process.nextTick vs setTimeout(fn, 0)

nextTick.js

for (var i = 0; i < 1024 * 1024; i++) {
  process.nextTick(function () { Math.sqrt(i) } )
}

results.md

Results

Intel i7 890 @ 2.93 GHz x64, node compiled with -march=native -mtune=native:

$ time node nextTick.js 

real	0m0.344s
user	0m0.276s
sys 	0m0.067s

$ time node setTimeout.js 

real	0m9.125s
user	0m8.707s
sys 	0m0.410s

Feel free to fork and add your results!

setTimeout.js

for (var i = 0; i < 1024 * 1024; i++) {
  setTimeout(function () { Math.sqrt(i) }, 0)
}

"setTimeout" usually has built-in minimal delay value (which is a few ms). Even if you call "setTimeout" with 0 ms delay it will still run with at least with few ms delay. This is an another reason why it is so slow in these "benchmarks".

See: https://developer.mozilla.org/en/docs/Web/API/window.setTimeout#Notes

400 MHz ARM926EJ-STM ARM® Thumb® Processor
32 Kbytes Data Cache, 32 Kbytes Instruction Cache, MMU

[root@alarm ~]# uname -a
Linux alarm 3.1.0-g6853fe7 #6 PREEMPT Tue Mar 4 21:53:22 CST 2014 armv5tejl GNU/Linux
[root@alarm ~]# node -v
v0.10.26
[root@alarm ~]# time node nextTick.js 

real    0m47.661s
user    0m46.309s
sys 0m1.352s
[root@alarm ~]# time node setTimeout.js 

real    2m10.043s
user    2m4.707s
sys 0m5.332s

Is expected this result on arm platform?

RAM 3.7 GiB
Processor Intel® Core™ i5-2467M CPU @ 1.60GHz × 4
Arch 64-bit
OS Linux, Ubuntu 13.10

$ time node nextTick.js
 real   0m0.749s
 user   0m0.630s
 sys    0m0.089s

$ time node setTimeout.js
 real   0m1.294s
 user   0m1.223s
 sys    0m0.080s

RAM: 4 GiB
Proc: [email protected]
OS: Win 7x64

$ time node benchmark/nextTick.js
real    0m0.843s +/- 0.078
user    0m0.015s
sys     0m0.016s

$ time node benchmark/setTimeout.js
real    0m1.468s +/- 0.048
user    0m0.000s
sys     0m0.015s

Ubuntu-14.04
node -v: v0.10.35

$ time node nextTick.js
0.42s user 0.04s system 99% cpu 0.465 total
$ time node setTimeout.js
1.08s user 0.23s system 100% cpu 1.306 total

😄 👍

$ time node nextTick.js && time node setTimeout.js 

real    0m1.044s
user    0m0.933s
sys     0m0.120s

real    0m1.499s
user    0m1.420s
sys     0m0.117s

Meh.

Can someone explain why there is a difference? In fact, I am facing an issue where setTimeout(f, 0) and process.nextTick(f) behave entirely different. Within a loop, the first will work as expected, whilst the second appears to not work.

Ubuntu 15.10 (VMWare player)
6 cpu | E5-1650 @ 3.2
Node 5.0.0

Average results after 3 runs

(With other node processes opened)

$ time node nextTick.js 
real    0m0.924s
user    0m0.584s
sys     0m0.348s

$ time node setTimeout.js 
real    0m0.732s
user    0m0.608s
sys     0m0.124s

(No ther node processes opened)

$ time node nextTick.js 
real    0m0.800s
user    0m0.736s
sys     0m0.072s

$ time node setTimeout.js 
real    0m0.714s
user    0m0.644s
sys     0m0.076s

I think you must compare setTimeout(fn, 0) to setImmediate(fn), but I think setImmediate will be still a bit faster.

$ uname -a
Darwin Wind.local 15.4.0 Darwin Kernel Version 15.4.0: Fri Feb 26 22:08:05 PST 2016; root:xnu-3248.40.184~3/RELEASE_X86_64 x86_64

$ node -v
v5.0.0

$ time node nextTick.js 
real    0m0.761s
user    0m0.691s
sys     0m0.075s

$ time node setTimeout.js 
real    0m0.728s
user    0m0.657s
sys     0m0.074s

$ time node setImmediate.js 
real    0m0.682s
user    0m0.621s
sys     0m0.064s

@ismoura No much difference right?

$ uname -a
 Linux  4.7.0-040700-generic #201608021801 SMP Tue Aug 2 22:03:09 UTC 2016 x86_64 x86_64 x86_64 GNU/Linux

 $ lsb_release -a
 Distributor ID:    Ubuntu
 Description:          Ubuntu 16.04.1 LTS
 Release:          16.04
 Codename:         xenial

  $ node -v
  v6.4.0

  $ time node nextTick.js 

  real      0m2.449s
  user     0m2.252s
  sys      0m0.124s

 $ time node setTimeout.js 

 real        0m2.693s
 user       0m1.604s
 sys         0m0.156s

Using let

$ time node nextTick.js 

real    0m2.875s
user    0m2.768s
sys 0m0.156s

$ time node setTimeout.js 

real    0m2.305s
user    0m2.128s
sys 0m0.108s

Using var

$ time node nextTick.js 

real    0m2.300s
user    0m2.256s
sys 0m0.088s

$ time node setTimeout.js 

real    0m1.699s
user    0m1.588s
sys 0m0.108s

RAM 8G
core 2

time node nexttick.js
real	0m0.863s
user	0m0.795s
sys	0m0.089s

time node timeoutqps.js
real	0m1.084s
user	0m0.978s
sys	0m0.111s

Darwin Kernel Version 16.3.0: Thu Nov 17 20:23:58 PST 2016; root:xnu-3789.31.2~1/RELEASE_X86_64 x86_64

node v7.2.0

node nextTick.js 0.62s user 0.12s system 99% cpu 0.747 total

node setTimeout.js 0.70s user 0.11s system 98% cpu 0.821 total

@IngwiePhoenix The explanation is here: https://nodejs.org/en/docs/guides/event-loop-timers-and-nexttick/#process-nexttick-vs-setimmediate

process.nextTick() allows you to "starve" your I/O events. setImmediate is safer.

$ uname -a
Darwin MacBook-Pro.local 18.7.0 Darwin Kernel Version 18.7.0: Tue Aug 20 16:57:14 PDT 2019; root:xnu-4903.271.2~2/RELEASE_X86_64 x86_64

$ node -v
v12.11.1

$ time node nextTick.js 
node nextTick.js  0.52s user 0.12s system 189% cpu 0.338 total

$ time node setTimeout.js 
node setTimeout.js  1.09s user 0.12s system 128% cpu 0.938 total

on ubuntu virtual machine

time node nextTick.js

real	0m0.669s
user	0m0.487s
sys	0m0.357s


time node setTimeout.js

real	0m1.513s
user	0m1.709s
sys	0m0.090s