8 ms·
Here's some comparison of input lag: # Title Min Max *Avg* SD 1 iTerm 3.3.0beta16 11.2 33.6 15.9 3.4 2 iTerm 3.2.9
by IfOnlyYouKnew 7y ago
Here's some comparison of input lag:
# Title Min Max *Avg* SD
1 iTerm 3.3.0beta16 11.2 33.6 15.9 3.4
2 iTerm 3.2.9 10.8 29.5 21.3 4.5
3 hyper 3.0.2 22.3 46.9 27.6 3.6
4 Terminal 2.9.1 2.4 8.3 5.8 0.5
5 Alacritty 0.3.3 5.2 16.8 6.3 1.4
6 Kitty 0.14.3 7.7 11.6 9.6 0.8
Measured using https://github.com/pavelfatin/typometer https://github.com/pavelfatin/typometer
Edit: Added iTerm 3.2.9 (2) added alacritty (3) added Kitty
- penagwin 7y agoI'm I reading that right that the input lag has increased on average? And while the minimum might have improved by .4 or is within measurement error?
- Peter-Jan 7y agoI am running a relatively barebones version of st(https://github.com/Peter-JanGootzen/st https://github.com/Peter-JanGootzen/st) and am getting these numbers on a modern machine. # Title Min Max Avg SD 1 st 12.6 38.1 20.7 6.8 I am quite suprised how low your numbers are(eventough I am aware that comparing between machines is probably unreliable), especially Hyper because my experience with it was that it was really slow compared to other terminal emulators.
- ihuman 7y agoIs the average for iTerm 3.2.9 supposed to be 24.5, not 4.5?
- viraptor 7y agoThe text is shifted. Min starts before its column. 4.5 is the sd.
- IfOnlyYouKnew 7y agofixed, sorry
- djhworld 7y agoCould you test Alacritty too? https://github.com/jwilm/alacritty https://github.com/jwilm/alacritty
- IfOnlyYouKnew 7y agodone
- j605 7y agoAfter testing it feels Kate has similar numbers as Terminal mentioned here but most emulators are just as slow. The minimal ones are actually worse off.
- chipotle_coyote 7y agoThis does lend credence to my perception that Terminal is faster. I know everybody else love love loves iTerm2, but I could never get past the feeling of "ugh I am typing in molasses what is going on here." Later versions have been better, but later versions of Terminal have also had more functionality. It's possible I still don't know what I'm missing, but so far the must have features I've tried apparently just weren't ones I must have. (Yes, I know about split panes, but I can run tmux. Yes, I know about iTerm2's tmux integration and I tried it and it was kind of [shrug emoji] for me, I dunno. Terminal has a find-on-page feature and I so far haven't been saddened it doesn't support regex. My shell has autocomplete and iTerm hasn't sold me on its implementation being better. Alfred has a clipboard history and iTerm hasn't sold me on its implementation being better. And so on.)
- Cort3z 7y agoIt might not be for you, but I would recommend reading through this https://www.iterm2.com/documentation-highlights.html https://www.iterm2.com/documentation-highlights.html I personally find the following either cool or really productive; Open Quickly, Undo close, Triggers, Smart Selection, Instant Replay, Regular Expression Search, Save Mark/Jump to Mark
- fiddlerwoaroof 7y agoThe killer features of iterm for me are three: 1. Built in password manager (command-alt-f) for the Times you’re stuck with ssh passwords 2. Output triggers: iterm can run actions in response to certain text on stdout: I use this to insert a mark at each terminal prompt line and when the phrase “NOTICE ME:” is detected. This way, I can use iterm’s mark navigation features to quickly jump to interesting lines 3. Finally, the biggest advantage of iterm is its ability to auto detect soft boundaries like vim vertical splits and restrict selection to one side or the other of the split.
- dorian-graph 7y agoI've spent at least 6+ straight months on iTerm2 3.x, kitty, and Alacritty (my current primary terminal), and briefly tried Terminal. Terminal is waaay faster. The only thing that's stopping me from using it as my primary terminal is the poor colour support and poor split window support (and setting shortcuts to go between them).
- dgrabla 7y agoCan you test Kitty as well?
- mikeash 7y agoMeasuring the time between a simulated keypress and pixels changing in the framebuffer seems kind of inadequate, since it’s missing out on delays in the input and display hardware.
- apetresc 7y agoSurely that would affect all the applications being compared equally, though?
- mikeash 7y agoNot necessarily. For example, if one app synchronizes with screen refresh and another does not, the latter will perform much better on this test but it’s not actually getting the changes to your eyes any faster. It would also just be nice to know the magnitudes. If the hardware adds a constant 100ms (for example, not a real number) then a 10ms advantage on the software side looks less interesting.
- trufas 7y agoThe hardware should be the exact same between all terminals right? It seems accounting for that would only add noise.
- giobox 7y agoIn many cases the display lag can be one of the biggest sources of input lag, often dwarfing the sub 20-30 ms lag times the earlier post measured across iTerm and it’s alternatives.
- Rebelgecko 7y agoWas that with or without the GPU rendering enabled?
- IfOnlyYouKnew 7y agoEnabled in Iterm2, whatever the default configuration is for the others.
- sonthonax 7y agoCould someone explain to me terminal performance and what these numbers practically mean? It's a complete mystery to me. It's so incredibly variable, the fact that I notice variability as a user really perturbs me, it's 2019, and displaying a matrix of a few thousand characters is still a challenge. I live in fullscreen iTerm, Tmux and Vim. I have 256 colours enabled. Vim has lazy redraw enabled. I'm on a 2018 MacBook pro. Scrolling is smooth only when connected to power (probably related to CPU throttling - but it's 2019 and it shouldn't matter). iTerm is significantly faster than Terminal when resizing Tmux panes and Vim windows. I've had similar issues on Linux. The only terminal emulator that has decent performance scrolling and resizing Vim buffers Alacritty. What's going on? Why is there a bottleneck here? Is it rendering the text? Is that still hard to do in 2019? I don't understand the technical challenge of parsing a protocol that would literally work on a real VT100 sans some escape sequences and colours.
- jakear 7y agoI have found iTerm to have very bad performance relative to Terminal.app. I honestly have no clue why its so popular, I cannot stand it.
- dijit 7y agoI agree with you, or, I did, until they switched to the metal renderer. Now iTerm is as fast as terminal.app (provided you use the same settings)
- bradknowles 7y agoSo, in this case, why would I prefer iTerm over Terminal.app? Seriously, I don't understand what the benefits are, if it's only "just as fast as Terminal.app".
- ShamelessC 7y agoIt has quite a few nice features that Terminal doesn't have. But yeah, if you don't need those features then just use Terminal.
- warvair 7y agoPease test ZOC Terminal: https://www.emtec.com/zoc/ https://www.emtec.com/zoc/. I use it on Windows, like it very much, and don't notice any lag. I'm wondering how it compares to the rest of these on a Mac.
- TimTheTinker 7y agoI can see how that would be a great solution on Windows, since there's no decent built-in TTY (prior to recent builds of Windows 10). It's advertised as an "ssh client" which is great there. But I don't know why anyone would choose to use that on a Mac. The UI requires initiating connections to remote systems (or "localhost") through the GUI -- which feels really backward when you can just open a new window/tab on any other terminal app (for local tty sessions) or run ssh from the command line (for remote sessions).
- warvair 7y agoI wasn't even thinking about local term vs. remote until you mentioned it, but you're right. I guess it's because I don't use a Mac very often and when I do I don't usually have a reason to use the local terminal. I find that's true on Windows as well, I don't often need to use the command line. OTOH, I'm SSHed into remote Linux servers 40+ hours a week.
- IfOnlyYouKnew 7y ago# Title Min Max Avg SD 1 Zorc 2.9 33.2 12.5 5.6 Quite fast, but feels a bit like it's 1999 again otherwise.
- warvair 7y agoThanks!
- norswap 7y agoIs that milliseconds? Aren't those values in the non-noticeable range? I tracked a paper recently saying that users don't notice feedback for less than ~96ms (a bit less in some scenarios).
- capnahab 7y agoAre these milliseconds? The visual evoked response latency ( time taken from the eye to electrical activity in the occipital cortex ) is about 88 ms in Women and 92 ms in men. There will be additional time for the visual object ( eg cursor) to be processed and appreciated in higher centres.
- capnahab 7y agoSo we are unable to differentiate speed of anything appearing on screen of less than about 100ms, making this benchmarking of academic interest only. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4572953/ https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4572953/
- ksaj 7y agoIs it an OS version issue? I haven't updated my OSX in ages (mainly because I don't like the direction its going, and updating seems to be guaranteed to fuck up my Logic Audio install like it does every damn time), and iTerm runs amazingly smoothly on it as is. I feel no lag with it and I spend a lot of time on the command line (Lisp programmer on vim ... go figure...), so I'm wondering if these numbers even matter after a certain point, or if later OSX versions are doing something crappier than the usual.