Fix High Memory Usage on Linux: 8 Steps

High memory usage on Linux is often not a problem at all—it’s the kernel using free RAM for disk caching—but when it is a real issue, the fix usually involves identifying the specific process or service consuming memory and then either restarting it, adjusting its limits, or adding swap. This article explains how to diagnose high memory usage, distinguish between harmless caching and actual pressure, and apply practical fixes that work on most current distributions.

Understanding Linux Memory Management

Linux treats unused RAM as wasted RAM. The kernel aggressively uses free memory for page caches, buffer caches, and other temporary storage to speed up file access. This is why free -h often shows very little “free” memory even when the system is healthy. The key metric is not “free” but “available,” which accounts for memory that can be reclaimed instantly when an application needs it. If your system shows high usage but the “available” column is large, you are probably fine.

Real memory pressure occurs when the sum of all running processes’ working sets exceeds the available physical memory plus swap. At that point, the kernel starts swapping, which can cause system-wide slowdowns. Understanding this distinction is the first step in any fix.

Diagnosing What’s Using Your RAM

Before changing anything, you need to know which processes are consuming the most memory. The following commands give you a clear picture:

free -h – shows total, used, free, shared, buffer/cache, and available memory.
top or htop – interactive process viewers; press M in top to sort by memory usage.
ps aux --sort=-%mem | head – a static snapshot of the top memory consumers.
smem – a tool that reports proportional set size (PSS), which accounts for shared libraries more accurately than RSS.

Look for processes that consistently occupy hundreds of megabytes or gigabytes. Common culprits are web browsers (especially with many tabs), Electron-based apps, Java applications, databases, and memory-hungry desktop environments like GNOME.

When High Memory Usage Is a Problem

High memory usage becomes a problem when it leads to excessive swapping, out-of-memory (OOM) kills, or sluggish response. Symptoms include:

System monitor shows swap usage climbing.
Applications take a long time to open or become unresponsive.
The kernel logs OOM killer events in dmesg or journalctl -k.

If you see these signs, the issue is not the caching—it is that your workload genuinely needs more memory than you have. At that point, you can either reduce the workload, add more RAM, or optimize how memory is used.

Practical Fixes to Reduce Memory Usage

Here are the most effective steps to lower memory consumption, in order of impact:

Restart the heaviest process. A simple restart of a browser or a long-running service often reclaims memory that has become fragmented or leaked. For example, systemctl restart myservice.
Remove unnecessary startup applications. Check your desktop environment’s autostart folder (~/.config/autostart) and disable anything you don’t need.
Use lighter alternatives. Replace GNOME Software with a lighter package manager frontend, or switch from a full desktop to a window manager if you have very little RAM.
Adjust swappiness. The vm.swappiness parameter (default 60) controls how aggressively the kernel swaps. Setting it lower (e.g., 10) makes the system prefer to keep processes in RAM, which can improve responsiveness on systems with enough memory. You can set it temporarily with sudo sysctl vm.swappiness=10 and make it permanent by adding vm.swappiness=10 to /etc/sysctl.conf.
Limit a process’s memory with systemd. For a service that uses too much, you can add MemoryMax= and MemoryHigh= directives to its unit file. For example, MemoryMax=2G caps it at 2 GiB.
Clear the page cache (temporarily). If you want to free cached memory for benchmarking, run sudo sync; echo 3 > /proc/sys/vm/drop_caches. This is not a permanent fix; the cache will rebuild.

Common Misconceptions and Mistakes

Many users panic when they see “used” memory at 90% and immediately kill processes or disable caching. That is usually counterproductive. The page cache exists to make your system faster; clearing it slows things down until it repopulates. Another mistake is relying solely on top’s RSS column, which overcounts shared libraries. A process that looks like it uses 500 MB may only actually own 100 MB of unique memory. Use smem or check /proc/[pid]/smaps for a more accurate picture.

Also, do not confuse memory usage with a memory leak. A leak is a gradual increase over time that never drops, even when the process is idle. If you suspect a leak, monitor the process for a day or two with ps -o pid,rss,cmd -p PID before taking drastic action.

When to Add Swap or More RAM

If your system consistently runs out of memory even after optimization, you have two main options:

Add swap space. Swap is a disk-based extension of RAM. It is slower but prevents OOM kills. You can create a swap file with fallocate -l 2G /swapfile, set permissions, and enable it with mkswap and swapon. Modern systems often use a swap file by default, so you can simply resize it.
Add physical RAM. This is the only real fix if your workload genuinely needs more memory than your hardware provides. Before buying, check your motherboard’s maximum capacity and the type of RAM it supports.

For most desktop users, 8 GB is comfortable for light work, 16 GB for multitasking and virtual machines, and 32 GB or more for heavy development or media editing. Servers vary widely; monitor your usage over time to determine the right size.

Frequently Asked Questions

Why does Linux use so much RAM when I have nothing open?

Because the kernel caches files and directories in unused memory. This is by design and speeds up disk access. Check the “available” column in free -h; if it is high, you have no problem.

How do I clear cached memory on Linux?

You can drop caches with sudo sync; echo 3 > /proc/sys/vm/drop_caches, but it is not recommended unless you are benchmarking. The cache is automatically reclaimed when applications need memory.

What is the OOM killer and how do I stop it?

The Out-Of-Memory killer is a kernel mechanism that terminates processes when the system runs out of memory. To prevent it, you need to reduce memory usage or add more RAM/swap. You can also adjust OOM scores with oom_score_adj to protect critical processes, but that is advanced.

Is it safe to set swappiness to 0?

Setting swappiness to 0 disables swapping of anonymous pages, which can cause the OOM killer to trigger earlier under memory pressure. A value of 10 is a safer low setting. Only set 0 if you know your workload fits entirely in RAM.

Why does my system use more memory after a kernel update?

Newer kernels may change memory management defaults, or the update may enable additional services. Check what services are running with systemctl list-units --type=service and disable any new ones you do not need.

How do I check which process is using the most memory?

Use top and press M to sort by memory, or run ps aux --sort=-%mem | head -10. For a more accurate picture, install smem and run smem -t -k.

General steps for current mainstream versions; menus can differ by device and version. See the disclaimer or report an error.

More in this bed

The whole Tech How-To bed →