and the top of the first megabyte are historically corrupt-prone

Device memory and memory-mapped I/O — firmware may describe ranges as E820_TYPE_RESERVED rather than RAM
crashkernel regions — a dedicated region held in reserve for the kdump capture kernel

CMA pools — contiguous memory for DMA that doubles as ordinary movable memory until needed

Hardware quirks — some chipsets corrupt specific physical ranges (Sandy Bridge iGPU is a canonical x86 example)

At most, there are only 3 humans in this thread. Myself included. Every other post is literally an AI training module

>>108638551
>Myself included
sounds like what an AI would say

>>108638505
>https://kernel-internals.org/mm/memory-reservation/

Doesn't answer why windows 10 makes less than 200 MiB unavailable and still works fine.

>>108638430
Sure, but a gig and a half of this for commodity hardware?

>>108638623
Prove to the court Windows does in fact not lay claim to those memory addresses and simply obscures the fact it is doing so from the user.

Pro tip: you can't

>>108638623
Windows is misleading you because it doesn't report its kernel memory in the "hardware reserved" section

>>108638210
>>108638108
64/1.024 == 62.5
You are a faggot who can't do math

>>108638714
It's off by that factor for each unit, retard. So congrats: you just calculated kilo-mebibytes.

File: 1260065887014185.jpg (41.4 KB)

41.4 KB JPG

>>108637757
someone tell this absolute normie newfag the difference between counting bytes in 1024 and 1000 units?

>>108638636

Windows 10 does still let you use most of your ram (like 63.2 GiB) by moving loads of stuff to virtual memory.

Less than 200 MiB is reserved.

>>108638798

I have 64 GiB of ram (68719476747 bytes).

btw: 64 GB is 59.6 GiB not 62.6 GiB.

>>108638628
unused ram is wasted ram. you want to do as little virtual swapping as possible to prevent hard faults.

>>108638846
>unused ram is wasted ram
Ah yes, this explains perfectly why the kernel reserving a few gigs of space for a look up table makes sense. Thanks for your well thought out reply.

https://www.linuxatemyram.com/

>>108638846

Ram that shows up as unused in mate system can still be used as cache which should improve performance.

>>108638869

Not how the mate system monitor reports ram usage.

>>108638859
as opposed to storing all that information in page and making your cpu do a ton of pointless interrupts

>>108638845
no.

You have about ~64000000000 RAM
since MB is kilo based.

and in MiB it is ~62500 MiB

which is more or less what your OP pic shows.

>>108638869
Dipshit. The reserved memory wouldn't even show up there. It reports only usable memory. Read the fucking man page.
>>108638888
>in page
The issue is why the kernel needs gigs worth of pages, not why it is reserved. Stop being obtuse.

>crying about RAM when you have 30+gigs available

File: images.jpg (5.7 KB)

5.7 KB JPG

>>108639033
Yeah, exactly! You have 25+ available, what do you need 15+ for?

>>108639049
>waaah I am "missing" a gig of RAM
>I STILL have more free RAM than most people have total
>I'm gonna cry about this

>>108639071
>paying for RAM so you can watch the kernel play with it
What did anon mean by this. Is this a form of cuckoldry?

>>108638604
kek, I'll add you to the list of human folk

>>108639096
What do you think a kernel does?

>>108639156
Ah fair point. Inflation fetishists then?

>>108639165
Windows/NT kernel "plays with RAM" as does the Darwin/XNU kernel on Mac, it's literally what the kernel does, interface with the hardware.

>>108639213
It's supposed to make a useful system. Swallowing up your entire physical address space for record keeping isn't it. Sad I have to explain this to you, but you're that special kind of autistic retard who can't infer anything. I anxiously await your next low-int nitpick.

>>108638909

Some people might think you are just trolling at this point.

But you are probably that stupid for real.

62.6 GiB is more than 64 GiB retard.

File: Screenshot_20260419_194758.png (51 KB)

51 KB PNG

>>108638210
>>108638714
>>108638764

>>108638714
NTA. But please learn how to use google. pic related.

regarding the question of OP: Google MMIO. tl;dr the kernel might reserve a space of RAM for your GPU (Not your iGPU) so it can map the memory and make it easier to move the textures, assets, etc to and from VRAM.

>>108639236
Nah "girl" I'll let you wallow in ignorance

Troll thread
Sage
Your 200MiB is in my asshole and you can free it up by kissing my belly button from the inside
Nigger
go back to windows

>>108639358
There's no need to get embarrassed now that I called you out on being functionally illiterate due to your genetic defect. Go ahead and say the line, autist.

File: 1753239456354496.png (7.4 KB)

7.4 KB PNG

>>108638810
Works on my Linux machine

>>108639417

32.7 GB is 30.45 GiB meaning that there is 1.55 GiB being reserved in your case.

>>108639449
no? I bought two 16GB sticks and I have 32GB
>redditspacing

>>108637757
On my system Windows reports 15.2 GB in task manager as available
On Linux it reports 14.9 GiB available in system-monitor and 15323752K/15986300K available in dmesg, and 15582212 in free
15986300K is 15.2 GiB
15582212, is 14.86 GiB, which rounds up to 14.9 GiB

In Windows 1KB = 1024 bytes, so I think Windows is just reporting your total memory as usable while Linux reports accurately.

>>108639381
Back to Windows kiddo

>>108639465

You paid for 32 GiB but in linux only 30.45 GiB is accessible.

File: 67.2 GB ram.png (7.6 KB)

7.6 KB PNG

gnome system monitor reports the ram usage in GB (not GiB) and unlike the mate system monitor it also reports how much is being utilized as cache.

File: 1756169676396342.png (4.4 KB)

4.4 KB PNG

>>108639512
I paid for 16GB just like it's advertised

>>108639520

You kinda got more than you paid for (it's technically 34.3597 GB = 32 GiB).

>>108639532

All 32 GiB of ram is there. You just cannot access all of it due to linux being inefficient when handling it.

File: GSM.png (25.7 KB)

25.7 KB PNG

>>108639517
Then change it to GB.

that's where the nsa put their spyware. it's got a large event cache.

GiB notation is so fucking gay

>>108639694
Computers are binary.
Base 10 is marketing.

>>108639694

No it's superior and all alternatives are stupid.

>>108639737
the unit inherently tells us the correct notation

anyway i missed it under all the "you're holding it wrong" kvetching, why is linux stealing op's ram?

File: 200MB missing.png (17.3 KB)

17.3 KB PNG

>>108637930
>>108639803
Because it isn't.

>>108637902
are these kernel errors in the room with us right now?

>>108638505
>lincucks just accept this bs laying down
this is why i will always use windows

>>108639858

The 63.83 GiB is only available if i boot memtest86+ (which isn't linux) or windows.

If i boot linux there is only 62.6 GiB available to use.

>>108638859
eviction exists and mappings will scale down with ram pressure. these things aren't static. unless you're nearly out of ram worrying about these numbers is retarded, and complaining without actual knowledge of how the kernel works even more so

>>108639941

I have not seen it scale at all. The amount of usable ram after boot is capped to 62.6 even if there is a bad freeze due to running out of ram.

>>108639929
Anon, nobody here knows shit. Use the slabtop command, which shows kernel memory. eg:

sudo slabtop --human --sort=c

This place is a cesspool of phone ricers.

>>108637757
> 1.2 GiB of my ram
That's more than my old Linux Box with Debian Potatoe even has.
Setting it up was so much work that I never felt the wish to upgrade.

>>108639976

I am surprised i got even 1 useful reply here (which is still better than reddit).

The main culpris is radix_tree_node which takes up 1238304

>>108639929
Because you didn't buy 64 GiB of memory but 63.8 GiB.
Windows and memtest report your TOTAL memory in GiB (1 GB = 1 GiB in those) as available.
Linux is reporting 63.8 GiB, minus allocations for BIOS, kernel, and other shit, as available memory.

>>108639911
accept? they immediately go full "it's not happening and it's good that it is"

In Windows I have 63.9GiB of RAM. Maybe you should just get a less wasteful OS.

>>108640108

Windows does use a good amount of it by default though so you end up having to use the pagefile (which will wear down your SSD) to actually utilize more ram than what you can in linux.

>>108640106

I had someone in reddit claim that it would be bad for stability to let you actually access all your ram and have system processes potentially be moved to a page file.

But as far as i have seen the stability in windows 10 is fine even if you use like 63 GiB of ram. The only problem is the reduction of the SSD lifespan.

>>108640034
1238304 what? It defaults to KiB. If so that's pretty fucking big for the radix tree. AFAIK, its only used for key-value lookups (think hashtable without the hashing). Not sure what module in the kernel is using so much but it feels wrong. For reference mine is around 140 MiB on 32 GiB.

>>108640151

Yea there is obviously some pretty bad bloat somewhere.

>>108637757
Windows does the same shit. It depends on how much you've set as a hardware reserve. Doesn't help windows and linux count space differently.

>>108640163
Let me just say that's not typical. My workstation uses btrfs which by default uses a ton of kernel space memory for caching. My laptop is in the 200MiB range for the whole kernel. What distro is this? What filesystem are you using for your partitions?

>>108639975
>radix_tree_node is the main cuprit
well it looks like the right thing to tweak is vfs_cache_pressure, then
question is, what are you running where that extra 1.2gb out of 64gb will actually make a difference? you'll still be running oom and your system will continue to freeze anyway
now admit this thread was just bait

>>108640178
Also, do you use virtual filesystems (VFS)? I don't, so maybe that's where it's coming from.

>>108640186
He's probably generating slop. That will eat memory up like a motherfucker.

>>108640197
the fuck you mean you don't use VFS

>>108640208
Shit, my bad. I'm low on sleep. Meant virtual volumes, lol.

>>108637830
>I have 64 GiB in total.
no, you have 64 GB in total, GiB != GB

File: 67.2 GiB.png (7.4 KB)

7.4 KB PNG

>>108640216

Another Dunning-Kruger case.

>>108640493
If everyone here is so dumb, why are you asking for tech support here?

>>108640714

Most people are dumb but sometimes you do get something useful out of it.

>>108640768
Read the sticky, go to >>>/wsr/ for your tech support threads

>>108640788

It's more of a "linux is bad and people are trying to deny it" issue.

>>108640882
Sounds like you don't need help then. You want people to go "Bill Gates-sama... I kneel"
But I was curious if this is an issue, since I only have 14.8 GiB of RAM and figured it was because the used laptop had 2 shitty 8GiB sticks or somethiing.
What distro are you using?

>>108641030

The issue isn't your ram stick. It's a general bad problem with linux.

I use Arch Linux (syslinux, lightdm, XFCE4).

>>108640714
Why do YOU come here? Real question.

>>108641030

We had a person earlier who could only utilize 30.45 GiB out of the 32 GiB they had paid for which is worse than our cases but even in your case it's still pretty bad since you really need all ram you paid for if you only have 16 GiB physical ram.

The issue is almost certainly the linux kernel itself rather than anything else. It might be that this cannot be fixed without rewriting parts of the kernel itself (working for free so valve can benefit from it).

>>108639929
memtest is in fact a linux colonel

>>108637757
If you don't install proper drivers the system may reserve RAM for the system to deal with those components in a basic level. You probably got a bunch of missing drivers and that's what you got. Same shit happens on windows.

>>108641069
Technology discussion not tech support, read the goddamned sticky

>>108640216
least retarded /g/ poster

>>108641146

I have tried various nvidia drivers.

Doesn't affect how much ram i can utilize.

>>108637757
you've got to pay the troll toll

linux is trash
there's no excuse for this

>>108641449

It's sadly to be expected with these non-profit open source projects.

Because microsoft is a private centrally controlled company with a lot of resources they are able to put effort and make it work fine even if you use say 63 GiB out of 64 GiB by simply having system processes move to the page file.

Linux instead is fragmented where people are expected to work for free and there is no real coordination and people are convinved doing things in the old inefficient way is necessary and people who try to claim otherwise are shut down. You get the typical stupidity of crowds where instead of fixing real problems defenders of the dysfunctional status quo resort to misinformation and gaslighting.

>>108641147
This isn't tech support though. It's unironically the start of a discussion.

>>108641691
It's tech support, Linux ate my RAM, grade school shit

>>108641724
You've never anything worth reading. I don't see why you're so upset.

>>108637930
66935632K (KiB) is available to Linux, or 68,542,087,168 bytes. Like all OSes that run on x86-64, the default page size is 4096 bytes, which is a property of the hardware. So, you have 16,733,908 pages. Linux needs a 'struct page' and PTEs (page table entries) as well as some other metadata that comes out to about 92 bytes per page or 1,539,519,536 bytes, or about 1,503,437 KiB, which is about 1% less than the "1517888K" value in your dmesg output. In general, you should expect about 2% of your memory to be used in keeping track of your other memory, so this is about right. Any OS that needs to use the MMU to map pages needs this bookkeeping, it's just that Linux doesn't lie to you about it.

>>108641861
>implying that the entries aren't sparse to combat this exact issue
>implying that the number of entries has anything to do with the physical capacity, not the size of the virtual address space
Try more like 0.2%

There are a lot of bad attempts to explain it but it is likely iGPU reservation.
Even when you have a dGPU as well, it's possible to switch between integrated and dedicated while the system is running or run both at the same time, so unless you have actually disabled the iGPU in BIOS it may still reserve RAM to use for it.
ZRAM is also worth checking.

>>108637757
Yeah, well I slapped in 128gb into my rig running mint and it says I have 134gb, so ha, better luck next time, bucko.

>>108641925
You have to reserve the space so you can run maxxed out. There is no such thing as sparse when under full memory pressure.

>>108642069
I'm talking about the table entries, you absolute tard. The majority of virtual memory will be EMPTY for every fucking process.

>>108642090
The reason the memory is reserved is so you can allocate them immediately upon a memory utilization spike. Hence, reserved for kernel use and not available for direct userspace use.

>>108642134
You don't even know what I'm talking about. You don't even realize you don't know what I'm talking about.

>>108642149
The memory overhead for bookkeeping when running maxxed out is accounted for up front at boot time. Hence, reserved.

>>108642158
Gunmaxx your beta hole

File: IMG_0622.jpg (231.3 KB)

231.3 KB JPG

>>108642178
Ask google AI
>linux kernel struct page overhead per 4096 byte page
>pte overhead per 4096 byte page linux
>what besides struct page and pte overhead uses reserved memory linux

The rule of thumb is 2% when running maxxed out. You don't have to take my word for it, if you want see picrel gAyI

>>108642200
You've already established you're retarded. You don't have to keep going

>>108641645
let me guess, you disabled swap to improve performance?

>>108642209
Lots of seething trannies today, I wonder why.

>>108642209
>>108642249
Ask stupid questions get answer you don't like...call them transgender?

>>108637757
post the output of

sudo dmesg

paste.debian.net
you won't though because for one thing you know what the problem is and are expecting a random internet user to use the process of divination to tell you what command to run to fix this problem instead of even trying to be facetious about it.

>>108642249
Cope and rope, man(page)let.

Looks like OP has some experience coming up short where it matters most ;^)

>>108642283
ywnbaw

>>108642304
Typical tranny imitating roastie cope

>>108642358
I'm a man and have no desire to be a woman. You must be projecting your fetishes onto me.

>>108642233

I disabled SWAP to free up SSD space and improve the lifespan of the SSD.

>>108641955

13900KF don't have a iGPU (pretty sure it's fused off so you cannot use it at all).

>>108642287

Memory: 65373456K/66935632K available (20562K kernel code, 2951K rwdata, 16740K rodata, 4776K init, 4600K bss, 1517888K reserved, 0K cma-reserved)

Don't know if it would be safe to post the full output and this is also not a tech support thread.

From what i have read the amount of memory reserved by the kernel is normal given the amount of ram i have.

>>108642200

Seems like the Struct Page is indeed the biggest problem and that windows doesn't have the same issue.

The amount of gaslighting and misinformation i have gotten when i have asked about this problem indicates to me that this is a linux specific problem (windows does let me utilize up to 63.8 GiB after all).

Seems like you can reduce it in linux but this is badly documented and may cause problems.

https://blogs.oracle.com/linux/minimizing-struct-page-overhead

>>108637902
I usually have only 3, lucky me

>>108643538
Nah, Windows just considers the cost of their "struct page" to be part of the user's memory usage.

Has anyone actually managed to successfully reduce the amount of ram reserved by the kernel to less than 1% of your total ram?

File: linux-inefficient.png (122.7 KB)

122.7 KB PNG

>>108643658

According to grok the windows equivalent takes wastes 25% less memory by default.

https://youtube.com/watch?v=2EFG3BO6oVY

>>108637757
Doesn't happen on my 4GiB Debian+XFCE install, I get 4GiB on both Windows and Linux. Skill issue.

>>108640216
He does indeed have 64 GiB of ram retard. This is RAM, not a mechanical hard drive.

>>108640714
He is making a tech discussion thread, which is what the board is for, not a tech support thread.

>>108643727
when i boot into windows it eats up 4gb of ram idling, even with no applications besides task manager running

File: Screenshot From 2026-04-20 11-52-26.png (8.6 KB)

8.6 KB PNG

>>108637757
I seem to have the opposite problem...

>>108645754
Flamewars != Discussion
This website person hates Linux, cool, not a discussion.

>>108643727
>according to LLM slop
both still need to be padded/aligned to 64b

>>108643389
>1517888K reserved
ya. sounds like iGPU

>>108645807

67.1 GB = 62.49 GiB

>>108645754
This is not a tech discussion thread.
This is a low effort shitposting bait thread.

>>108645754
I meant 3.9 GiB on both, brainfart. But yes, still a giant skill issue. Inspect /proc/iomem and find out why. Maybe post it here.

>>108645810
He is discussing an actual topic on how linux apparently reserves more memory for kernelspace on his puter, which is a very valid discussion. Plus, he's definitely slightly smarter than the average /g/ poster, which is a welcome rarity given that defying all statistics, the average /g/ user manages to be significantly dumber than the average /g/ user somehow.

File: Screenshot From 2026-04-20 12-42-47.png (9 KB)

9 KB PNG

>>108645888
you're right, I'm retarded.

>>108637757
GiB is not the same as GB

>>108646181
nowhere in OP was GiB confused with GB

>>108645838

Wouldn't that show up under "cma-reserved" ?

It would be pretty bad if linux reserved 1 GiB + for a non-existent iGPU.

>>108646002

It's hard to find any good information on how to actually reduce the amount of ram reserved by the kernel.

The biggest culprit seems to be the "Struct Page" which does grow linearly with the amount of ram you have (which is pretty bad in my case since i have 64 GiB of DDR5).

>>108646281
Do you want the kernel to work? It needs RAM. THE MORE YOU HAVE THE MORE IT NEEDS.

Bitches in 2038 will be like "I LOST 16GB OF RAM" nigga you have 1TB of RAM stop sweating the small stuff. Yes, 16GB is a lot right now, but in a future context where people have terabytes of RAM it will be insignificant.

>>108646254
idk man. I'm trying to figure it out as well. but it may be a quirk with AMI BIOS, because I can't figure it out either. the iGPU doesn't show up either.

>>108646313

I might try freeBSD later (if i get some more available disk-space) to see if it lets me use more ram.

You could also try switching linux distro but i don't think it will help much.

>>108646308
The funny part is that if Linux did userspace memory accounting in units of (page_size + overhead, e.g., 4096 + 92 == 4188) there'd be no "problem" because the moment a process used 1 byte of the page, they'd get charged for the full Hitler. But since Linux devs are more interested in "how many regular hw pages can I mmap and actually use" it makes more sense to account like Linux does.

The problem is rolling up developer info to userfaggot GUI info because users are no longer predominantly actual devs.

>>108637757
>Has 64GB of ram. Only uses half of it. Doesn't want his OS to reserve 1GB ram in case it needs it
You really think your games or whatever you are running is more important than the system that allows you to run said programs?

>>108646798

That would just hide the inefficiency.

According to unreliable sources the overhead from that was smaller in wndows.

Perhaps i could find a way to do a proper test.

>>108647105
You are literally retarded. This is why generations of people are blackpilled on teaching anyone anything; faggots like you are best off completely uninformed so at least you're not mostly uninformed, like you are now. But go get 'em faggot.

File: image.png (67.8 KB)

67.8 KB PNG

>>108639976
Is this normal? I have 64GB RAM and a 4TB SSD.

>>108647105
Linux and the NT kernel are both doing similar things by reserving ram for kernel needs. This is separate and different from hardware ram reservation. The only difference is that Linux is transparent in how it reports this information to you. You should be more upset at Microsoft for having misled you and wasting the past 24H of your time

>>108647156

Your Radix_tree_node is smaller than mine for some reason.