Inserter power usage (aka don't use fast inserters)

[5th Earth]

So, a reply to a comment I made on a youtube video prompted me to break out the spreadsheets...

The question is, what's the most energy efficient way to use inserters? In other words, how do we maximize items per second per kW? Well, I crunched numbers and found some interesting things.

First, my base assumptions: all my calculations were based on the 0.18 chest-to-chest IPS listed on the wiki. I don't think anything changed for 1.0, and I didn't want to get bogged down in the minutia of how inserters and belts interact.

Now, some conclusions:

1: Idle power usage is basically irrelevant for performance calculations. Every inserter's active power usage is so much bigger than its idle power that the differences between different types of inserters are obliterated the moment you start moving any reasonable quantity of items.
2: Burner inserters suck. But that's not really a surprise. By far the highest energy usage per IPS. BUT, they are the only ones with zero idle power usage. So for extremely low IPS, (less than .027, or about 1.6 items per minute) they are actually the best option. There may be scenarios where this is relevant, I guess?
3: Long arm inserters use almost exactly as much power per IPS as regular inserters. They have a higher maximum throughput due to their higher speed, but this is exactly cancelled by their increased energy usage. I assume this is by design.
4: Fast inserters suck. I was surprised by this, but unless you actually need the high IPS in limited space, you are always better off using multiple regular inserters instead. 3 regular inserters will move more IPS, for less energy, than one fast inserter. You should only ever use fast inserters when space constrained.
5: Filter inserters are like fast inserters that use more power. This should be obvious from their stats, but you really shouldn't ever use them unless you have to.
6: Stack inserters suck in the early game. Like fast inserters only more so, any combination of regular or fast inserters that produces the same IPS will use less power.
7: Stack inserters are much better by the time you research stack bonus 7. They actually beat fast inserters on IPS/kW, by a significant margin. You should always use them instead of fast inserters for high IPS applications. (another reason fast inserters suck). In fact, the IPS/kW gets so good that they almost beat even regular inserters for efficiency, assuming the regular inserter is at max capacity.

I want to repeat the biggest conclusion here, because I think it will actually impact my designs: Fast inserters are really inefficient. Unless you are space constrained, you are always better off using multiple regular inserters. And, in the late game, stack inserters are always better for high IPS in limited space and you pretty much should never use fast inserters.

[5thHorseman]

Huh. I never did the math but always just assumed the opposite of your conclusion, that a single fast inserter was in all ways better than 2 regular inserters.

I guess it's the price you pay for compactness.

[Yoyobuae]

Don't think many care all that much about power efficiency. And if you did you'd be worrying 100x more about drills/furnaces/assemblers.

[sparr]

2: Burner inserters suck. But that's not really a surprise. By far the highest energy usage per IPS. BUT, they are the only ones with zero idle power usage. So for extremely low IPS, (less than .027, or about 1.6 items per minute) they are actually the best option. There may be scenarios where this is relevant, I guess?

I wonder if there's a situation where I have an inserter handling fuel and inserting fewer than 1.6 items per minute...

Screenshot_20200823_012608_50percent.jpg (97.17 KiB) Viewed 9388 times

[Koub]

@OP : I'm not surprised in the slightest manner with your conclusions. Eyeballing consumption and speed, it seemed to me obvious. The only things eyeballing said nothing is the item per minute threashold a burner inserter gets more efficient than a regular inserter, and the stack bonus research at which stack inserters get better than fast inserters.

However you didn't tell us what fuel you used for the burner inserter : burner inserters have to refuel themselves more often with coal than with nuclear fuel. That should have an impact on the overall consumption. Or maybe you didn't account for that extra swing to fuel themselves ?

Overall, the fast inserters being less efficient than normal inserters ips wise is kind of expected. It's how tiered entities usually work.
AM1 : 77.5 kW, speed 0.5
AM2 : 155 kW, speed 0.75
AM3 : 388 kW, speed 1.25
if you normalize that to powerconsumption for equivalent speed 1, you get :
AM1 : 155 kW per speed 1 craft
AM2 : 207 kW per speed 1 craft (rounded to the closest kW)
AM3 : 310 kW per speed 1 craft (rounded to the closest kW)

Or said otherwise, 6 AM3 produce as fast as 10 AM2 and 15 AM1.
6 AM3 : 2328 kW
10 AM2 : 1550 kW
15 AM1 : 1162.5 kW

As you get up tiers, you get more features (fluid input, more module slots, smaller footprint) as you give up efficiency.

As an additional note, you're restricting yourself to inventory to inventory transfer. You did well, because it's the only way to get predictable and repeatable results. However if you were to test belt to inventory or inventory to belt transfers, I suspect your results would significantly vary (but you'd have to make them for all 4 orientations).

[Khagan]

regular inserters will move more IPS, for less energy, than one fast inserter. You should only ever use fast inserters when space constrained.

As Koub has pointed out, this is not surprising. The simplest tool that does the job is usually also the cheapest.

Stack inserters are much better by the time you research stack bonus 7. They actually beat fast inserters on IPS/kW, by a significant margin. You should always use them instead of fast inserters for high IPS applications.

Your statement about IPS/kW is true (for chest-to-chest), but your conclusion is not, because you are overlooking the hugely higher capital cost of the stack inserter.

If we have a fast inserter running flat out chest-to-chest, and replace it by a stack inserter moving items at the same rate, the energy consumption (ignoring drain) drops from 46kW to 33kW. But the stack inserter costs 90 resources to make, and the fast inserter only 12.5. Even using solar, 13kW of power only costs about 29 resources to produce (and nuclear power about a twentieth of that). So it's much cheaper to use the fast inserter, and increase power production to compensate. The stack inserter wins only if it is replacing three or four fast inserters (and you are using solar power).

And of course, belt-to-chest or chest-to-belt, the fast inserter is more power efficient to start with, since the stack inserter is only about twice as fast instead of four times. Once again, the simpler tool is cheaper.

[5th Earth]

@OP : I'm not surprised in the slightest manner with your conclusions. Eyeballing consumption and speed, it seemed to me obvious. The only things eyeballing said nothing is the item per minute threashold a burner inserter gets more efficient than a regular inserter, and the stack bonus research at which stack inserters get better than fast inserters.

However you didn't tell us what fuel you used for the burner inserter : burner inserters have to refuel themselves more often with coal than with nuclear fuel. That should have an impact on the overall consumption. Or maybe you didn't account for that extra swing to fuel themselves ?

(...)

As an additional note, you're restricting yourself to inventory to inventory transfer. You did well, because it's the only way to get predictable and repeatable results. However if you were to test belt to inventory or inventory to belt transfers, I suspect your results would significantly vary (but you'd have to make them for all 4 orientations).

Yeah, I haven't figured out the break-even for stack research that makes stack inserters better than fast inserters because a full table of stack inserter IPS per research level wasn't on the wiki. At some point I'll have to do my own measurements and figure it out.

Re: the type of fuel for burner inserters, you're right, I hadn't taken that into account. The wiki says one coal will last for 57 swings, so that means the inserter will use 1/57 of its fuel to power itself. That's 1.8% lost. Solid fuel lasts 178 swings, or 0.6%. So worst case scenario, we'll actually need to transfer no more than one item every 1.6*1.018 minutes, which is only a few seconds longer. It has an effect but it's pretty small. And if you are using it for a nuclear centrifuge because you are a genius madman like sparr, then the energy lost will be trivial.

I'm aware that inventory-to-inventory leaves out most of the real-use applications we'll see. And of course, I know you often can't reliably use regular inserters on the faster belt types, so that's also a limitation to the applicability. My assumption is that even though the specific kW/IPS used changes in different configurations, the trends between types of inserters will stay the same except in odd edge cases. Though I know that's an assumption that may be wrong, especially for stack inserters that act really weird in some arrangements. YMMV.

[5th Earth]

If we have a fast inserter running flat out chest-to-chest, and replace it by a stack inserter moving items at the same rate, the energy consumption (ignoring drain) drops from 46kW to 33kW. But the stack inserter costs 90 resources to make, and the fast inserter only 12.5. Even using solar, 13kW of power only costs about 29 resources to produce (and nuclear power about a twentieth of that). So it's much cheaper to use the fast inserter, and increase power production to compensate. The stack inserter wins only if it is replacing three or four fast inserters (and you are using solar power).

And of course, belt-to-chest or chest-to-belt, the fast inserter is more power efficient to start with, since the stack inserter is only about twice as fast instead of four times. Once again, the simpler tool is cheaper.

In the long run, the initial capital cost will be overshadowed by the continuous and ongoing fuel costs. I don't know what the break even point is, but it's like buying a more expensive car that has better fuel economy. Eventually the money you save on fuel will be more than the initial price difference. And since the lifespan of an inserter is infinite (barring biter attacks) we have as much time as necessary to recoup the initial investment.

You're right that restricting myself to chest-to-chest transfers excludes the differences involved in belt-involved transfers. It was a simplification to find out the initial trends, since there are so many variables otherwise. That'll be for future research I suppose.

[Khagan]

In the long run, the initial capital cost will be overshadowed by the continuous and ongoing fuel costs. I don't know what the break even point is

For solar power there are no fuel costs, and no break-even point, ever.

For mature nuclear power, it takes something like 1600 hours at capacity for the fuel costs to add up to the same as the capital cost of solar. The break-even point for replacing a fast inserter by a stack inserter would be over 4000 hours of continuous operation, and that is assuming both

1. that uranium ore is as valuable as iron and copper (it isn't, not really) and
2. that a unit of resources in thousands of hours time is as valuable as the same unit now (it isn't, not remotely).

Any vaguely sane discount rate for future resources will result in the same conclusion as for solar: no break-even point.

[Fact.Oreo]

While this all sounds calculated and precise it feels too theoretical to me, so I set up a simple experiment:

I have 2 identical power systems on separate power grids containing 1 Boiler, 2 Steam engines and a chest for Coal.
System 1 is fed by a Burner inserter (stack size 1).
System 2 is fed by a basic electric Inserter (stack size 1).

System 1 was started up by placing 50 coal into the chest, at which point the Burner inserter loaded itself and got to work (and reloaded when there were about 5 coal left in the chest).
System 2 was started up by placing 1 Coal into the Boiler and 49 into the chest, at which point the electric Inserter got to work.

Each power grid is serving an identical bank of Electric mining drills (productivity +50%) pumping directly into chests (details below).
Test 1: 4 mining drills
Test 2: 16 mining drills

When each system ran out of coal and steam, here is how much ore was produced:

Test 1:

System 1 (Burner Inserter): 1596
System 2 (Electric Inserter): 1656 <-- produced 60 (~3.8%) more ore than System 1
System 2 was roughly 3.8% more fuel efficient (approx 1.9 pieces of Coal).

Test 2:

System 1 (Burner Inserter): 1584
System 2 (Electric Inserter): 1648 <-- produced 57 (~4.0%) more ore
System 2 was roughly 4.0% more fuel efficient (approx 2 pieces of Coal).

So as you ramp up the volume the difference gets magnified, but is it enough to definitively choose one inserter over the other? No, but I found it interesting anyhow.

Either way, if I run out of Coal and don't have any laying around to manually refuel my Boilers...well, there goes all my production anyway.

[theolderbeholder]

The advantage of regular inserters is only given as long as they do not need to pick up from red oder blue belts -which only fast inserters can do relieably? Or is this deprecated?

[foamy]

The advantage of regular inserters is only given as long as they do not need to pick up from red oder blue belts -which only fast inserters can do relieably? Or is this deprecated?

Nah, it's still true. You can even manage to mess up fast inserters if they're pulling from undergrounds and the belt isn't near-saturation. Can be worked around a split onto a tag end, something I regularly do in my boiler plants to allow burner inserters to work properly there.

[Fact.Oreo]

Foamy, that’s actually a great idea, I’ve never thought about that. Feed boiler area with high speed belts to increase quantity of boilers but break off to small yellow belt sections to feed burner or regular inserters so they can grab it.

It seems so obvious now that I type it out...

I just always put my rows of boilers too close together to add faster belts and split anything off.

[foamy]

Foamy, that’s actually a great idea, I’ve never thought about that. Feed boiler area with high speed belts to increase quantity of boilers but break off to small yellow belt sections to feed burner or regular inserters so they can grab it.

It seems so obvious now that I type it out...

I just always put my rows of boilers too close together to add faster belts and split anything off.

What I do is run the faster belts down the same row the inserters are on, dodging under them with undergrounds, then back up for the splitters. I have a really neat 120 boiler layout (triple boiler rows to either side of the fuel feed, twenty long) that runs on two red belts of coal and drives ~216MW (~90% of the belt capacity) of steam engines. It needs full inserter capacity researches to work at full capacity with burner inserters, but it's quite easily possible to either feed better fuels or to use better inserters, reserving the burner ones purely for a bootstrap power plant. Also, it can be built in stages and without, initially, needing anything more than the starting tech of wood power poles, steam engines, boilers, burner inserters, piping, and yellow belt; all the extra bits can be retrofit, painlessly.

[KeepResearchinSpoons]

Oh, what you've done is nice and games but first consider this:
wiki for inserters (thank bilka and all the crew for it exists)
then you might get the idea of how it works.
Then you could get the tests with your ticks and combinators with lua scripts and things wired up
Good luck!

There are at least 3 known usages for burner inserters that are widely spread nowadays so you could calc how much efficiency is lost each tome or per inserter per minute. Anti burnout boiler setups being the most used.

Also it costs 3 irons and very few coals up front so what are the situations you use the b. inserter with? (Imagine a furnace with burner drill in the distance left for 7 minutes or so and the fact it is cheaper to have a burner with a wood chest then having it stalled and loose the products. The gains are more then the cost obviously.)

Red inserters are a hidden grace actually since they allow belting between the assemblers' lines.
Blue inserters are known for supplying the coal to the furnace lane as it is cheaper then having a splitter and yellow ones are even cheaper at this role.

Well most of us know the stack inserters are useless with items that have low stack sizes: imagine arty shells or pretty much any gear. They are pretty the same speed with blue ones and have a huge consumption in comparison. SAME applies to a belt that is NOT compressed to a whooping point. All your main busses that went wrong and introduced balancers instead of priority shifts as a courtesy to 0.14 bus types and taps. Well it is probably a design issue but having a radiated belt costs you lots of swipes for a stack inserters so remeber this as wel when talking about efficiency.
Also since "the early game" for stack inserters usually starts at cap bonus 2 (red green tier!) when they are about twice as fast as blue ones I doubt it is so much one sided as "loose to blue ones, period" Should we get a more elaborate calculation results with different usage scenarios?

As a design gamey you can count the number of inserters with 0 grades for 1-1 copper wire - green circuit build that you need to use for full throughput (the common answer is:

one yellow and one blue is enough till tier 2 (aka cap bonus 2)

)

And as a bonus, no the idle power usage is not to be counted off the board so easily. Since you have a TON of inserters all over the place it builds up to some MW or GW which is even more so with stack inserters (the ones rarely active as an crying example). Just shift the dot aside for each 10/100/1000 inserters you use at your build and remember your smelting array has usually 96 inserters per yellow lane each (also think of cap2 bonus here for how much sci costs to how much swipes are cut down with it in grand total).

[4xel]

Filter inserters can be used with clocks for UPS efficient design (deactivated filter inserters are dormant).

energy per item isn't the only metric to optimize against.

In addtion to other ewamples already given, burner inserter can be used to feed gun turrets that aren't expected to fire often or ever. It's not just more resource efficient, it's also arguably sturdier (you don't need power)

[blazespinnaker]

Another perspective is that placement becomes very critical in any time sensitive approach (achievements like there is no spoon, scenarios like rocket rush, wave defense). Both hand placing and bot placing are time proportional to # of entities.

Anything that can reduce entities required is vital when time constrained. A good example are steel furnaces, which are very expensive, but they cut down entities required in half.

But yeah, that aside, I used to prefer to use burner inserters sometimes when there is limited swinging involved and powered by coal. Eg, feeding furnaces. My biggest problem is they slow down time to initially feed the furnace with coal, so I stopped using them. They cause a lot of pollution as well I believe.

Red inserters are magical in so many different ways, in particular for the ability to form creative factories that are semi universal. Designs where you can just change the recipe of the assembler and it still works are very interesting to me. I'm trying to convert everything to red inserters.

Basic inserters are useful for throttling throughput.