Productivity modules: where do you draw the line?

[AutoMcD]

This is a chart I made using Foreman in preparation for an upcoming optimized base build:

My goal for this is 20 science of each type per minute, which should support 10 labs. In addition it will make upgrade modules at a modest rate for late game toys without crippling the science production.

In general my feeling about productivity modules are to start at the top and work down. So putting them in the labs has the most effect since an extra of each science saves a ton of resources. Next, blue science because that is effectively the same as producing more red circuits, smart inserters, batteries, and steel. Like, why would I use modules in just the inserters or the red circuits when most of them go to blue science where a module there gets better coverage?

T3 modules are expensive, so placing them I would want the highest bang for the buck. labs, I will have 10 of them. That's too many T3 modules to make, and not everything is for science. Red circuits, there's 6 assemblers. But green circuits? Only 3 assemblers and it's gobbling up 85% of the copper and 50% of the iron! I feel that the single biggest point of improvement is there. From there it gets more distributed, 1/4 of the circuits go to processors, 1/4 to red circuits, and most of the remaining to science. And 3/5 of red circuits also to science so safe to say the majority of everything still covered under the lab+blue science umbrella.

Iron is much more distributed, conserving this via production modules would take a lot. But the single biggest point of use again is green circuits, and if you trace where it's used it seems the same strategy applies.

While this would reduce resources needed to hit my goals dramatically, copper and iron are abundant. What I usually have trouble with is OIL. And most of it goes to plastic.. 64% of it anyway. The rest sulfur. These are only 2 buildings so easy to hit with modules. But naturally depending on how bad the oil situation is could potentially hit every building related to oil consumption. And again, nearly all of this goes to red circuits and batteries, which mostly goes to blue science.

[Frightning]

Since the resources on the map are technically finite, but massive (except Oil, which has infinite yield, but finite rate). If you want to produce as much stuff as you can, you should always use Productivity Modules where ever you can, because it effectively increases the yield of the resource deposits on the map. Note that it will be important to get off steam power and onto solar+accumulators to remove the Coal/Solid Fuel/Wood-based costs of producing things due to energy consumption. For things that can't take Productivity Modules, the best choice is Efficiency Modules to lower energy costs, and hence reduce the fixed costs of the achieved production rate. Why not Speed Modules? Because they are expensive (especially higher tier modules), and increase energy cost per unit. It is cheaper to simply build more Assembling Machines or Chemical Plants, etc...than it is to use Speed Modules. The only place where Speed Modules make sense is in Pumpjacks, where they can increase production rate, thereby increasing the (long term) Oil production rate of the Oil Fields on the map.

[taiiat]

i will not use Productivity Modules unless i have Beacons.
otherwise it's a mixture Efficiency and Speed.

luckily that's fine since by the time the Factory starts consuming inordinate amounts of everything, you definitely have Beacons available.

[BlakeMW]

It is cheaper to simply build more Assembling Machines or Chemical Plants, etc...than it is to use Speed Modules. The only place where Speed Modules make sense is in Pumpjacks, where they can increase production rate, thereby increasing the (long term) Oil production rate of the Oil Fields on the map.

Cost isn't everything in a game where you can have so many resources, cost becomes decreasingly relevant as a factory becomes more developed. Speeds are good for reducing space used. While space is generally infinite we could say that real estate inside the factory is more valuable than that out in the middle of nowhere. Speed1 modules are a cheap way to help balance out ratios (for example 5:6, or 4:5 where a speed1 can turn it into 4.8:5). Heck, they're good enough that antielitz uses them in his speed runs to help balance out the ratios for blue circuits.

[silverkitty23]

Space is infinite, but if you involve logibos, spreading out can hurt performance.

[bigyihsuan]

I rarely use Productivity. I only ever use Speed. Efficiency is for biter-friendly chumps :p

Then again, I only play in Peaceful, so no need to worry about pollution and biter attacks.

[vanatteveldt]

Also I don't agree when you say that the effect of the productivity module is "multiplied". 1x Prod 3 saves you 10% ore, no matter how fast the machine runs. That part of the effect is not multiplied at all, it's always 10%. The amortization occurs faster, of course..

You addressed BlakeMW, but I'll bite We were having a similar discussion here (viewtopic.php?f=18&t=25218#p159799)

The thing is, productivty modules both increase output per second (by giving 10% extra production) and decrease it (by giving -15% crafting speed). But the essential thing is that the extra production is multiplicative with crafting speed, while the decreased crafting speed and speed module crafting speed bonuses are additive. What I mean with this is that output per 'cycle' (i.e. recipe time divided by base crafting speed) is (1+productivity bonus) * (1 + crafting speed bonus). In other words, every extra speed modules gives a fixed effect of extra output per cycle, but the productivity bonus is multiplied by the speed bonus, so with 10% extra output every speed module has 10% more effect since the productivity bonus kicks in with every normal production round.

So, in an extreme example with 14 effect sources (ie 2 slots and 12 beacons as in the pump jack example), 14 speed modules give crafting speed 16, but 2 prod and 12 speed give crafting speed 13.4, multiplied by 1.2 = 16.08 total speed.

This is why setups of productivity modules in the plant with speed beacons around it are so useful. If you have 4 prod3 modules (for -60% crafting speed) you more than double output with a single speed beacon,upping crafting speed to -10%, outputting .9(*1.4) instead of .4*(1.4). Assuming that you can place a beacon to affect at least 2 plants (this is very conservative), the 2 speed3 modules increase output by a factor of 2.25, or in other words save more than 8 productivity3 modules to get the same output.

Finally, you really shouldn't underestimate the savings on a whole production chain. If you have 2xprod3 in your smelter you only save 1 - (1/1.2)= 17% ore. However, assuming you use green circuits in bulk and have 4xprod3 modules in both cable and green circuit assemblers, you suddenly save 1-1/(1.2*1.4) = 40% iron ore and 1-1/(1.2*1.4*1.4) =57% (!) copper ore. This goes up further for red circuits -> blue circuits -> rocket control units -> rocket parts.

For petroleum the savings are also pretty big. Assuming you turn all crude into PG, without productivity 10 crude yields (heavy/light/gas) 1/4.5/5.5 -> 0.75 + 4.5=5.25/5.5 -> 3.5+5.5=9 PG. With 20% productivity in each stage, you get 1.2/5.4/6.6 -> 1.08+5.4=6.48/6.6 -> 5.2+6.6=11.8 gas, or 31% more. Turning that into plastic you get another 20% productivity for 36% savings, turning it into batteries you get 1.2 (sulfur) * 1.2 (gas) * 1.2 (batteries) for a total productivity of 2.26 or 56% savings.

[siggboy]

You addressed BlakeMW, but I'll bite We were having a similar discussion here (viewtopic.php?f=18&t=25218#p159799)

You are 100% correct (and so was Blake; actually I had realized my mistake earlier but did never get around to correct it here).

I think what's most important is that productivity modules apply to machine cycles. If you make more cycles happen (= increase cycle frequency by a factor), you amplify the effect of the productivity module (by that factor). Everything else follows from there quite easily.

The interesting question remains if that makes Prod 3 in furnaces a good idea (if we include Speed modules from beacons as well). The main investment here, and actually the only one that matters in Factorio, is player time (and player enjoyment during that time). Setting up mining outposts (and maintaining them and laying the rails etc.) in particular is not the most enjoyable aspect of the game, at least not when you're doing it for the 30th time. So if the modules in the smelting area can cut that activity down significantly (in the long run), I'm all for it .

Regarding petroleum: I think that's a completely different ball game, because oil is a lot more rare than ore; so it makes sense to go after smaller savings in that area, and also the investment (for the modules) is not in oil but mostly in other stuff. That changes the amortization dynamics, too.

Finally, you really shouldn't underestimate the savings on a whole production chain. If you have 2xprod3 in your smelter you only save 1 - (1/1.2)= 17% ore. However, assuming you use green circuits in bulk and have 4xprod3 modules in both cable and green circuit assemblers, you suddenly save 1-1/(1.2*1.4) = 40% iron ore and 1-1/(1.2*1.4*1.4) =57% (!) copper ore. This goes up further for red circuits -> blue circuits -> rocket control units -> rocket parts.

Now this is misleading.

The savings in the Electric Furnace are independent of the savings "down the line". Sure, the effect is multiplicative here as well, but saving "Iron Ore" is different from saving "Iron Plates", even though one is made from the other.

In the furnace you save 17% (ore). In the assembler you save 29% (plates). If you remove the modules from the furnace you still save 29% plates in the assembler. Saving resources is always good. Multiplying the various savings through the chain together does not really convey useful information (of course it feels good to know you save this and this much, but it doesn't help you).

It's always better to put the modules late into the chain first, before you move to earlier steps. We know this, it's easily deduced, we do not need to know the total cumulative effect.

[Frightning]

Finally, you really shouldn't underestimate the savings on a whole production chain. If you have 2xprod3 in your smelter you only save 1 - (1/1.2)= 17% ore. However, assuming you use green circuits in bulk and have 4xprod3 modules in both cable and green circuit assemblers, you suddenly save 1-1/(1.2*1.4) = 40% iron ore and 1-1/(1.2*1.4*1.4) =57% (!) copper ore. This goes up further for red circuits -> blue circuits -> rocket control units -> rocket parts.

Now this is misleading.

It's not misleading at all, imo. Savings in terms of raw resources is the only useful way to measure the cumulative effect of have prod. modules throughout the process of making goods. It shows that the resource saving effect can become far more pronounced for high level products (things that have more steps from raw to ready), than basic products (and it's not hard to reason that it's best to start w/ prod. modules in the final product first and work backwards concentrating on what constitutes the bulk of the resources used first).

[BlakeMW]

It's always better to put the modules late into the chain first, before you move to earlier steps. We know this, it's easily deduced, we do not need to know the total cumulative effect.

Not quite true! It's actually best to place modules into the highest yielding recipes/assemblers in the chain, in accordance with the amazing Productivity Module Math post, but you can also take into consideration individual resources.

For example if you want to stretch copper (I always do) and most your copper is going to circuits, this is the copper ore usage of various intermediates (in best available machine):

Electronic Circuit: 3.75 copper/s
Blue Circuit: 3.33 copper/s
Copper Wire: 2.5 copper/s
Red Circuit: 0.78 copper/s
Furnace: 0.57 copper/s
Mining drill: 0.53 copper/s

So the first place you put prod modules is the relatively lowly electronic circuits, next the blue circuits, next the copper wire. The red circuit is a grossly inefficient place to use prod modules despite being higher in the chain. A prod module in electronic circuits is doing 4.8x as much work as one in red circuits - another way of looking at it is you're expending 4.8x as much investment and extra electricity to save 1 copper in red circuits, than saving 1 copper in electronic circuits.

[Frightning]

It's always better to put the modules late into the chain first, before you move to earlier steps. We know this, it's easily deduced, we do not need to know the total cumulative effect.

Not quite true! It's actually best to place modules into the highest yielding recipes/assemblers in the chain, in accordance with the amazing Productivity Module Math post, but you can also take into consideration individual resources.

For example if you want to stretch copper (I always do) and most your copper is going to circuits, this is the copper ore usage of various intermediates (in best available machine):

Electronic Circuit: 3.75 copper/s
Blue Circuit: 3.33 copper/s
Copper Wire: 2.5 copper/s
Red Circuit: 0.78 copper/s
Furnace: 0.57 copper/s
Mining drill: 0.53 copper/s

So the first place you put prod modules is the relatively lowly electronic circuits, next the blue circuits, next the copper wire. The red circuit is a grossly inefficient place to use prod modules despite being higher in the chain. A prod module in electronic circuits is doing 4.8x as much work as one in red circuits - another way of looking at it is you're expending 4.8x as much investment and extra electricity to save 1 copper in red circuits, than saving 1 copper in electronic circuits.

Good point, I was thinking purely in terms of how to produce the most of x using given resources. In which case, you can't really improve upon final item first. But when you make more than one thing with your factory (typical), then yes, it's where the greatest amount of the (raw) resources you want to stretch are being used. Technically, if you want to be as efficient as possible w/ order that you 'module up' as prod modules are made, the best places to start the individual production centers with highest raw resource throughput rates.

[taiiat]

For example if you want to stretch ore usage

So the first place you put prod modules is the relatively lowly electronic circuits, next the blue circuits, next the copper wire. The red circuit is a grossly inefficient place to use prod modules despite being higher in the chain. A prod module in electronic circuits is doing 4.8x as much work as one in red circuits - another way of looking at it is you're expending 4.8x as much investment and extra electricity to save 1 copper in red circuits, than saving 1 copper in electronic circuits.

i see it in yet another way - the things i'm producing of which i need to use for other crafting the most (i.e. the things i need to craft the most of) would be where i'd be interested in using Productivity Modules.
so, ideally the things that i need to use as crafting reagents as well as actually use, as those will generally have the highest demand. or atleast the things i use to craft a bazillion different things with because i need an arseload of those items, as my entire factory critically relies on the incredibly high volume production of those items to function.

[Frightning]

For example if you want to stretch ore usage

So the first place you put prod modules is the relatively lowly electronic circuits, next the blue circuits, next the copper wire. The red circuit is a grossly inefficient place to use prod modules despite being higher in the chain. A prod module in electronic circuits is doing 4.8x as much work as one in red circuits - another way of looking at it is you're expending 4.8x as much investment and extra electricity to save 1 copper in red circuits, than saving 1 copper in electronic circuits.

i see it in yet another way - the things i'm producing of which i need to use for other crafting the most (i.e. the things i need to craft the most of) would be where i'd be interested in using Productivity Modules.
so, ideally the things that i need to use as crafting reagents as well as actually use, as those will generally have the highest demand. or atleast the things i use to craft a bazillion different things with because i need an arseload of those items, as my entire factory critically relies on the incredibly high volume production of those items to function.

If you're trying to be efficient with resources spent on the expensive prod modules (e.g. level 3 especially), then the best place need not be in something you make a lot of because of how many assemblers you have making that item. It really does solely come to down to resources used/sec, nothing else matters.

[vanatteveldt]

Finally, you really shouldn't underestimate the savings on a whole production chain. If you have 2xprod3 in your smelter you only save 1 - (1/1.2)= 17% ore. However, assuming you use green circuits in bulk and have 4xprod3 modules in both cable and green circuit assemblers, you suddenly save 1-1/(1.2*1.4) = 40% iron ore and 1-1/(1.2*1.4*1.4) =57% (!) copper ore. This goes up further for red circuits -> blue circuits -> rocket control units -> rocket parts.

Now this is misleading.

The savings in the Electric Furnace are independent of the savings "down the line". Sure, the effect is multiplicative here as well, but saving "Iron Ore" is different from saving "Iron Plates", even though one is made from the other.
[...]
It's always better to put the modules late into the chain first, before you move to earlier steps. We know this, it's easily deduced, we do not need to know the total cumulative effect.

I disagree. The "big" question in this thread is whether productivity modules are worth it, and where you should use them. My simple answer, from a long-term solar-powered end-game megabase perspective is: yes, they're worth it, and you should use them everywhere. The main argument for this is that the overall effect is huge: you save a little here and there, but since it multiplies, the overall savings are huge. Sure, the circuit savings are independent of the smelting savings, but the bigger picture is that you want them both

Also thanks for linking to @DaveMcW's post. My thinking was to start from smelters and then move up the chain, but I guess I should have probably started with green circuits, even though you probably want to do green circuits and copper cable at the same time since they require a redesign (going from 3:2 to 1:1 ratio) which you might as well combine with adding beacons. It's also interesting to see that if oil is scarce (which it often is), it's better to add productivity to plastic and acid production before adding it to oil processing.

[taiiat]

If you're trying to be efficient with resources spent on the expensive prod modules (e.g. level 3 especially), then the best place need not be in something you make a lot of because of how many assemblers you have making that item. It really does solely come to down to resources used/sec, nothing else matters.

ah, but if those machines can pump out stuff faster, i don't need to have dozens and dozens more of those Assemblers making those materials.

there's many places where the extra output is nice once the factory is large enough.
that's just where i'd want the extra output the most, first.

[Frightning]

If you're trying to be efficient with resources spent on the expensive prod modules (e.g. level 3 especially), then the best place need not be in something you make a lot of because of how many assemblers you have making that item. It really does solely come to down to resources used/sec, nothing else matters.

ah, but if those machines can pump out stuff faster, i don't need to have dozens and dozens more of those Assemblers making those materials.

there's many places where the extra output is nice once the factory is large enough.
that's just where i'd want the extra output the most, first.

I think you're confusing how speed modules work with how production modules work. The units/sec produced is actually lowered by productivity modules, but the yield (per resource used) is increased (so is energy cost as well). You will need more assemblers with prod modules than without to achieve a desired units/sec production rate of the given object, but corresponding resource drain will be notably lower.

[siggboy]

Finally, you really shouldn't underestimate the savings on a whole production chain.

Now this is misleading.
It's always better to put the modules late into the chain first, before you move to earlier steps. We know this, it's easily deduced, we do not need to know the total cumulative effect.

I disagree. The "big" question in this thread is whether productivity modules are worth it, and where you should use them. My simple answer, from a long-term solar-powered end-game megabase perspective is: yes, they're worth it, and you should use them everywhere.

That's of course, true (not taking pollution into account). If you have the prod modules, it's best to use them everywhere, because they'll amortize eventually and from then on they give you free resources.

The main argument for this is that the overall effect is huge: you save a little here and there, but since it multiplies, the overall savings are huge. Sure, the circuit savings are independent of the smelting savings, but the bigger picture is that you want them both

Yes, but I have to try and explain again what I meant by "misleading".

We were discussing the benefit of combining +speed with +productivity. The +speed makes +productivity objectively stronger because of the reasons discussed. How much "stronger" can be figured out by multiplying the effects (it's what you pointed out).

So far, so good.

You then, however, made the mental leap to combining several +productivity effects in the same production chain, again giving an example by multiplying the effects together -- but this is not the same kind of multiplicative effect as combining +speed with +prod in the same machine.

If you have +prod in, say, your gear wheel factory, then putting some additional +prod into your smelters does NOT make the gear wheel factory in any way "better". All it does is make you produce some iron plates for free, and as long as you can use those that's fine and dandy but does not affect your existing IGW production at all.

One cumulative effect (+speed with +prod in the same machine) is qualitatively different from the other cumulative effect (+prod through a production chain) -- but you kind of put them both in the same basket, at least it was my impression, and that's what I meant when I said "misleading".

Now, Blake again was spot on when he mentions that it's best to put the +prod into the machines with the highest throughput in raw materials. That's the most effective use if you look at the entire factory.

When I said it's best to put the modules at the end of the chain first, I was concerned with only a single product, from source to sink. For example, if you look at the complete chain with perfect ratios from raw materials to rocket part, then a +10% productivity in the rocket silo will stretch everything in that chain by 10% (including, e.g. the electronic circuits). Now why would I stretch only electronic circuits by 10% when I could instead stretch everything by that amount if I put the module at the end of the chain?

Of course this line of reasoning is quite moot and not really useful, as I now realize -- since you usually do not make an isolated production chain with perfect ratios for something and then get the modules one-by-one and have to decide in which order you use them. In an actual factory you have production of "stuff" that gets used in various ways, and if your goal is to be resource efficient, then the +prod needs to go where the resource throughput is highest.

Also thanks for linking to @DaveMcW's post. My thinking was to start from smelters and then move up the chain, but I guess I should have probably started with green circuits, even though you probably want to do green circuits and copper cable at the same time since they require a redesign (going from 3:2 to 1:1 ratio) which you might as well combine with adding beacons. It's also interesting to see that if oil is scarce (which it often is), it's better to add productivity to plastic and acid production before adding it to oil processing.

Yes, that post is useful indeed.

[vanatteveldt]

I think you're confusing how speed modules work with how production modules work. The units/sec produced is actually lowered by productivity modules, but the yield (per resource used) is increased (so is energy cost as well). You will need more assemblers with prod modules than without to achieve a desired units/sec production rate of the given object, but corresponding resource drain will be notably lower.[/quote]

This is of course correct, but I don't think anyone ever uses just productivity modules, because they are so much more effective together with speed modules (as argued above). Thus, it's not really about building more assembly plants, it's about boosting them with more speed beacons.

[BlakeMW]

If you have +prod in, say, your gear wheel factory, then putting some additional +prod into your smelters does NOT make the gear wheel factory in any way "better". All it does is make you produce some iron plates for free, and as long as you can use those that's fine and dandy but does not affect your existing IGW production at all.

I would say with the very important exception of if you are actually iron ore limited either by mines or logistics (for example you might be limited by 2 fully compressed basic belts of ore and be too lazy to upgrade the whole logistics chain to support 3 compressed belts). Usually I use prod3 modules because I'm limited by the ore, so putting prod modules into the furnaces actually does improve the throughput of the rest of the chain. Arguably it's an expensive way to overcome an ore bottleneck compared with expansion, but unless you're doing a speedrun it's a solid investment.

[siggboy]

Usually I use prod3 modules because I'm limited by the ore, so putting prod modules into the furnaces actually does improve the throughput of the rest of the chain. Arguably it's an expensive way to overcome an ore bottleneck compared with expansion, but unless you're doing a speedrun it's a solid investment.

The investment is fine, as long as you keep in mind that it takes ~15 HOURS to amortize with a perfect setup (i.e. including beacons and all); it's what Dave has determined in that other thread.

I think all this discussion was very interesting, and we've basically all agreed but talked at cross-purposes for a while. Maybe the biggest mistake we made was that we looked at resource savings, when in fact what we get are extra products; the savings are the inverse effect but only if you use up all product (or if you're actually starved of product, as per your reasoning above).