Inserter Throughput

[JasonC]

I want to add inserter throughputs (for various belt orientations if its dependent) to the wiki. There doesn't seem to be a post dedicated to this 100% so let this one be the reference.

The numbers (for transferring from a chest to the side of a straight belt), as of 0.12.29, are (±0.004, I think):

• Coal (not including self loading coal): 0.588 i/s
• Basic: 0.831 i/s
• Long: 1.153 i/s
• Fast: 2.306 i/s
• Smart: 2.306 i/s

This information is from the following post (I do not know the original source of these numbers):

... I'll just add the exact inserter ratios:

Inserts per second (as of 0.12.30)
Fast: 2.31 (26 ticks/insert)
Long: 1.15 (52 ticks/insert)
Basic: 0.83 (72 ticks/insert)

I've also experimentally verified the above numbers as well as tested the number for a coal inserter (pre-loaded with coal, not self-fueling) with the following vaguely pictured setup:



Where the throughputs are displayed on the right, measured by dividing the number of items removed from a chest by time passed:



---

My original post, to which Koub was responding below, was:

I want to add inserter throughputs (for various belt orientations if its dependent) to the wiki.

Before I set up an experiment, has anybody already determined this? I can't find it in the forum but it seem like something somebody would have done.

[Koub]

viewtopic.php?p=146115#p146115
This should help

[JasonC]

viewtopic.php?p=146115#p146115
This should help

Thanks. I double checked those, and tested coal too, and updated the first post here as well as added https://wiki.factorio.com/index.php?tit ... throughput.

[BlakeMW]

Thankyou for verifying my numbers, the setup I use includes the LED display blueprint, but because I'm not a wizard with combinators I use a simpler method, I make a combinator clock set to count from 0 to some really high number. I then link the smart chests to the LED displays and can compare the items inserted with the time in ticks, which is also shown on an LED display. A snapshot is taken with the OS screenshot utility (I couldn't figure out a good in-game way to freeze the setup on a certain tick without doing more with combinators than I'm comfortable with...)

Here are a few other interesting data points, these are mainly things I wanted to confirm still occur:

The first is: Fast Inserters which are orientated North are slower


So turns out that East/West Inserters insert 3.95% faster than North/South Inserters when dealing with saturated belts.
(Note: I made sure all inserters were picking from the end of the belt)

The second is Insertion speed depends on Belt orientation


To my surprise, it turns out perpendicular orientation (near side insertion) is nearly 8.7% faster compared with parallel orientation.

Final one Nearside Inserting is faster than Farside Inserting:


Well hey those numbers look familiar... so it turns out that near side inserting is faster by 8.7%, and apparently picking from the end of the belt counts as far side inserting. There is no exhibited difference in insertion speed by orientation with forced side inserting, the south inserter is just as fast as the west inserter (or just as slow in the case of farside inserting).

There are many more possible permutations which could be tested especially chest insertion but the above covers some of the more common/important cases.

[orzelek]

Nice test setup

Would you be willing to test near/90 degree inserters from mods? (Side inserters and Near inserters mods)
It seems for me that 90 degree inserters can do their job significantly faster - much shorter arm movement.

[JasonC]

Here are a few other interesting data points, these are mainly things I wanted to confirm still occur:
...
The first is: Fast Inserters which are orientated North are slower

Using the same experimental setup as above:

• Vertical belt, east/west of inserter:
  • See original numbers.
• Horizontal belt, south of inserter:
  • Same as east/west numbers.
• Horizontal belt, north of inserter:
  • Burner: 0.579 i/s (1.53% slower)
  • Basic: 0.820 i/s (1.32% slower)
  • Long: 1.129 i/s (2.08% slower)
  • Smart/Fast: 2.220 i/s (3.73% slower)

It's not just fast inserters, burner, basic, and long inserters show the same behavior.

Weird. Wiki updated.

[BlakeMW]

Something else relevant to this topic:

I weaponized the difference in insertion speed to make a pure fast inserter low-power condition:



The out inserter is on dedicated power, via sideloading the fish swap sides and the in-inserter is on the main grid and will tend to insert about 8% faster because it's picking from the near side. That means when the main grid is fully powered up (or no less than 8% underpowered) the chest will be gradually filled. Because it is so finely balanced it makes a very sensitive blackout/brownout detector. Also after a prolonged blackout it takes a long time for the in-inserter to refill the chest so the low power condition will remain true for some time.

Surprisingly but logically the fully-powered equilibrium state (and also the chest level you use to detect full power) is 3 fish left on the belt, the in-inserter cannot put those 3 fish back because to pick up the 3rd fish from the end it has to swing a full 180 degrees and becomes just as slow as the out inserter, as such it constantly picks up from the 4th slot on the belt where it is in equilibrium with the out inserter - it has a speed advantage when picking from what we might call the "4th and a half" slot (that is grabbing an item which is entering it's reach) which is why it is able to clear the backlog.

This is instructive, near-inserting is faster only when belts are fully saturated. It is faster because the angle turned by the inserter is less.

This fact can also be weaponized, while terminal inserting is generally slower, terminal-near-inserting can be even faster:



In this case the in-inserter turns through an extremely short angle and extends a short distance and can easily clear the belt. This version has an empty belt fully powered state and is still very sensitive. Why doesn't it work if the items are on the far lane? Because the inserter isn't smart enough to grab the closest item, instead it goes for the furthest away item on the far side of the belt and the tile and also shows a strong tendency to "windmill" (turn 360 degrees).

[JasonC]

I weaponized the difference in insertion speed to make a pure fast inserter low-power condition:

Well, the difference is so subtle that you may also have made a Factorio-was-just-updated detector.

[AntiElitz]

I also did those experiments some time ago and created a wiki page about it. However i never found the time to overhaul the Inserters wiki page, so i never linked to it at all.
The Numbers on the smart Inserters are outdated since they behave the same as blue ones due to a recent patch. However the others are still valid.
https://wiki.factorio.com/index.php?tit ... xperiments

[ignatio]

I've put a bit of effort into measuring inserter throughput in the 0.13+ world where we have varying stack sizes. I've documented my findings on the wiki page. There's also a spreadsheet with all the numbers and plenty of graphs, although I haven't put a whole lot of effort into making it comprehensible.

I found that even the slowest inserters can pick items from express belts nowadays, but they are not particularly fast at it.

One thing that was a bit of a surprise to me is the belts at which the different inserters are most efficient at different stack sizes. There doesn't seem to be a clear pattern to it. E.g. the long-handed inserter is quickest with basic belts even at stack size 3, where both burner and basic inserters get faster with fast belts.