From Gentoo Wiki
Jump to:navigation Jump to:search

This article describes various methods of system benchmarking on Linux. It covers CPUs, power usage, RAM, and graphics.

Power usage

sys-power/powertop can guess the computer's current power usage in watts.

root #powertop -d | grep usage
Power usage (ACPI estimate): 15.7W (0.5 hours)


Some examples with different cryptographic commands. OpenSSL tries many different combinations of ciphers, key lengths and modes:

user $openssl speed

This could be bound to a single core (as root):

root #taskset -c 1 openssl speed

sys-fs/cryptsetup instead has less algorithms but includes kernel's cryptographic API if available:

user $cryptsetup benchmark


Command time measures execution time of a sub command. Use a reproducible command to measure CPU performance:

root #time emerge -1 coreutils
real	2m33.483s
user	2m6.310s
sys	0m55.257s

Using emerge is not very reliable since the task is very complex and also requires I/O time and does not run parallel in all stages. Instead with sys-devel/bc calculation of PI burns CPU time for a single core only:

user $time echo "scale=5000; 4*a(1)" | bc -l



app-benchmarks/ramspeed can measure the computer's ram speed.

ramspeed is incompatible with ACCEPT_LICENSE="@FREE". It is a proprietary program.
user $ramspeed -b1 -m4 | grep 4096 && ramspeed -b2 -m4 | grep 4096
INTEGER & WRITING      4096 Kb block: 1948.94 MB/s
INTEGER & READING      4096 Kb block: 5208.66 MB/s


Some of the available packages to measure input-output-performance in general:


Historically used to measure hard disk performance:

root #hdparm -tT /dev/sda
 Timing cached reads:   1076 MB in  2.00 seconds = 538.08 MB/sec
 Timing buffered disk reads: 242 MB in  3.01 seconds =  80.32 MB/sec


Using time to examine time spent in user and system space:

user $time bzip2 -c patch-3.3 > patch-3.3.bz2
real	0m13.272s
user	0m7.606s
sys	0m0.054s
user $time eix search

real	0m0.363s
user	0m0.066s
sys	0m0.008s




root #emerge --ask games-fps/nexuiz

Reset default options

If you have played nexuiz you should reset to default options for canonical benchmark results.

user $rm -r ~/.nexuiz/


user $nexuiz-glx -benchmark demos/demo1 -nosound 2>&1 | egrep -e '[0-9]+ frames'
1910 frames 42.0330749 seconds 45.4404063 fps, one-second fps min/avg/max: 35 46 63 (90 seconds)



root #emerge --ask games-fps/xonotic


If you have played Xonotic you should reset to default options for canonical benchmark results.

user $rm -r ~/.xonotic/


user $xonotic-glx -benchmark demos/the-big-keybench 2>&1 | egrep -e '[0-9]+ frames'
12568 frames 556.0637400 seconds 22.6017255 fps, one-second fps min/avg/max: 19 23 27 (207 seconds)


glxgears is a very basic OpenGL support test, it is not a real benchmark tool!

Obtain glxgears by emerging x11-apps/mesa-progs.

user $vblank_mode=0 glxgears & sleep 15 ; killall glxgears
5762 frames in 5.0 seconds = 1152.369 FPS
5806 frames in 5.0 seconds = 1161.017 FPS

See also

  • PowerTOP — a Linux utility that can monitor and display a system's electrical power usage.

External resources