Why use Beacons?

[leitk]

Summary: If you use productivity modules, use speed beacons as well.

For a long time I didn't understand the real use of beacons.
I thought they were to hyper speed a mega base, and for bases that were designed for rocket a minute.

In my case, I'm not often interested in rocket a minute nor mega bases. So I've mostly ignored speed modules and beacons.
Also, with the new nuclear plants, power is not a problem, so I ignore the efficiency modules.
But, the free goods from productivity modules is hard for me to ignore, so I apply prod 3 modules everywhere...
That slows down things, but the factorio answer to that is more, so build 2 factories with instead of 1.

The real reason for beacons is that they make building your factory cheaper, you need fewer assemblers and fewer modules if you use beacons.

There was a post on the math of productivity modules that gave me a clue viewtopic.php?f=134&t=5705. This is a very good post and one I refer to frequently.

That hint in DaveMcW's post kept bugging me, and eventually I took a look at speed modules in beacons with prod modules in assemblers. It's really more than a hint, but I didn't pay much attention to it the first time.
The numbers are amazingly good.

For example, a level 3 assembler using 4 prod 3 modules (PM3) runs at speed .5 and makes .1 high tech beaker a second (on average).
If you want 1 a second you need 10 of them.
If you do something like this:

Those 2 assemblers each run at speed 3, so they make .6 beakers a second, and to get 1 a second you need slightly less than 2.
The 2 assemblers use 8 PM3s, and the 4 beacons use 8 SM3s (speed module 3s), for a total of 16 modules 3's vs 40 needed for 10 assemblers in the case if 1 high tech beaker per second.
Since level 3 modules are expensive, it really helps to save those 24.

If you make this instead:

That 1 assembler runs at speed 5.5 so it makes 1.1 beakers a second.
That uses a total of 4 PM3s and 16 SM3s, which while it's more than the prior case, it does allow other types of factories to exist beside it and split the total cost.

That's with one assembler. When you add additional assemblers the costs go down a lot, because for each additional assembler you only have to add 1 or 2 beacons. They share the beacons from the existing assembler, so in the 4 beacon case the first assembler saves 24 modules, the second saves 30.
In the 8 beacon case the first saves 20, the second saves 32. This assumes you add the assemblers in a line each touching the next, that can require some creative designs.

If you don't use modules or use other than productivity modules, then beacons are not as effective because the counter of slow production to speed makes the whole setup much more effective.

For power and pollution, if you compare 4 PM3s with and without beacons the 8 beacon setup uses about 1/2 the power, and about 1/5th the pollution (assuming clean electricity).

There are some things to be aware of. If your factory is not running full time, assemblers will decrease their power usage, but beacons will not, so an idle factory with beacons will still use a lot of power.
The other thing to be careful about is the feeding of fast assemblers. especially for 1/2 second recipes a 5.5 speed factor means it finishes in less than .1 seconds, so keeping them fed 100% is impossible since the assembler will trigger a request when it gets below 2x ingredients, and it takes longer to load them than 1/3 a second. A stack inserter (or 2) will keep that fairly infrequent however. 1 second recipes can be kept full if you can keep enough material at the inserters, but that may require stack inserters or multiple inserters. Emptying shouldn't be a problem for most recipes, although if you have a wire (1/2 second 2 wires per recipe) making 30.8 (4/sec default, time 5.5 plus 40%) wires a second, you'll need 2 stack inserters going to a chests.

To a lesser degree this applies to labs and furnaces as well. I do recommend it for labs because it does reduce module needs and research is expensive (see DaveMcW's post). I don't recommend it for furnaces, they only take 2 modules and they're so slow that productivity isn't generally worth it, but if you do put prod mods in your furnaces you should use beacons as well, as that lowers the total number of modules required.

[Frightning]

It's really about economy of scale, the 8 Beacons to 1 Assembly machine style layout can be tiled so that the long term trend is a 1-to-1 ratio of Assembly machines and Beacons, despite having each machine affected by 8 Beacons, which means that you can get extreme throughput with maximum productivity bonus at a relatively low fixed cost (it's expensive in absolute terms, but you can get a ton of throughput this way, so that the cost per unit of productive capacity is greatly reduced; obviously, if you don't care for productivity bonuses, then it's far cheaper to just make more assembly machines).

[Engimage]

This kind of a setup scales pretty well if you try to tile it BOTH ways. With long rows it comes as close as 1 beacon per assembler on larger installations.
Take a peak at my set of designs for beaconed processing
viewtopic.php?f=202&t=49452
You can see that the most optimal structure for beaconed setups is a set of long rows that do share a line of beacons with each other making adding new rows as simple as adding one new row of beacons with machines inside.

And one thing to notice is that machines should have an offset relative to beacons by 1 tile so that 4 of beacons have an effect on the machine. If you line up machines with beacons only 3 of them will cover machines.
Another aspect of creating a beaconed setup is the UPS hit. The more machines you have the more is CPU usage for your factory. So when reaching large scale megabase you absolutely need to think about it.

[leitk]

Frightning and PacifyerGrey, you are entirely correct and those designs are impressive.

But that's exactly why I never looked at beacons before.
A big factory needs beacons, but what I didn't realize was that a small factory can greatly benefit from them as well.

If you have as few as 4 assemblers with productivity module 3s making the same thing and can put them in the same place, 4 beacons and 1 assembler can do the same job for 12 modules, saving 4. And when it comes time to expand, the savings are even bigger.

4 is the point where beacons are cheaper than more assemblers. 4 beacons and one assembler does the work of 6 assemblers without beacons, so if it is replacing 4 assemblers, it will be idle some time.

[Ringkeeper]

funny, the same discussion "why beacon with speed and assembler with prod if you can use efficency modules" i saw on reddit? last week.

I'm building atm my first mega base (1000 science per minute) and i'm in transition to it. So i simulated the needs with beacon+prod , effi and without.

With beacon i need 3,5GW. Just for the science production, rest of base not counted. With effi3 + 1 speed3 you need 400mw less but 5x the amount of factories. With effi2 it gets worse GW wise then beacon+prod. And without any module you need 2 GW more xD

So, yes, beacon+prod help in any base.

[uukgoblin]

Also, with the new nuclear plants, power is not a problem, so I ignore the efficiency modules.

I thought about it a bit: viewtopic.php?f=16&t=59989. It turns out, if power or pollution IS a problem, then Efficiency modules are sometimes actually worse than Speed ones. And I believe this is ALWAYS true if you put them in Beacons. I haven't yet found a case where putting Efficiency modules in a Beacon would make sense (especially when you use Productivity modules in the main building).

The surprising fact, for me, was that adding Beacons with Speed modules actually makes buildings more power/pollution efficient, even when you take the Beacons' power usage into account! (although you must be using Prod modules in that building)

[HurkWurk]

the only place i have ever used eff modules is around my rocket silos. most of the beacons were speed, with a few eff to cut the massive power use.

[m44v]

The way eff modules work with beacons, a 2xeff3 beacon needs to touch 5 assemblers or 6 furnaces to break even, which is hard to do except in the most dense layouts. And that's for break even, so the energy savings are actually very small in comparison with the cost of the modules.

So if you're going to use beacons speed modules are pretty much the only choice.

[quyxkh]

if you're going to use beacons speed modules are pretty much the only choice.

Refineries used to have only two module slots, I'd put eight of them around a beacon and eff everything up, that worked great and making everything fit was a fun puzzle. Now putting 3×E1 in every refinery early works great, I miss the puzzle-solving reward there. I'd better quit typing before I yell GET OFF MY LAWN! oh, darn.

[bobucles]

If beacons could affect the energy consumption of beacons, efficiency modules would suddenly become very important. A speed3 beacon sitting next to two partner speed3 beacons would burn 1.4MW, so beacon lines would become nearly 3 times as expensive and square grid 12x1 setups even more so. An Eff3 beacon would cut beacon energy consumption by nearly 200kW each, as well as reducing assembler3's by 84kW a pop.

Speed beacons get nerfed, and efficiency beacons get buffed. Is it enough of a balance swing to make efficiency beacons worthwhile? Probably not. The game balance is seriously tilted THAT far in favor of speed beacons. But it might create some new exciting builds such as using eff beacons to subdue the multiplying energy demand.

[Aeternus]

Problem with a heavily beaconed setup like listed above is that there's little to no room for logistics. Often resulting in logistics bots handling local goods deliveries. Yes, the throughput is insane on a prod-modded 8-speed-beaconed assembler. Yes, you get a lot of manufacturing compressed into a small footprint. The flipside is that belts are a no-go, they can't carry the goods fast enough (unless you have a large input -> small output deal like LDS or RFuel) which defaults to bothive.
Me, I prefer a hybrid strategy. Full production mods and 2 to 3 beacons per production structure in basically a column allows for full production bonus while maintaining an above-normal speed (factor 1.75 to 2.25 for assemblers). This still leaves room for belts. To discourage that kind of setup, limiting the amount of beacons that affect a single structure could help.

Efficiency beacons are pointless. I tried using hem. Because they work off of the base energy consumption of the device they are boosting, their effect is negligable once production mods come into play. Often their own energy consumption exceeds the savings made by boosted structures. The only use for beacons I've found is speed.

[Hedning1390]

Problem with a heavily beaconed setup like listed above is that there's little to no room for logistics. Often resulting in logistics bots handling local goods deliveries. Yes, the throughput is insane on a prod-modded 8-speed-beaconed assembler. Yes, you get a lot of manufacturing compressed into a small footprint. The flipside is that belts are a no-go, they can't carry the goods fast enough (unless you have a large input -> small output deal like LDS or RFuel) which defaults to bothive.

I wouldn't call belts a no-go. Most products don't require that much throughput. For things with extreme demand like electric mining drills they can't have productivity modules, so you don't have to have them between beacons. Worst perhaps are green circuits which is in great demand so you need compact blueprints, but also a case where a single factory saturates a belt lane (and wire would saturate a complete belt). However there are always ways around this. For example in the case of green circuits they mostly go to blue circuits, so why not create a mixed line turning ore directly into blue? Like this: https://i.imgur.com/XF9zfM3.png

Yes, some sacrifices have to be made, but no-go is too strong a word.

Efficiency modules are not completely pointless. In my mall I have several buildings with just a few speed modules in them, or no modules at all. The free space can be used for efficiency modules. Also to slow down things like mining drills which can only use speed modules (for better ratios) you can put in efficiency modules. Finally for mining drills. YOu may not care about the tiny +30% productivity when you have a lot of research already, nor care to make the patch go away faster with speed modules, or even want to make the huge investment in t3 modules. In that case t1 eff modules are a great alternative. Neither of these 3 examples are examples where you'd put them in beacons, however you can save some power by efficiency beaconing the rocket silo.

[Aeternus]

Problem with a heavily beaconed setup like listed above is that there's little to no room for logistics. Often resulting in logistics bots handling local goods deliveries. Yes, the throughput is insane on a prod-modded 8-speed-beaconed assembler. Yes, you get a lot of manufacturing compressed into a small footprint. The flipside is that belts are a no-go, they can't carry the goods fast enough (unless you have a large input -> small output deal like LDS or RFuel) which defaults to bothive.

I wouldn't call belts a no-go. Most products don't require that much throughput. For things with extreme demand like electric mining drills they can't have productivity modules, so you don't have to have them between beacons. Worst perhaps are green circuits which is in great demand so you need compact blueprints, but also a case where a single factory saturates a belt lane. However there are always ways around this. For example in the case of green circuits they mostly go to blue circuits, so why not create a mixed line turning ore directly into blue? Like this: https://i.imgur.com/XF9zfM3.png

Because it is still more sensible to compress it into green. Green with full production mods converts 1.5/(1.4^2) copper and 1/1.4 metal into 1 green circuit. You transport less materials down belts if you turn them into green circuits earlier.
For rail transit it's a definate nobrainer - green circuits stack to 200, copper/iron to 100. So moving greens to the blue circuit makers means less then half the trains on the grid. It may -seem- smart to co-locate green circuit production at blue circuit production, but in raw logistics you're at a disadvantage for doing so.

Green circuits are for me the biggest headache I'm trying to crack in my budding megaplant v2 (v1 failed miserably - fully centralizing smelting was a disaster due to throughput bottlenecks). I want to do them by belt and co-locate the greens plant between my copper and iron smelter so I can directly tap the needed resources and eliminate the biggest transport hog alltogether. Rough calculation shows I need 24 compressed belts out, and 15 belts of each resource in. I consider them to be part of the "base resources", on-par with the smelted iron and copper. Considering how much of them are consumed, that seems reasonable.

[Zavian]

For rail transit it's a definate nobrainer - green circuits stack to 200, copper/iron to 100. So moving greens to the blue circuit makers means less then half the trains on the grid

I think you are missing a possible strategy. For a belt based main bus setup, you are absolutely right that you want green circuits on the bus, rather than making green circuits on site, and having extra iron and copper plates on the bus. But for a train based system your blue circuit factory could receive coal, petroleum, plus iron and copper ore. It could make plastic, sulphuric acid, red and green circuits on site. (It could also have its own refining as well, but personally I'd rather centralise refining and ship petrol, lube and rocket fuel). That is 4 stations of good inward, and 1 station of good outward. 5 sets of trains (admittedly most of those routes will need multiple trains, and probably multiple source outposts). Compare that to a setup with one or more large iron smelters, large copper smelters, one or more large green circuit outposts, a red circuit outpost, a refinery producing plastic (and other things).

Now count those routes:
iron ore to iron smelter;
copper ore to copper smelter;
coal to refinery (or where you make plastic);
iron to refinery (or where you make acid);
iron to green circuit factory;
copper to green circuit factory;
green circuits to red circuit factory;
copper to red circuit factory;
plastic to red circuit factory;
green circuits to blue circuit factory;
red circuits to blue circuit factory;
acid to blue circuit factory;

I make that 12 different routes, (13 if you count the output blue circuits like I did above). Each of those route needs trains, and loading and unloading stations. (Think how many inserters and balancers are in a belt based train station. Loading and unload in a belt based factory is something you want to minimise if you need to save UPS). You mention that green circuits stack to 200, and you claim you save trains when moving green circuits over moving the equivalent amount of iron and copper, but that ignores the fact that you still have to have trains moving the iron and copper ore to the smelters, then from the smelters to the green circuit factory. Overall shipping green circuits actually requires more trains than just shipping the ore (or the plates) and making green circuits on site. Depending on how you divide your production outposts, you probably still want a green circuit factory, but you can help keep it a manageable size by making green circuits on site for products where it makes sense. (If you make red and green science in one factory, then you are already shipping in iron and copper, so you green science might as well make green circuits on site; red circuits could either import green circuits or import iron. if blue science is importing red circuits, then it probably also imports green circuits for mining drills (If it makes red circuits on site, then it might as well make green circuits as well); production science would also import red and green circuits, or make them on site).

Here is a prototype belt based blue circuit factory that I built a while ago that imports iron and copper plates, coal, petrol and acid, and makes about 1250 blue circuits/min. Just making green + red circuits at your blue circuit factory will cut your green circuit factory from 20 belts to about 8 for a 1k science/min factory.

BlueCircs2.jpg (927.54 KiB) Viewed 30576 times

[Hedning1390]

Because it is still more sensible to compress it into green.

I often transport ore to a smelting place. Taking the trains to the blue circuits instead is no loss. Ore is transported in both cases. If you smelt at your mine then the same logic applies, but you ship in the plates instead, like in Zavian's design. Regardless; nothing differs between belts and bots here. You are talking about compressing ore to greens? Well they are even more compressed as blue. We are eliminating transport steps, not adding. In my case I am giving some ore trains a different destination in order to eliminate about 5/9 of green circuit trains, 1/5 of iron plate trains and 2/5 of copper plate trains.

[uukgoblin]

you can save some power by efficiency beaconing the rocket silo.

Surprising, but true! I've done some analysis:

Case 1: You're going for efficiency only, don't care about resource usage (really?): the most power-efficient setup is 3x eff3s, 1x spd3 in the rocket silo and no beacons

Case 2: You're putting 4x spd3 in the rocket silo for some reason (not sure why you'd do that): the most power-efficient setup is 7 beacons filled with eff3s

Case 3 (most reasonable IMO): You're putting 4x prod3 in the rocket silo for massive resource savings: the most power-efficient setup is 20 beacons, 15 of which should be filled with eff3, and 5 with spd3.

I'm assuming you can't use more than 20 beacons around a rocket silo (please let me know if that's false!), and that the beacons won't be used for anything else. Also, you should be switching off your beacons when the rocket silo is idle. EDIT: I also haven't tested using lower-tier modules, or putting mixed modules in a beacon.

I'm thinking of making a post to summarize my findings (for other buildings too), and with pretty 3d graphs... See how it goes.

[Aeternus]

For rail transit it's a definate nobrainer - green circuits stack to 200, copper/iron to 100. So moving greens to the blue circuit makers means less then half the trains on the grid

I think you are missing a possible strategy. For a belt based main bus setup, you are absolutely right that you want green circuits on the bus, rather than making green circuits on site, and having extra iron and copper plates on the bus. But for a train based system your blue circuit factory could receive coal, petroleum, plus iron and copper ore. It could make plastic, sulphuric acid, red and green circuits on site. (It could also have its own refining as well, but personally I'd rather centralise refining and ship petrol, lube and rocket fuel). That is 4 stations of good inward, and 1 station of good outward. 5 sets of trains (admittedly most of those routes will need multiple trains, and probably multiple source outposts). Compare that to a setup with one or more large iron smelters, large copper smelters, one or more large green circuit outposts, a red circuit outpost, a refinery producing plastic (and other things).

Now count those routes:
iron ore to iron smelter;
copper ore to copper smelter;
coal to refinery (or where you make plastic);
iron to refinery (or where you make acid);
iron to green circuit factory;
copper to green circuit factory;
green circuits to red circuit factory;
copper to red circuit factory;
plastic to red circuit factory;
green circuits to blue circuit factory;
red circuits to blue circuit factory;
acid to blue circuit factory;

Sensible, but I'll make the comparison. I prefer to keep all things petrol-y co-located at the refinery so I can do away with tanker trains for all things except minor lube/acid out and big oil in. The rest is just pumps-and-pipes, which granted, looks messy but that's the usual look of a refinery anyway. Plastic production I do within piping range of the refinery. It's essentially 1 train of coal for 1 train of plastic, with Petrol added, so why do the petrol trains? My refinery thus has:

Oil in (raw good, high volume)
Coal in (raw good, high volume)
Iron In (low volume)
Copper in (low volume)
Rocketfuel out (high volume)
Plastic out (high volume)
Lube out (low volume)
Acid out (low volume)
Batteries out (low volume)

Next to the refinery I have an iron ore smelter cluster, followed by the Electronic Circuit cluste, then the Copper smelter on the other side. The EC cluster is beltfed directly from the Iron and Copper smelters, so no extra traffic on the rail grid to keep that running. This copper smelter feeds the EC cluster almost exclusively, it only feeds the refinery too with a small local train to the batteries facility. The iron smelter feeds all other production too.
Iron Ore in (high volume)
Copper Ore in (high volume)
Iron Out (high volume
EC out (high volume)
Copper out (low volume, local train to smelter)

Routes for this setup are:
Oil to refinery
Coal to refinery
Copper ore to smelter
Ironore to smelter
Copper to redcirplant
Plastic to redcirplant
EC to redcirplant
Redcir to bluecir plant
EC to bluecir plant
Acid to bluecirplant

10 routes, not counting output. And it reduces the types of goods arriving at your red and blue circuit production facilities, reducing their complexity and eliminating one of those unloading stations you mention. My reason for not doing the metal and copper to a remote EC plant is simple: The high volume of EC needed makes local production very tempting. Tie the smelters directly to the EC plant and you eliminate a -very- frequent long copper and iron train from the grid, reducing bottlenecks.

You are talking about compressing ore to greens? Well they are even more compressed as blue. We are eliminating transport steps, not adding.

Your philosophy about design is obviously different. I don't want ore trains in my core factory rail net. They are high volume, irregular and very long. I design my factory so the ore is delivered to the smelter at the edge of the factory, almost isolating the ore loop from the rest of the factory and eliminating that rail traffic entirely. I also try to reduce the traffic in-factory by placing production centers close to where their goods will be needed, reducing route length. For instance, processing units are needed for RCS and yellow science. So the ProcUn plant (along with the speedmod plant) are located near these. And sure, I could put the ProcUn plant dead smack at the refinery and smelters, but then you'd just end up with a bus-driven factory. Besides, you still need EC for tons of other places - Electric Mines, Green science, Electric Engines... to name a few. So you need to have EC trains anyway. And at a stack size of 200, they've got max compression stackwise, so minimum trains on the (shared) tracks.
As for shipping ore directly... One EC train requires roughly 4 ore trains of equal length.

--------------

I make that 12 different routes, (13 if you count the output blue circuits like I did above). Each of those route needs trains, and loading and unloading stations. (Think how many inserters and balancers are in a belt based train station. Loading and unload in a belt based factory is something you want to minimise if you need to save UPS). You mention that green circuits stack to 200, and you claim you save trains when moving green circuits over moving the equivalent amount of iron and copper, but that ignores the fact that you still have to have trains moving the iron and copper ore to the smelters, then from the smelters to the green circuit factory.

Yea. On virtually separate rail grids. Ore on the outside. Goods on the inside. And as stated above, by colocating the bulk EC production at the bulk smelters, you can beltfeed from those directly, eliminating the iron and copper trains. Ore in, EC out.

Overall shipping green circuits actually requires more trains than just shipping the ore (or the plates) and making green circuits on site.

You're right in absolute terms. Oretrain + EC train > Oretrains. But in logistical terms you may want to keep the ore trains outside of busy parts of your railgrid to prevent gridlock and junction bottlenecks. And do you want to build separate smelters at every location where you need heavy EC production? Smelters at the red circuit, blue circuit, green science, speed mod plant? Could get a little messy I think.

Depending on how you divide your production outposts, you probably still want a green circuit factory, but you can help keep it a manageable size by making green circuits on site for products where it makes sense. (If you make red and green science in one factory, then you are already shipping in iron and copper, so you green science might as well make green circuits on site; red circuits could either import green circuits or import iron. if blue science is importing red circuits, then it probably also imports green circuits for mining drills (If it makes red circuits on site, then it might as well make green circuits as well); production science would also import red and green circuits, or make them on site).

I'm keeping red and green sci production separate. Iron+Copper in for red, Iron+EC in for green. Blue does import red circuits, but I prefer to keep Electric Mine production near the iron smelter since it requires -gobs- of iron. Engine production as well (Engines and Electric Engines in one plant, output ships to blue/purple science respectively). Electric Furnace production is something I've not fully decided on. Brick "smelters" onsite or not? It needs to be close to the steel smelter since that's the main demand for it is highest. Which leaves LDS production that requires copper, plastic and gobs of steel.

Here is a prototype belt based blue circuit factory that I built a while ago that imports iron and copper plates, coal, petrol and acid, and makes about 1250 blue circuits/min. Just making green + red circuits at your blue circuit factory will cut your green circuit factory from 20 belts to about 8 for a 1k science/min factory.

BlueCircs2.jpg

I'm trying for a 10k sci/min with 8x 1250 in a square pattern. Keeps it managable per cluster, especially for refining and smelting... I've done something very similar to what you've got going in my initial busfed mall/early science factory - albeit without the heavy beaconing you do. It does work, but you're importing so many varied goods that your stations become large and ou risk spaghettification. At work so can't do screens now, but it looks very similar.

[Hedning1390]

You are talking about compressing ore to greens? Well they are even more compressed as blue. We are eliminating transport steps, not adding.

Your philosophy about design is obviously different. I don't want ore trains in my core factory rail net. They are high volume, irregular and very long. I design my factory so the ore is delivered to the smelter at the edge of the factory, almost isolating the ore loop from the rest of the factory and eliminating that rail traffic entirely. I also try to reduce the traffic in-factory by placing production centers close to where their goods will be needed, reducing route length. For instance, processing units are needed for RCS and yellow science.

There's nothing about the blueprint which excludes you from isolating the ore from the rest of the network. Ore comes in at one side, all other products enter or exit at the other side.

As for keeping things close this is as close as it can possible be for the high count items. Plate goes directly to greens and greens directly to blues. If blues have to be on the fringe of your base then so be it. A blue circuit demands 14.3 electronic circuits each, so for every blue circuit train you make you save that many green circuit trains making the delivery directly. In addition by having the smelting in the blueprint you save 3 plate trains per circuit trains. That's a total of 56.5 trains saved per blue circuit train. This easily makes it worth placing this on the fringe together with the rest of your smelters even though that means having to transport the blues a little longer.

By doing this your green facility will shrink by 5/9, making all logistics over there easier. The reason to have green going directly to blue is the same as why you make plate go directly to mining drills, (mining drills eat less iron than blues btw). I really don't see the difference.