A comprehensive look at inserter power consumption

[D0SBoots]

There's been a lot of good work done on measuring inserter speed/throughput (viewtopic.php?f=18&t=23409), but I haven't seen high-precision measurements of inserter power use. In particular, I've seen people assuming that a basic inserter uses a constant 13kW when it's operating, and multiplying that by the cycle time to get energy use, but if you look at the power graph it's obvious that's not what's going on:

Inserter Consumption.png (36.04 KiB) Viewed 8557 times

So, let's measure this stuff for real!

The setup

Inserter Test.jpg (162.27 KiB) Viewed 8557 times

My setup is pretty simple: A couple of chests, maybe some belts depending on what I'm testing, and the super-useful electric-energy-interface item, which you need to use the console to give yourself. (That's not an accumulator.) For those (like me) who weren't aware of it, it takes all the pain out of dealing with electricity in benchmarking setups.

All the work is being done by a custom script-let, which is watching a chest or assembler as a trigger, and whenever the number of items in the chest changes, it reports how many ticks it's been, and how much energy has been pulled from the nearest electric-energy-interface. To use it, set up the EEI so that it's producing electricity exactly counterbalancing the inserter's drain, taking into account that it's measured in Joules/tick. So for a basic inserter, that would look like:

Code: Select all

/c game.player.selected.power_production = 400/60

And then run the script-let while selecting the target chest:

Code: Select all

/c local finish = function(msg)
  game.print(msg)
  script.on_event(defines.events.on_tick, nil)
end
local dist2 = function(pos1, pos2)
  local dx = pos1.x - pos2.x
  local dy = pos1.y - pos2.y
  return dx * dx + dy * dy
end
local last_tick = 0
local last_count = 0
local last_energy = 0
local chest = game.player.selected
if chest == nil then
  return finish("Nothing selected")
end
local inventory_type = defines.inventory.chest
if chest.type == "assembling-machine" then
  inventory_type = defines.inventory.assembling_machine_input
elseif chest.type ~= "container" then
  return finish("Don't know what to do with selected item " .. chest.name)
end
local interfaces = game.surfaces[1].find_entities_filtered{name = "electric-energy-interface"}
if #interfaces == 0 then
  return finish("Need a god-mode electric-energy-interface on the map")
end
local interface = nil
local best_dist = 999999999
for i, v in ipairs(interfaces) do
  local dist = dist2(v.position, chest.position)
  if (dist < best_dist) then
    best_dist = dist
    interface = v
  end
end
local fun = function(e)
  local inv = chest.get_inventory(inventory_type)
  local count = inv.get_item_count()
  if count ~= last_count then
    local energy = interface.energy
    game.print("Period: " .. e.tick - last_tick .. " ticks, Energy: " .. last_energy - energy .. "J, Items: " .. count)
    last_tick = e.tick
    last_energy = energy
    last_count = count
  end
end
script.on_event(defines.events.on_tick, fun)

The code requires a little tweaking to read the fuel from a burner inserter, but otherwise works as-is.

Chest-to-chest
Sadly this forum doesn't seem to support tables.

Code: Select all

Name                  | Period (ticks) | Frequency (Hz) |  Energy N/S (J) |  Energy E/W (J) |
Burner inserter       |            102 |          0.588 | 140000.09313226 | 140000.03501773 |
Basic inserter        |             72 |          0.833 | 7000.0006519258 | 7000.0005774200 |
Long handed inserter  |             52 |          1.154 | 7000.0046566129 | 7000.0024959445 |
Fast inserter         |             26 |          2.308 | 9799.9962449074 | 9799.9987483025 |
Filter inserter       |             26 |          2.308 | 11199.995708466 | 11199.998569489 |
Stack inserter        |             26 |          2.308 | 27999.996423721 | 27999.989271164 |
Stack filter inserter |             26 |          2.308 | 27999.996423721 | 27999.989271164 |
                                                        |     Energy N->S |     Energy E->W |     Energy S->N |     Energy W->E |
Burner self-fueling   |             94 |          0.638 | 272171.58610003 | 222171.59522700 | 222171.59522700 | 272171.64179312 |

Surprisingly, the energy usage came out to some very round numbers, with a little bit of noise presumably due to animation differences in how far the arm can swing per tick. There was an interesting slight discrepancy between horizontally aligned inserters and vertically aligned ones, which showed up across the board, but the magnitude is so small that no one would ever notice it in a normal game. Also, the speed numbers match what was already on the wiki, as expected.

What's more interesting is the list of things that didn't matter:

• Left-to-right vs. right-to-left and up-to-down vs. down-to-up (except for burners self-inserting fuel, where it matters a lot)
• Changing the stack size (it changes the avg energy per item, but not the energy per cycle)
• Introducing pauses. (Manually put one coal in the chest, let it cycle, then manually add another coal.)
• Destination chest being full
• Inserting to/from things that aren't chests, like assemblers, boilers, etc.

In all the above cases, the total energy used per cycle was exactly the same (once you factor out the drain, which varies based on how long it took).

Chest-to-belt
This section turns out to be very simple: It's exactly the same as chest-to-chest. Belt orientation doesn't matter, belt type doesn't matter, even the stack size doesn't matter - it always takes exactly the same amount of energy, only varying based on the orientation and type of the inserter. Of course, all those things make a big difference for cycle times and throughput, but that's well-covered elsewhere.

Belt-to-chest
At this point it becomes an unholy mess, and hard numbers are only available for certain specific circumstances. I've moved all the raw numbers for these tests into a spreadsheet: https://docs.google.com/spreadsheets/d/ ... KFOPm3C4w/

Even though there's no exact numbers to be gleaned here, you can see the general pattern: Grabbing from the near lane is always more expensive than the chest-to-chest scenario, even for a single item. Grabbing from the far lane is always cheaper than chest-to-chest, even when grabbing multiple items. (This might not hold true with the stack inserter grabbing 12 items, though.) Increasing the stack size increases the cost, but not too significantly.

Of particular note is that it costs about 113% of base energy to pull items from the near lane on a full belt, and 77.2% of base energy to pull from the far lane, when there's no stack bonus.

[Mehve]

Interesting.

When you say that stack size doesn't matter, are you saying that a stack inserter grabbing 2 items consumes the same total energy as it does while grabbing 12? Because I would have expected that the process of groping around around for multiple items due to higher stack size would increase overall energy consumption, even if the rate didn't increase.

Also, just for reference, where on that power graph would the inserter's idle power level be?

[Fuller]

Hi.

First of all ,thx for all the work!I
There might be something wrong with the stack inserters especially in these cases:
-stack inserter bonus (up to 12 items,AFAIK energy consumption is independent of stack size=> energy/item should decrease with stack bonus)
-stack inserters picking up items from different sources (y-,r-,b-belt vs. chest => with fixed stack bonus, the energy/item should decrease with the belt speed* and reach its minimum with picking from a chest)


*if the amount of items per second is a constant.

[D0SBoots]

When you say that stack size doesn't matter, are you saying that a stack inserter grabbing 2 items consumes the same total energy as it does while grabbing 12? Because I would have expected that the process of groping around around for multiple items due to higher stack size would increase overall energy consumption, even if the rate didn't increase.

If you watch a stack inserter going chest-to-chest, it doesn't pause at all when it's picking up 12 items from the chest. It apparently doesn't consume any extra power when doing so, either. Similarly, when dropping items on a belt, it just sits there, and apparently sitting there and dropping items doesn't cost power either. Only movement (either turning, or hand extension, which is hard to see unless you're really zoomed in) costs energy.

When picking up items from a *belt*, it definitely "gropes around," and that definitely costs extra energy.

Also, just for reference, where on that power graph would the inserter's idle power level be?

That inserter was running all out - I'm not sure that its idle level is on there at all. If it is, it would be the spike downwards just before the big spike up. The broad flat, level is where it is turning.

There might be something wrong with the stack inserters especially in these cases:
-stack inserter bonus (up to 12 items,AFAIK energy consumption is independent of stack size=> energy/item should decrease with stack bonus)
-stack inserters picking up items from different sources (y-,r-,b-belt vs. chest => with fixed stack bonus, the energy/item should decrease with the belt speed* and reach its minimum with picking from a chest)

Yes, the energy consumption is independent of the stack size, so the energy/item decreases with stack bonus. I edited the post to make this a little clearer. As to picking up items from different sources, my early results do show that picking up from a chest will probably be cheapest for a stack inserter with max stack bonus, but getting a good, repeatable test setup when picking up from belts is a mess.

[D0SBoots]

I tidied up the main post, pulling the bulk of the raw data into a spreadsheet and leaving just the big conclusions.

Now that I've collected enough numbers, we can take another look at the age-old question of basic inserters vs. burner inserters for supplying boilers. It looks like the burner inserter is 20x the energy/cycle in all configurations (including pulling from belts), so with the assumption that we have no stack bonus and are pulling from the near lane, the basic inserter is 7908.177969J/item, and the burner is 20x that. Except the burner gets a 2x bonus, since electricity from boilers is only 50% efficient. If K=7908.177969J, the break-even point is when 10K=K+X*drain, X = 9*K/drain = 9 * 7908.177969J / 400W = 2.97 minutes, which is not very long at all.

Furthermore, there is a savings in iron and copper if you replace basic inserters with burners, and you can also get rid of the power poles needed to supply the inserters. It turns out that this resource savings is larger than the cost of the additional fractional coal mine needed to cover the extra coal, and the electricity savings is slightly larger than electricity cost of the fractional mine. So in terms of resources/MW, it's (slightly) cheaper to use burner inserters to feed boilers, even for non-backup plants. The full math is at https://docs.google.com/spreadsheets/d/ ... MxG3PXc3k/

[Exasperation]

Now that I've collected enough numbers, we can take another look at the age-old question of basic inserters vs. burner inserters for supplying boilers. It looks like the burner inserter is 20x the energy/cycle in all configurations (including pulling from belts), so with the assumption that we have no stack bonus and are pulling from the near lane, the basic inserter is 7908.177969J/item, and the burner is 20x that. Except the burner gets a 2x bonus, since electricity from boilers is only 50% efficient. If K=7908.177969J, the break-even point is when 10K=K+X*drain, X = 9*K/drain = 9 * 7908.177969J / 400W = 2.97 minutes, which is not very long at all.

Correction, the break-even point for energy is 2.97 boiler-minutes per coal consumed. If your boilers are running at full utilization that means .037 boiler-minutes per coal consumed (each boiler can consume .45 coal/second), so 2.97 means your boilers are running at 1/80th of their capacity. At this point I think you should be less worried about the relative expense of inserters, electric miners, and burner inserters, and more concerned with the cost of having made 80 times as many boilers as you actually need to run your factory.

Edit - fixed a silly math error - thought the results seemed a bit off.

[D0SBoots]

Yes, I suppose that's another way of looking at it. I didn't make it adequately clear, but the scenario I was examining in that first paragraph was for a backup steam plant, which is (by design) idle most of the time. So if it only inserts one coal per boiler every 2.97 minutes, or equivalently (as you say) the boilers are running at an average of 1/80th capacity or less, then burner-fed inserters will consume less energy (coal, in this case) than the basic inserters. It's not a apples-to-apples comparison, because in reality the basic inserters' drain would be coming from solar instead of coal, but it's good enough for a rule of thumb.

However, it's also sort of moot since it turns out that burners are (surprisingly) better even for a regular steam plant when you account for the full circumstances.

[BlakeMW]

Furthermore, there is a savings in iron and copper if you replace basic inserters with burners, and you can also get rid of the power poles needed to supply the inserters. It turns out that this resource savings is larger than the cost of the additional fractional coal mine needed to cover the extra coal, and the electricity savings is slightly larger than electricity cost of the fractional mine. So in terms of resources/MW, it's (slightly) cheaper to use burner inserters to feed boilers, even for non-backup plants. The full math is at https://docs.google.com/spreadsheets/d/ ... MxG3PXc3k/

Hmmm interesting logic can't say I'm sold on it though since if you took that fractional miner and used it to mine say, iron, you could then make more miners. The burners are simply burning the output of that miner.

[D0SBoots]

Hmmm interesting logic can't say I'm sold on it though since if you took that fractional miner and used it to mine say, iron, you could then make more miners. The burners are simply burning the output of that miner.

Yes and no. It's true that you can use a miner to mine iron to make more miners to mine more iron etc.; Factorio is theoretically exponential in that way. But I don't see how that's relevant to this analysis. If you look at the spreadsheet, there are two scenarios I'm comparing: A tileable boiler layout with 20 boilers, 20 basic inerters, 40 steam engines, 1 offshore pump, and 17.25 coal miners, and virtually the same layout with 20 burner inserters and 17.591 coal miners. (OK, the spreadsheet doesn't actually show the layout itself, but the boiler part is tileable.) They both produce the same gross power of 36MW, but the one with burners produces slightly more net power, and also requires fewer resources to build.

You can't just take some fraction of those coal miners away and turn them into iron miners; the plant wouldn't have enough coal anymore to function at peak power, and it would be drastically less efficient. This is equally true for both the basic inserter and burner inserter scenarios; if you're saying "the coal is just being burned" by the burner inserters, you can say the same for the boilers.

[BlakeMW]

But Return on Investment is a fundamentally valid concept. Basic inserters are a larger investment but have a lower ongoing cost. As far as I can tell your analysis only holds true if the active consumption of coal is considered to have no opportunity cost (by that I mean that consuming coal is like consuming solar power - you don't lose anything by not doing it except the cost of the machine to gather it). While I would not put coal at the same value as iron since coal is not in general a "progress limiting resource", it does run out quickly enough to be bothersome. Altough I guess for a spoon run (assuming max richness) you might realistically never need to relocate a miner so all resorurces can be considered unlimited. I think I actually might have used burners quite a bit in the early game of my spoon run - but that was mainly on the grounds of crafting time saved (and also since they don't require copper meaning it could be done from the initial bank of burner miners on iron).

[D0SBoots]

Yes, that's true. Because my analysis was initially in the context of a spoon run, where everything will only ever be placed once (I'm making sure that all mineral patches are rich enough to last 8 hours), I'm treating miners much the same as solar panels. I think this is still mostly valid even outside that context, specifically in the case of coal, which has few uses outside of boilers. The slightly decreased time before you have to relocate to another coal patch won't be noticed by most people; they will have transitioned to solar or nuclear with steam only as a backup by then.

But in practice, the difference is very slight; other considerations such as getting automated basic inserters as part of green research and "I don't want burner inserters taking up space in my inventory" can be enough to tip the balance the other way.

[Exasperation]

The difference is actually even slimmer than your spreadsheet shows. Your spreadsheet calculates drain as if the inserters never actually move anything; at full utilization (for the boilers, not the inserters), the average drain should be something like 46% of your calculated value. You also underestimate the coal consumption of burner inserters by ~2% (the coal spent refueling themselves). Together, they reduce the difference in your net energy production values from ~48.5 kW to ~44 kW. Doesn't really change the evaluation in the big picture; game settings and mods make a much bigger difference to which version suits your needs better. You're either more concerned with dwindling coal reserves (my last game, the most accessible coal after my starting patch was ~2000 tiles away behind some heavy biter opposition *shakes fist at RSO*) or not concerned with coal running out at all (your spoon run). In the former case, basic inserters make sense, in the latter burners make more sense.

[D0SBoots]

The difference is actually even slimmer than your spreadsheet shows. Your spreadsheet calculates drain as if the inserters never actually move anything; at full utilization (for the boilers, not the inserters), the average drain should be something like 46% of your calculated value.

Drain affects all entities in Factorio constantly, regardless of whether they're operating or not. If you re-read my experimental procedure for calculating inserter energy, I'm relying on this.

You also underestimate the coal consumption of burner inserters by ~2% (the coal spent refueling themselves). Together, they reduce the difference in your net energy production values from ~48.5 kW to ~44 kW.

Good point that I had forgotten to measure energy costs for burner self-fueling! I've updated the main post and my numbers spreadsheet with the new experiments. Surprisingly, burner fuel self-insertion is *much* more expensive than normal inserts, and also not round numbers at all, even when inserting from a chest. It's also strongly direction-dependent, unlike all other inserter mechanics. There's other oddities with it, too: Even if the burner is completely empty, and stack bonus is enabled, the inserter will only grab one fuel for itself. However, if it had already picked up a stack of multiple coal at the time when it runs low on fuel, it will self-load with that full stack. Needless to say, this makes it tough to model the precise average cost of fuel self-insertion.

In the end, I added a fudge-factor of 1.25x to the measured cost of self-fueling from the near lane, to account for the additional cost incurred when the inserter diverts in mid-swing to self-fuel. It barely changes the numbers, though - total miners goes from ~17.591 to ~17.602, and net energy goes down from 34.417MW to 34.416MW.

[Exasperation]

Ah, so your energy cost is the delta between idle cost and total energy cost. Fair enough.

It's not really surprising that burner self insertion is more expensive, if you've ever played with a configurable inserter mod. Self insertion involves changing arm extension, which is usually much slower/more expensive than inserter rotation. The stack bonus stuff is weird, though.

[BlakeMW]

I'd think that the most expensive part of self-fueling is the 2 coal a burner inserter picks up before it starts doing anything useful and maintains. Which means the real cost is:

Burner inserter: 3 Iron, 2 Coal
Basic Inserter: 4 Iron, 1.5 Copper + ~1/3 a small power pole (0.33 wood, 0.33 copper)

Burner inserter is still cheaper especially in terms of crafting time because coal is a basic resource which requires zero intermediate processing - altough it does still narrow the cost difference.

[featherwinglove]

There was an interesting slight discrepancy between horizontally aligned inserters and vertically aligned ones

I wonder if that's related to the stretchy trains problem...