Belt balancers: sizes, splitter counts and throughput

What a belt balancer actually is

A splitter takes up to two input belts and divides the flow evenly across up to two outputs. Chain splitters together and the even division spreads across more lanes. A balancer is that chain, arranged so that every input can reach every output.

This matters in two situations. Pulling from a mine, uneven patch depletion leaves some belts full and others empty, and a balancer evens them out so the consumer sees one steady flow. Feeding a smelter column, it stops the first few furnaces taking everything and starving the rest.

How many splitters a balancer needs

For sizes that are powers of two, the answer follows a fixed structure: log₂(N) stages, each holding N/2 splitters.

Balancer	Stages	Splitters per stage	Total splitters
2 to 2	1	1	1
4 to 4	2	2	4
8 to 8	3	4	12
16 to 16	4	8	32

The familiar 4-to-4 built from four splitters and 8-to-8 built from twelve both fall out of that formula rather than being facts to memorise. Note how the total grows: doubling the width roughly triples the splitter count, which is why 16-wide balancers are rare in practice even in bases that could use them.

Sizes that are not powers of two

There is no clean formula here, and this page will not invent one. A 5-to-5 or 6-to-6 balancer is built by taking the next power of two and trimming the unused lanes, so its splitter count depends on which design was trimmed and how.

Wanted	Built by trimming
3 to 3	4 to 4
5 to 5	8 to 8
6 to 6	8 to 8
7 to 7	8 to 8
9 to 9	16 to 16
12 to 12	16 to 16

Throughput: the part that catches people out

A balancer being "balanced" and being "throughput-unlimited" are separate properties, and mixing them up is the most common balancer mistake. Balanced means each output receives an equal share. Throughput-unlimited means any input can send a full belt to any output without the network choking in between.

The distinction bites when inputs and outputs are unequal in number. Feeding 1 belt into 4 outputs divides the flow; each output gets a quarter of a belt, no matter how balanced the design is. On transport belt at 15.0 items per second, that is 3.75 items per second per output.

Shape	Total in	Per output	Outputs saturated?
4 to 4	60.0/s	15.0/s	Yes
2 to 4	30.0/s	7.50/s	No, outputs run partly empty
4 to 2	60.0/s	30.0/s	Yes
3 to 2	45.0/s	22.5/s	Yes
8 to 8	120/s	15.0/s	Yes
1 to 4	15.0/s	3.75/s	No, outputs run partly empty

Figures are for transport belt. The pattern holds at every tier; only the absolute numbers change.

Belt tiers and what they carry

Throughput is not a separate stat in the game data. It follows from belt movement speed: each tile of belt holds eight items across two lanes, and the game runs at sixty ticks per second.

Belt	Tiles / tick	Items / second	Relative to transport belt
transport belt	0.03125	15.0	1.00x
fast transport belt	0.0625	30.0	2.00x
express transport belt	0.09375	45.0	3.00x
turbo transport belt	0.125	60.0	4.00x

The tier added in 2.0 carries 60.0 items per second, 4 times the base belt. That changes balancer planning more than it first appears: a lane count chosen to move a given throughput on older belts is oversized once the same flow fits on fewer faster belts, and every lane removed takes a stage of splitters with it.

Choosing a balancer for a non-power-of-two width

The trim table above tells you which larger design to cut down from, but not all trims are equal. A 5-lane balancer trimmed from an 8-to-8 uses twelve splitters; a 6-lane uses the same twelve; a 7-lane also uses the same twelve. The three designs differ in which lanes are left unconnected, not in how much hardware they need. Once you are trimming from the same parent, the splitter count is fixed and the choice is about which unused outputs are easiest to leave dangling.

That changes at the next boundary. A 9-lane balancer trims from a 16-to-16 and costs thirty-two splitters, nearly triple a 7-lane design. In practice this means there is a dead zone: widths of 5 through 7 are cheap, but 9 through 15 carry the full 16-wide cost even if you only use nine lanes. When your bus is nine belts wide, the honest choice is often between accepting the overbuilt 16-lane balancer and redesigning the bus down to eight.

Wanted width	Parent balancer	Splitters	Practical note
3	4-to-4	4	One unused lane; trivial trim
5–7	8-to-8	12	Same hardware regardless of width
9–15	16-to-16	32	Consider whether 8 lanes suffice

Balanced versus throughput-unlimited: which one you need

A balanced output is what most players mean by "a balancer": every output belt receives the same share of whatever arrives. A throughput-unlimited balancer guarantees something stronger — that any single input can saturate any single output even when all other inputs are empty. The first is an even-division property. The second is a non-blocking property. They overlap but are not the same.

Consider a mine patch feeding a 4-to-4 balancer. As miners deplete, some input belts run full while others trickle. A merely balanced design divides the trickle evenly, but a belt that receives a quarter of a nearly-empty input may not be able to pass a full belt from a different input through to the same output. A throughput-unlimited design can. If your consumers care about total throughput rather than evenness — furnaces pulling from a shared bus, for example — the non-blocking property is the one that matters.

Evenness matters when the outputs feed identical machines that each need a fair share: a row of assemblers, a set of labs, a smelter column. In those cases a balanced output prevents the first machine draining the belt. The two properties coincide for equal input and output counts that are powers of two, which is why the distinction is easy to miss until you build an unequal shape.

The shapes table makes the boundary concrete. A 4-to-2 balancer can saturate both outputs (each receives two input belts' worth) but a 2-to-4 cannot saturate any output (each receives half a belt). The first is throughput-unlimited in the direction that matters; the second is merely balanced. If you are merging two belts onto four furnaces, balanced is correct and the half-belt per furnace is by design. If you are pulling four belts of ore onto two belts of furnaces, you need the throughput-unlimited direction.

What turbo belts change about an existing bus

A turbo belt moves sixty items per second, four times a yellow belt and double a red one. Upgrading belts on an existing bus does more than speed up the items already on it. It changes how many lanes are needed for the same total throughput, and every lane removed removes a stage of splitters at every balancer the bus crosses.

A main bus built with eight yellow belts carries 120 items per second. The same flow fits on two turbo belts. The balancer at the end of that bus shrinks from a 16-to-16 (thirty-two splitters) to a 4-to-4 (four splitters) if the bus itself is narrowed — a reduction that has nothing to do with the belts' own speed and everything to do with how many of them there are.

Target throughput	Yellow belts	Red belts	Blue belts	Turbo belts
30 items/s	2	1	1	1
60 items/s	4	2	2	1
120 items/s	8	4	3	2
240 items/s	16	8	6	4

This does not mean turbo belts are always the right upgrade. A bus that is already eight lanes wide with taps along its length gains no balancer savings from merely upgrading the belt items in place; the splitters are still sized for eight. The savings appear when the bus is redesigned narrower, which is a larger decision than a belt upgrade. Turbo belts also require power and a later technology, and a yellow belt bus that is working is rarely the bottleneck worth fixing first.

When a full balancer is overkill

Not every junction needs every input to reach every output. A splitter alone merges two belts into two outputs without balancing across more than two lanes. Two splitters in a row merge four into two. These simpler arrangements use fewer splitters than a full balancer and are the correct tool when the goal is to combine flows rather than to redistribute them evenly.

The distinction is whether the outputs need to be interchangeable. Feeding two belts of coal into a single boiler line needs a merge, not a balancer: the boilers do not care which belt the coal arrived on. Feeding four belts of mixed ore into four identical smelter columns needs a balancer, because each column should receive the same blend. Building a full 4-to-4 balancer where a two-splitter merge would work costs hardware and footprint for a property you do not use.

A useful heuristic: if you would be happy with any output belt receiving any input belt's full flow, you need throughput, not balancing. If you specifically need every output to receive the same fraction regardless of how the inputs are filled, you need balancing. Many factory junctions are the first kind.

Why 16-wide balancers are rare in working bases

The formula gives a 16-to-16 balancer at thirty-two splitters across four stages. That is a large footprint, and the splitter count itself is not the main cost. Each stage adds belt length, and the total structure spans roughly sixteen tiles in each direction. In a bus that is already crowded with underground belts and power poles, finding a sixteen-tile rectangle for a balancer is often harder than supplying the splitters.

There is also a diminishing-returns argument. A 16-wide bus on yellow belts carries 240 items per second. Most intermediate products — circuits, gears, plastic — do not need that rate in a single stream. When a bus reaches sixteen lanes, it is usually carrying several materials side by side, each tapped independently. Balancing across all sixteen mixes materials that should stay separate. The balancers that actually get built in large bases are more often 4-to-4 or 8-to-8 at material-specific tap points, not one giant 16-wide at the bus terminus.

Turbo belts shift this calculus. Four turbo belts carry the same 240 items per second that sixteen yellow belts do, and a 4-to-4 turbo balancer uses four splitters in a fraction of the footprint. The rare 16-wide yellow balancer becomes an unnecessary 4-wide turbo balancer once the bus is upgraded — another reason belt tier and lane count should be planned together rather than treated as independent choices.

The footprint argument cuts both ways. An 8-to-8 balancer uses twelve splitters across three stages and fits in a compact rectangle that most bus designs can accommodate. A 16-to-16 uses thirty-two and demands a straight run of belt long enough to be worth planning around. In megabase builds where every tile is counted, the 16-wide balancer is sometimes replaced by two 8-wide balancers running in parallel, each handling half the lanes. The two arrangements carry the same total throughput but split the footprint into two smaller blocks, which is easier to route around a crowded bus even though the total splitter count is slightly lower (24 versus 32).

Applies when…

• Splitter counts assume a network where every input can reach every output. Simpler designs that merely merge or split lanes need fewer splitters and do not balance.
• Throughput figures assume both lanes of the belt are full. A single-lane feed carries half the stated rate.
• Counts for sizes that are not powers of two depend on which larger design was trimmed, so no single number is quoted for them.
• Belt speeds are read from the prototype data for the stated game version. Mods that add belt tiers change these figures.

Related

• Production Ratio Calculator — check whether a flow fits on the belts you have
• How production ratios actually work