Factorio Forums

There are many ways to improve CPU performance in your mining and smelting setups. These include:

• Mining drills output directly into furnaces.
• Furnaces output directly into cargo wagons.
• Very large, very rich resources.
• High levels of Mining productivity technology.
• Use as many beacons as possible.

But after doing all that, what is the most UPS efficient way to fill a cargo wagon? I ran performance tests on 5 different blueprints to find out.

Blueprint #1
3 furnaces, 6 beacons per furnace, long-handed inserters.
Furnace speed: 10
Inserter items per stack: 2.476
Note: Uses 33% less train stops than the other blueprints


Blueprint #2
2 furnaces, 8 beacons per furnace, long-handed inserters.
Furnace speed: 12
Inserter items per stack: 2.971
Note: The inserter runs at 99% capacity, which hurts its ability to empty a buffer when the train moves.


Blueprint #3
2 furnaces, 7 beacons per furnace, stack inserters.
Furnace speed: 11
Inserter items per stack: 1.362
Note: I don't expect this blueprint to win, but it is a useful comparison for the next two.


Blueprint #4
2 furnaces, 7 beacons per furnace, circuit controlled stack inserters.
Furnace speed: 11
Inserter items per stack: 11.995
Note: The inserters are activated once every 229 ticks.


Blueprint #5
2 furnaces, 7 beacons per furnace, circuit controlled stack filter inserters.
Furnace speed: 11
Inserter items per stack: 11.995
Note: The inserters are activated once every 229 ticks.



Testing method

I built a test scenario with the goal of producing 2640 copper plates per second. Output is to chests instead of trains, to avoid train performace affecting results. Every 10000 ticks a script empties the chests.

Test map downloads: mining-performance-1.zip mining-performance-2.zip mining-performance-3.zip mining-performance-4.zip mining-performance-5.zip

I benchmarked each map using the command: Results

Blueprint #1
Performed 100000 updates in 44856.280 ms
Performed 100000 updates in 45480.041 ms
Performed 100000 updates in 44954.518 ms
Performed 100000 updates in 45176.745 ms

Blueprint #2
Performed 100000 updates in 42370.623 ms
Performed 100000 updates in 43612.927 ms
Performed 100000 updates in 41403.191 ms
Performed 100000 updates in 42735.350 ms

Blueprint #3
Performed 100000 updates in 46345.014 ms
Performed 100000 updates in 46153.803 ms
Performed 100000 updates in 45539.896 ms
Performed 100000 updates in 46446.667 ms

Blueprint #4
Performed 100000 updates in 31330.315 ms
Performed 100000 updates in 30590.865 ms
Performed 100000 updates in 30464.489 ms
Performed 100000 updates in 30281.557 ms

Blueprint #5
Performed 100000 updates in 32769.336 ms
Performed 100000 updates in 32819.492 ms
Performed 100000 updates in 32577.326 ms
Performed 100000 updates in 32445.764 ms


Conclusion

Blueprint #4 is the clear winner! This blueprint includes my 229-tick timer, but you should reduce the cycle time a bit so it can empty the buffer when the train moves. Also you only need one timer per map.

Blueprint #3 is the worst, the inefficient inserters drag it down even though it has faster furnaces than Blueprint #1.

"Enable/disable" circuit condition is slightly faster than "Set filters".

Why not use prod3 modules in such beacon heavy setup?

Because ore is free.

(ore is not free unless you have infinite ores without liquid requirement)

Prod modules give you ore for free and prod modules in the furnace give you free plates. Are you sure you don't get plates faster with some combination of prod modules?

Some quick notes.

All but #1 will lose some beacon range on the miners when the design is actually expanded to work with trains (as they can't fit a 2 block gap that is misaligned to the rail 2x2 grid).
By not using trains you miss measuring the performance of a more compact design requiring fewer trains, for example.
This design will be a pain in the real world when the miners deplete.
Chests on miner 4 will fill after 1400 seconds and be idle the remainder 266 sec Well acutally chests are cleared, how did you do that, I wonder

I've created a variant of #1 that performs quite a bit better if you want to run it in your machine. It works by bridging stack inserters across the 3 tile gap rather than using red inserters. Additionally the output from the furnaces are clocked (250 ticks). 1 red inserter remains per block for power reasons.
Picture:
My machine results: (I use an automated script that slightly overestimates what the actual benchmark reports) https://github.com/mulark/factorio_benchmark_scripts
Edit: using 1000 seconds
mining-performance-1.1.zip,mining-performance-1.zip,mining-performance-4.zip
0.3933,0.4134,0.3028
0.3909,0.4078,0.2925
0.3839,0.4208,0.2929
0.3777,0.4031,0.2953
0.3854,0.4187,0.2902

Coopl that you figured that out, the only issue I see (which, to be honest, is because I'm lazy) is that you have to move the miners because they can not mine the tiles under the middle of the furnaces and beacons.

I ran 3 more tests: mining-performance-6.zip mining-performance-7.zip mining-performance-8.zip

Blueprint #6
Blueprint #1 with circuit controlled stack inserters.
Furnace speed: 10
Inserter items per stack: 12
Note: 252 tick timer.


Blueprint #7
Blueprint #2 with circuit controlled stack inserters.
Furnace speed: 12
Inserter items per stack: 12
Note: 210 tick timer.


Blueprint #8
Blueprint #4 with productivity modules.
Furnace speed: 8.4 * 1.2 = 10.08
Inserter items per stack: 12
Note: 250 tick timer. The test map only produces 2638 plates per second.



Blueprint #6
Performed 100000 updates in 35852.388 ms
Performed 100000 updates in 36206.522 ms
Performed 100000 updates in 35993.606 ms
Performed 100000 updates in 35503.578 ms

Blueprint #7
Performed 100000 updates in 32939.640 ms
Performed 100000 updates in 32901.497 ms
Performed 100000 updates in 33449.394 ms
Performed 100000 updates in 32727.000 ms

Blueprint #8
Performed 100000 updates in 30685.536 ms
Performed 100000 updates in 30866.923 ms
Performed 100000 updates in 30725.498 ms
Performed 100000 updates in 30118.889 ms

It looks like I was wrong about productivity modules. They are basically tied with speed modules when you have 7+ beacons.

DaveMcW wrote:It looks like I was wrong about productivity modules. They are basically tied with speed modules when you have 7+ beacons.

They become increasingly more effective with the number of both speed beacons and module slots. So on a miner with 3 modules they are definitely more effective than speed beacons. However it seams your build is more backed up on smelters.

Wow, you UPS guys must all play on ridiculously high ore settings. I'd never consider a design that had to be re-positioned several times to eat up the whole ore patch. Maintaining enough mines is enough work already!

PacifyerGrey wrote:

DaveMcW wrote:It looks like I was wrong about productivity modules. They are basically tied with speed modules when you have 7+ beacons.

They become increasingly more effective with the number of both speed beacons and module slots. So on a miner with 3 modules they are definitely more effective than speed beacons. However it seams your build is more backed up on smelters.

Unfortunately, productivity modules only stack additively (if you had +10000% mining productivity, modules only bring you to +10030%). Thus, speed modules are much better in mining drills at medium to high levels of mining productivity.

Selvek wrote:Wow, you UPS guys must all play on ridiculously high ore settings. I'd never consider a design that had to be re-positioned several times to eat up the whole ore patch. Maintaining enough mines is enough work already!

I doubt very many people actually use this in-game, this is more of a theoretical test.