I've been using Xterm and enjoying it as it seems pretty efficient, and it comes with X typically. I've heard many people call it the slowest terminal emulator, which didn't match my experiences. The often recommended choice is urxvt, a unicode-compatible fork of rxvt. My only experience with urxvt has been watching a friend's program run very slow in it, while the same program was much faster in xterm, so I was not impressed.
For whatever reason, I grew curious and looked up what people had measured. It is possible that the friend's graphics drivers were not installed properly (ever important to drawing anything in X efficiently) or something else was screwy, as urxvt consistently ranks well while xterm ranks pretty poorly. So, I decided to measure a few myself.
I've immediately eliminated konsole and gnome-terminal for two reasons. For one, the output is not updated very quickly on these, so quickly scrolling or watching something change rapidly becomes a choppy experience, which I do not favor. The second reason is that I don't even want to install them - Konsole wanted about 200MB of KDE dependencies. Gnome-terminal was not so bad in this way, but it was not stellar either.
I elected to compare xterm, my old favorite, to rxvt-unicode (also known as urxvt) and xfce4-terminal. I had high hopes for XFCE4's terminal, as it is a desktop environment intended to be light-weight.
All terminals were configured with the same scrollback buffer size (16384 lines) and the bitmap font "Gohufont" at size 8. All ran an equivelant python one-liner, which prints out a line of text 1,000,000 times. The line was padded with spaces so it is the exact same length in each test.
Baseline standard: /dev/null
First place: urxvt
Second place: xterm
Third place: xfce4-terminal
The results surprised me: urxvt is the clear winner here. Xterm wasn't particularly fast, but it is certainly not the sluggish mess that some people make it out to be.
XFCE4's terminal's performance was embarassing.
I suppose urxvt is left as the clear choice.
A few things I did not post that I checked were whether the contents of the line (all spaces versus all pound signs) mattered (it did not), and whether the size of the window affected the results (it did not).
Back to main index