Factorio Forums

I've been thinking about a power usage of part of my Factory and came up with unexpected conclusion: Assembling Machine 3 is NOT power-efficient. More yet, in some instances using AM-2 is much more rational (from several points of view) than AM-3. Now I try to explain why I think this is the case. My conclusions apply to vanilla version of the game, as quality introduce a lot more factors to consider and may change results a lot (also, I've yet to master/bore of vanilla game, so I've never tried space extension). But you can apply my comparative approach to quality mod too. First, statistics for different Assembling Machines:

.. . . . . .. . . . . . . Power . . . . . . . . . . . . .
Machine .. Speed .. Active . . Idle . . Slots . Pollution
AM-1 . . . 0.50 . .. 75 kW . . 2.5 kW . . 0 . . . 4/m . .
AM-2 . . . 0.75 . . 150 kW . . 5.0 kW . . 2 . . . 3/m . .
AM-3 . . . 1.25 . . 375 kW .. 12.5 kW . . 4 . . . 2/m . .

Now, lets consider one AM-3 without any modules. To produce 1 craft with it we will need to spend an amount of energy given by a formula:

E = (375 kW + 12.5 kW) * (T / 1.25) = [(375 kW + 12.5 kW) / 1.25] * T = P * T
P = (375 kW + 12.5 kW) / 1.25 = 310 kW

Where E is energy, T — base time needed to produce 1 craft with speed 1 and P is normalized (or reduced) to speed 1 power consumption. As you can see, it's easier to compare normalized power P for different cases with the same product as you can forget about time T that comes as factor into energy E. Therefore henceforth I will calculate and compare only normalized power.

To produce 1 x AM-3 the following resources are needed: 2 x AM-2 and 4 x SP-1. Now, consider this: instead of producing Speed Modules 1 and then AM-3 we can produce 4 x EF-1 (Efficiency Module 1) and stuff them into our already built 2 x AM-2. What will we get? Well, of course you will need more space for two machines and additional investment of belts and inserters, but space in Factorio is cheap and investment is one-time thing. On the other hand, the energy is spent constantly. Two Efficiency 1 Modules will reduce active power consumption by 60%, which will introduce a 0.4 power factor in our formula for normalized power P:

P = 2 x (0.4 x 150 kW + 5 kW) / (2 x 0.75) = 86.6667 kW

Which is more than 3.5 times cheaper than with AM-3! And this "outdated" setup will work faster, as 2 x 0.75 = 1.5 > 1.25! (Also, take notice, that "number of machines" factors cancel each other in the formula just above, which is logical, therefore I will not use them henceforth, unless they are relevant). Only if you get full (and expensive) upgrade, adding 3rd EF-1 into shiny new AM-3 to get the maximum possible 80% power reduction (resulting in the power factor 0.2) you will achieve some progress (19% comparative reduction):

P = (0.2 x 375 kW + 12.5 kW) / 1.25 = 70 kW

But AM-3 have as much as 4 module slots and we haven't made full use of them yet in our calculations. So lets consider the most energy efficient setups for AM-2 and AM-3. For AM-2 it will be 2 x EF-2 which reduce active power consumption by the maximum 80%:

P = (0.2 x 150 kW + _5.0 kW) / (1.0 x 0.75) = 46.6667 kW

Now, for AM-3 it will be 3 x EF-3 + SP-3 which gives maximum power reduction AND speeds up production by 50% (factor 1.5) therefore reducing normalized power:

P = (0.2 x 375 kW + 12.5 kW) / (1.5 x 1.25) = 46.6667 kW

And it's the same value! But it's SOOOOO SOOOOOOO much more resource expensive investment to just save space and reduce pollution! Yes, the numbers for last one (normalized too) look much better:

Poll_1 = 1.0 x 4/m / (1.0 x 0.50) = 8.0/m
Poll_2 = 0.2 x 3/m / (1.0 x 0.75) = 0.8/m
Poll_3 = 0.2 x 2/m / (1.5 x 1.25) = 0.2133/m

Of course, the main use for module slots in Assembling Machines is to stuff them with Productivity Modules, as you can't put them into beacons. For the maximum achievable productivity, there is no other way but use AM-3 with its 4 slots. (Rationality of on what stages of the game how much productivity is good or just good enough lets put off for another discussion). But there is a lot of products in the game that you can't produce with productivity modules, including intermediate ones (all the components for Production Science Pack, for example). The only modules for such "end-products" are speed and efficiency ones.

Now, lets think about what happens when we stuff AM-3 full of SP-3's. Yes, production speed is beautiful to look at. For something like Copper Cables or Iron Gear Wheels you will most likely need fully upgraded Bulk Inserters just to keep up. But numbers for energy consumption and pollution are just awful!

f_power = 1.0 + 4 * 0.7 = 3.8
f_speed = 1.0 + 4 * 0.5 = 3.0
P = (3.8 x 375 kW + 12.5 kW) / (3.0 x 1.25) = 383.3333 kW
Poll = 3.8 x 2/m / (3.0 x 1.25) = 2.0267/m

You will save space at a great (and, imho, unacceptable) cost of power usage and pollution. Hereupon, I think, my point is made on all items stated in the title.

There are, also, Beacons to consider and they introduce additional variables into formulae above. But I've though about this puzzle for some time already and couldn't come with any better solutions that beacon-less ones already mentioned. The most power efficient configuration with Beacons is one where 1 Beacon influence 10 Assembling Machines (or 10 other such 3x3 structures for that mutter). The Beacon contains 2 x EF-3 and the 10 x AM-3 each contain 2 x SP-3 + EF-1 + EF-2. The formulae become:

f_power = 1.0 + 2 * 0.7 - 0.3 - 0.4 - 1.5 x 2 x 0.5 = 0.2
f_speed = 1.0 + 2 * 0.5 = 2.0
P = [10 x (0.2 x 375 kW + 12.5 kW) + 480 kW] / (10 x 2.0 x 1.25) = 54.2 kW > 46.6667 kW
Poll = 0.2 x 2/m / (2.0 x 1.25) = 0.16/m < 0.2133/m

Which is better pollution-wise, but worse energy-wise (although close, +16%). And other solutions I tried are much worse yet. The 1:10 configuration looks like this (B stays for Beacon and 3 — for Assembling Machine 3):

/`\/`\/`\/`\
|3||3||3||3|
\_/\_/\_/\_/
. . . . .. .
. . . . .. .
/`\ ./`\ /`\
|3| .|B| |3|
\_/ .\_/ \_/
. . . . .. .
. . . . .. .
/`\/`\/`\/`\
|3||3||3||3|
\_/\_/\_/\_/

Configurations where one Machine is influenced by more than one Beacon are much less power-efficient in total because ratio of Beacons to Machines is much worse than 1:10 (as there is no way to place them such), as well as efficiency with which Beacons project their effects inversely proportional to the square root of a number of Beacons influencing one building. All this highly increase power cost of every craft. Therefore, imho, the use of Beacons is reasonable only when you are dealing with drawbacks of productivity modules. In this case two or more Beacons per building can be more efficient solution despite all the topological disadvantages of placement and diminishing returns of multiple transmissions. I can't help but cringe a little when I see setups with tons of Beacons and SP-3's with no EF-3's at all and think to the authors of such things: "Have you ever considered your energy spendings?"

What do you people think about all this? Did I miss something in my considerations? Can you come up with better puzzle solution of 1:10 beacon problem (different placement of different modules maybe)?

.

FWIW, there are a variety of optimization criteria.

E.g. if you commit to using lots of productivity 3 modules, but aren't yet at the point where production costs are trivial, speed beacons are valuable in the way that, for a particular production target, the extra cost of speed beacons is outweighed by the decrease in the number of production 3 modules you need. And that may be something you're more concerned about than the cost of building extra power plants and extra pollution.

And I understand for super end-game factories where people start optimizing for how quickly your computer can actually run Factorio, minimizing the number of assemblers becomes super important so that there is less for your computer to do every tick. And since speed beacons and solar farms are free in this metric, the only remaining cost is how much compute the extra pollution costs you by spreading your cloud further. I don't know the relative concerns about that.

B@R5uk wrote: Thu Mar 05, 2026 11:37 am Did I miss something in my considerations?

Yeah, years of experience and discussions made by other players.

Two factories. One in your way:
https://factoriolab.github.io/1.1/list? ... VnB1g&v=11

And other with productivity+speed:
https://factoriolab.github.io/1.1/list? ... sfQ__&v=11

Second one requires more energy but this can be easily solved with nuclear power. It may surprise you, but it produce less pollution. Though it is not a main reason to choose it. Now look at the main thing - resources. Your way requires three times more resources. And much more complex logistic.
If you are considering cost - most of the second setup pay off in less than 3 hours, after that it is clear surplus. It takes much more time to setup additional outposts and logistic in your variant.

B@R5uk wrote: Thu Mar 05, 2026 11:37 am I can't help but cringe a little when I see setups with tons of Beacons and SP-3's with no EF-3's at all and think to the authors of such things: "Have you ever considered your energy spendings?"

Yes, and no. We all try to optimize our factories, and it seems there are different optimization goals, some contradicting.

For example, I optimize for resource usage. Usage in terms of raw input ratio to output ratio of the whole factory, and in terms of buildings used for the factory (factory footprint). I don't consider a specific machine, I consider the whole factory. It starts with mining resources with mining drills and pumpjacks, and it ends with science pack consumption in the lab.

This optimization is a 3 step process. The 1st step is to understand resource usage is minimized by using the best possible productivity modules for every recipe that allows it. The 2nd step is to speed up production with the best possible speed modules (in beacons, because module slots in the machines are occupied by productivity modules) so the factory size in general shrinks. This concept results in a certain input:output ratio, and the final step is to scale it to your desired production goal in terms of volume or SPM (science per minute).

Energy usage is no part of this, because energy is the first resource that becomes de facto infinite and unlimited (the copper wire of a mediocre small power pole is able to carry 100+ GW). The same with pollution. It just dissipates and vanishes. Turn off the display of the pollution cloud and it's no more. Aggro'ed enemies are kept away with defensive perimeter walls. Nests from expansion are destroyed by artillery. Fully automated and no effort.


Your optimization goal seems to be energy usage per craft. However you don't consider the whole factory. You have to include resource mining and production of every intermediate. You only look at single machines, but production is a pyramid. You need vast amounts of ore (mining drils), and this is condensed to intermediates, and the final product just needs a few machines. However the most saving is applied if you don't need some machine at all, and this is achieved by using vastly less mining and intermediate production by increasing productivity. @angramania showed an example.

To pick up your complaint, I cringe a little when I see factories without productivity modules. So much waste of resources! The best usage of ore is if it's not necessary to mine it in the first place.