Conveyor Belts, Splitters and Manifolds in Satisfactory

Most factory problems that look like a production problem are a belt problem. A machine that idles for two seconds every minute is not underfed by its recipe; it is underfed by whatever is upstream of it.

Belt throughput by mark

Mark	Items/min	Unlocked at	Cost per segment
Mk.1	60	Tier 0 — HUB	1 Iron Plate
Mk.2	120	Tier 2 — Logistics Mk.2	1 Reinforced Iron Plate
Mk.3	270	Tier 4 — Logistics Mk.3	1 Steel Beam
Mk.4	480	Tier 5 — Logistics Mk.4	1 Encased Industrial Beam
Mk.5	780	Tier 7 — Logistics Mk.5	1 Alclad Aluminum Sheet
Mk.6	1200	Tier 9 — Peak Efficiency	1 Ficsite Trigon

Two things to notice. The jumps are not even — Mk.2 to Mk.3 adds 150/min, but Mk.4 to Mk.5 adds 300. And belts draw no power at all, at any mark. Upgrading a belt line costs materials and nothing else afterwards.

Belt segments are 1 to 56 metres long, which matters when you are running a line across a gap: a single segment can bridge more than you would think, and fewer segments means less clutter.

The bottleneck is almost always the belt, not the machine

A Miner Mk.1 on a pure iron node produces 120/min. Put it on a Mk.1 belt rated 60/min and you have thrown away exactly half the node. This is the most common early-game mistake and it is invisible, because nothing is flashing red — the miner simply backs up and idles.

The habit worth building: whenever you place an extractor, check the belt mark against its output rate before you walk away.

Splitters do not split evenly straight away

A Conveyor Splitter distributes round-robin across its outputs, but only across outputs that can currently accept an item. That distinction produces behaviour people read as a bug.

If one downstream machine is full and the other is not, the splitter sends everything to the one that will take it. When the full machine drains, it starts alternating again. So a freshly built line looks lopsided for a minute and then settles.

It only settles if total demand is below total supply. If your two machines together want more than the belt delivers, the split stays uneven and both run starved. No amount of waiting fixes an undersupplied line.

Manifold or load balancer

This is the long-running argument, and the honest answer is that a manifold is right almost always.

A manifold is one belt running past a row of machines, with a splitter tapping off at each one. Build cost is trivial: one splitter and a short belt per machine.

A load balancer is a tree of splitters and mergers that mathematically divides the input into equal shares before anything reaches a machine.

The case for the manifold:

• It reaches the same steady state. Once the buffers in each machine fill up, a manifold feeds every machine equally. The only difference is the warm-up period.
• It is far cheaper and takes a fraction of the space.
• It survives changes. Adding a machine to the end of a manifold is one splitter. Adding a machine to a balanced tree usually means rebuilding the tree.

The case for the balancer is narrow but real:

• Startup time matters to you. A manifold fills machines front to back; the last machine in a long row can take several minutes to reach full rate. If you are benchmarking or you genuinely cannot wait, balancing is instant.
• The line is deliberately undersupplied. If supply is below demand on purpose, a manifold starves the far end permanently while a balancer shares the shortage evenly. This is the one scenario where balancers are clearly correct.

If you are building a normal factory with enough input, build the manifold and spend the saved materials on more machines.

Smart and Programmable Splitters

Two upgrades exist and they solve different problems.

Smart Splitter filters one output by item type, with the other outputs taking Any, Overflow, or nothing. The classic use is a mixed belt from a container: filter Iron Rod down one line, let everything else pass.

Programmable Splitter allows multiple filters per output. It costs Computers and High-Speed Connectors, so it belongs to the point in the game where you are sorting a genuinely mixed belt rather than splitting two things apart.

The Overflow setting is the useful one that gets overlooked: set an output to Overflow and it only receives items when the other outputs are blocked. That turns a splitter into an automatic pressure valve feeding a sink or a storage buffer, which stops a backed-up line from stalling everything upstream.

Mergers and the priority problem

A Conveyor Merger combines up to three inputs into one output, round-robin. When total input exceeds belt capacity, all inputs get throttled equally.

The Priority Merger changes that: one input is served first, and the others only get the remaining capacity. The standard use is fuel. Feed a generator line from your main supply on the priority input and from a backup or byproduct source on the others, and the backup only contributes when the main supply falls short.

Lifts save more space than they look like they do

Conveyor Lifts move items vertically in the footprint of a single tile, come in the same six marks as belts with the same throughput, and cost one plate each. A Lift Floor Hole lets a lift pass cleanly through a foundation rather than routing around it.

For any factory built on more than one floor, lifts are what makes vertical layouts practical. The alternative — ramping belts up and down — eats floor space that could hold machines.

Common mistakes

Running a Mk.5 belt into a Mk.2 belt. The whole line runs at 120/min. Throughput is set by the slowest segment, not the average.

Building a load balancer for a fully supplied line. You spent materials and space to save a warm-up period you will never notice again.

Assuming an uneven split is broken. Check whether total demand exceeds supply first. If it does, the splitter is behaving correctly and the problem is upstream.

Forgetting Overflow exists. A single Smart Splitter set to Overflow into a container prevents a whole class of stalls.