A slow Linux system may be short of CPU time, or it may be waiting on something else. The top monitor shows CPU usage alongside running processes, helping distinguish a busy application from a machine with spare processor capacity.

The procps version of top combines a system summary with a process table. The summary averages activity across logical CPUs, while the process table normally measures each task against one logical CPU. A single busy thread can therefore approach 100% CPU while the machine still has idle capacity, and a multithreaded process can exceed 100% CPU.

A brief spike does not establish sustained CPU pressure; several refreshes show whether the activity persists. Inside a container, the CPU summary and load averages may describe the host rather than the container's CPU allowance, while the task table only lists processes visible inside that container.

Related: How to check load average in Linux
Related: How to check process CPU and memory usage in Linux

Steps to show CPU usage in Linux with top:

  1. Start top in the terminal. 
    $ top

    No root privileges are required. The CPU summary appears below the load average and task counts.

  2. Read the %Cpu(s) line after the next screen refresh.

    us is time running ordinary user processes, sy is kernel time, and ni is time running processes with an adjusted nice value. A high id value means idle capacity; sustained low id with high us or sy indicates CPU activity. wa is I/O wait, rather than active CPU work. hi and si are hardware and software interrupt time, and st is virtual CPU time taken by the hypervisor.

  3. Press uppercase P to sort processes by CPU usage.

    The %CPU column shows activity since the previous refresh; PID and COMMAND identify the process. Here, yes uses 99.0% CPU while the combined summary remains 87.5% idle. Its high usage occupies roughly one logical CPU, rather than the whole machine. The TIME+ column is accumulated CPU time, not a current percentage.

  4. Press 1 to display usage for each logical CPU.

    The combined line becomes %Cpu0, %Cpu1, and further CPU rows. Here, Cpu1 has only 1.0% idle while most other CPUs are idle. A busy thread can move between CPUs, so the busiest row may change between refreshes. Pressing 1 again restores the combined summary.

  5. Press q to return to the shell.
  6. Capture two CPU readings one second apart for a copyable snapshot. 
    $ top -b -d 1 -n 2 -o %CPU
##### snipped #####
top - 01:10:07 up 10 days, 21:44,  0 user,  load average: 0.66, 0.25, 0.24
Tasks:   4 total,   2 running,   2 sleeping,   0 stopped,   0 zombie
%Cpu(s):  3.4 us,  9.4 sy,  0.0 ni, 86.9 id,  0.0 wa,  0.0 hi,  0.2 si,  0.0 st 
MiB Mem :  11946.3 total,   2621.1 free,   1116.7 used,   8470.2 buff/cache     
MiB Swap:   4096.0 total,   4016.6 free,     79.4 used.  10829.6 avail Mem 

    PID USER      PR  NI    VIRT    RES    SHR S  %CPU  %MEM     TIME+ COMMAND
   1629 user      20   0    2708   1552   1452 R  98.0   0.0   0:57.14 yes
      1 root      20   0    2708   1620   1528 S   0.0   0.0   0:00.01 sleep
   1621 root      20   0   32800  20980   7528 S   0.0   0.2   0:00.08 Xvfb
   1756 user      20   0    9032   5024   3016 R   0.0   0.0   0:00.00 top

    -b produces plain text, -d 1 sets the interval, -n 2 stops after two displays, and -o %CPU sorts by CPU usage. The first display is omitted here; the second measures the intervening second. It shows yes at 98.0% CPU with 86.9% idle overall, confirming a busy process alongside spare system capacity. On a CPU-bound machine, low idle time persists across readings. Actual values and process names vary.

Share this guide