Testing science- and energyefficiency of chained science lab

[Distelzombie]

Hi there.

I did some testing regarding laboratory chains and other similar setups. Here are my findings:

Basically this answers the following questions: How long can/should you chain labs? Whats the use? How good is it compared to direct insertion?
- You should always build blocks of labs instead of single lines because they are way more efficient.
- Its really ressource efficient, fast to build manually, easy to extend, uses less "volume" (smaller, compared to setup with four belts and four inserters)
- Its a certain stated percentage as good as direct insertion. This values can be used to deduce efficiencys of other researches.

Note: All tests were made with all research done using Creative-mod. Except infinite research: I used mining-productivity research to test everything because it needs six different packs. Please be aware that these findings only apply to research with six different science packs with equal need. In different researches the numbers change according to need.
I deliberately chose this setup for the late game in mega bases where mining productivity is a key for maintenance.
All numbers are taken from the 10min graph of the power consumption dialog and averaged. I dont know a better way to measure use of labs. I used the consumption of the labs in a setup where I directly insert in each lab as a
Remember not to loop the labs. This can stop them from working.


I started chaining 400 labs with 3 stack-inserter each and found that the science-pack distribution follows a sawtooth-wave like pattern. Because at some point the last lab will run out of a science pack and the inserter will grab the entire inventory of this pack out of the adjacent laboratory. (I know what you think: What if you change stacksize? Nope, no difference.)
I found that about 50% are not working because the inserters constantly pull pack out of them. I call this efficiency from now on and put this into relation to labs that have direct insertion. ("speed" works fine too)

Here is a picture of the sawtooth-wave like behaiviour. Looks like its leveling over time:

Lowering the number of labs did nothing good until I reached about 20. Here the efficiency started reaching reasonable amounts: ~86% (error 5%)
Because the power consumption dialog swaps to MW, essentially lowering my measurement accuracy to about +-5%, I used 16 labs from now on to test things thoroughly which put me at 960kW max., lowering the error to something about 0.1 or 0.2%

- 16 labs in a row, chained by three stack-inserter (Stack size: 12) each: 91% efficiency, essentially making one and a half of them wasted ressources.
- Using fast-inserters the efficiency dropped to 88.6%.
Because one and a half of the labs were waste, I decreased the number to ten:
- With 10 labs in a row, chained by three stack-inserter (Stack size: 12) each: 94.5% efficiency, "wasting" only half a lab.
- Test using stack-inserters with lowering stack size (10 to 2) shows a decrease in efficiency to 94.15% instead of 94.5%
- Test using stack-inserters with lowering stack size (9 to 1) shows a decrease in efficiency to 94.3% instead of 94.5%

-------------------
I realized this is not a good way of doing it after a guy named "sEiNZu | phanYuuh" showed me his setup: A grid of 4x4. That would obviously be better than a simple row. (See picture below)

Grid of 3x3, 9 labs in total is 97.2% efficient. (Way better already)

Grid of 5x5, 25 labs in total: This is more difficult since the power consumption is over 1MW and I lose accuracy. With these numbers alone the setup would be 100% efficient: 1.4MW in the grid and 1.4MW with direct insertion of 25 labs. But that is not the case, therefore I used the graph instead - as a visual measurement.

Because the amount of inserters in this setup is extremely high, I tested three different versions: Three inserter each, two inserter each and one inserter each. I pasted the graphs in one picure for easy comparison:

It looks like all three are equal. One stack-inserter seems to be enough. Also a grid of 5x5 appears to be the best amount because the higher you go, the higher the loss. It looks like about 94% efficient - the graph from the 10-in-a-row setup didnt look much different. Thats good.

Conclusion: If you want to save one-time ressources, arrange the labs in a grid of 5x5. Paste it four times with the input in the middle and you're good to go.

I know it doesnt have military science there..

-------------------------
EDIT: I realized that a grid 5x5 is quite bad when researching a science like "follower robot count". (7 packs, 30s) The efficiency drops to about 90%, judging by the looks of the graph. :/ I would say a grid of 4x4 would be the best solution then for lategame.

-------------------------
ENERGY EFFICIENCY:
I was wondering if lab setups with beacons really are more energy efficient than setups without beacons as some people here stated. So I tested it.

I tested four different setups. Popular ones and some others.
I use a fixed amount of energy stored in accumulators. The accumulators have a capacity of 1750MJ. The amount of research done with this amount of Energy is shown in the picture.
Every lab has two productivity modules and every beacon two speed modules. I chose "follower robot count" because it uses all seven science packs and only needs 30s per pack. Its a good benchmark research.
The best setup uses only Beacons that reach at least 6 Labs.

I also tried measuring the science pack consumption using combinators. (The setup can be found as a blueprint on Page three or four) This is not accurate due to the productivity modules increasing the amount of research done, but being unmeasureable by counting.

[Nich]

I am really confused all this setup changes is the number of requester chests required. You still need just as many robots to transport science packs and there average distance traveled will be about the same since they are in the middle of the grouping.

[Lav]

Note that you're testing this on infinite science techs, which take tons of time, which decreases the frequency of labs running out of science packs and causing a hiccup.

When you run this on earlier game techs, the situation is so much worse.

Personally I've made a number of attempts at chained labs, and the only conclusion I made was that I'm going to use sushi belts from now on.

[Distelzombie]

But it still takes 60s per science pack regardless of if its infinite or not, doesnt it?

[iceman_1212]

it takes significantly less time if all lab speed upgrades are completed and if the setup is speed beaconed (with prod modules inside lab). ~20 seconds for a ~60 second research, by my casual observation.

[Xeorm]

But it still takes 60s per science pack regardless of if its infinite or not, doesnt it?

No. Techs have different times per pack. Most mid-tier techs are 30s. It's only late-game with infinite tech that you routinely get 60s research times.

[NorwegianBlue]

Some techs require multiple packs of some, but not others, per time interval.

That can really mess up a finely tuned system.

[Distelzombie]

Lav: I deliberately tested lategame science mechanics.
Iceman: I did all research upgrades.
Did you guys even read what I wrote? I repeat myself. Theres a "Note" section in the post.
But I probably wasnt very clear...

Some techs require multiple packs of some, but not others, per time interval.
That can really mess up a finely tuned system.

Yea, you cant have one build thats good for all without compromise. But infinite mining productivity is the most wanted research of all. And I can always make more tests since all the shit's still there.

I am really confused all this setup changes is the number of requester chests required. You still need just as many robots to transport science packs and there average distance traveled will be about the same since they are in the middle of the grouping.

It was quickly thrown together. There are probably better builds. All this is is ressource efficient. I should have noted that. I forgot it while copying from reddit where its no use to anyone.

Basically this answers the following questions: How long can/should you chain labs? Whats the use? How good is it compared to direct insertion?
- You should always build blocks of labs instead of single lines because they are way more efficient.
- Its really ressource efficient, fast to build manually, easy to extend, uses less "volume" (smaller, compared to setup with four belts and four inserters)
- Its a certain stated percentage as good as direct insertion. This values can be used to deduce efficiencys of other researches.

I updated the initial post. (But the "note" section was there all along)

[Lav]

Erm... With that many stack inserters, I'm pretty sure it's less resource-effective than a red sushi belt. Need to check...

UPD. Ok, I tested the sushi belt quickly. No hard numbers yet, but essentials are as follows:

1. 7 packs on a looped red belt work fine for as long as labs only pull from one side of the belt, otherwise pack density on the belt is not enough. On the other hand, for as long as the labs only pull from one side, it doesn't matter what type of belt it is - faster belts have faster recovery rate from zero state, but for as long as there's enough time to replenish the belt between cycles, it doesn't matter how fast it moves (up to a point - as soon as the belt it too long to make a full loop before a science cycle is ready, it will start losing effectiveness, so faster belts are still preferable). You can run a belt loop around the lab complex, or make an internal loop and put the labs on the perimeter, it doesn't matter.

2. It's a good idea to keep at least 3 belt tiles per lab - this ensures that a pre-filled belt will have enough science packs for all labs.

3. When inserting to the belt, I track two belt tiles - the one inserted to, and the one before that. In this scenario it's important to keep track what lane do you use to insert - it's preferable to insert via the lane that's "up the stream" - so as soon as a pack enters the sushi belt, the inserting belt stops until the pack leaves the tile. This allows for a better spacing when filling from zero and prevents small unicolor patches from forming.

4. Resource cost is 3-4 red belt tiles + fast inserter per lab = ~50 iron and ~5-6 copper per lab (including the cost of the loading bay).

5. Labs seem to be active nearly100% of the time - didn't have time for accurate measurements, but can do so later.

6. Start time from pre-filled belt is practically instant. Start time from empty belt feels quite slow.

7. Construction density when running an internal belt loop is ~15 tiles of space per lab - roughly the same as with presented chained lab complexes.

[Daid]

7. Construction density when running an internal belt loop is ~15 tiles of space per lab - roughly the same as with presented chained lab complexes.

But you can do more space efficient then proposed:

And I'm pretty sure someone will come up with a bigger more compact setup.

[Lav]

If I made any attempt to compress my design in the first place, this argument would have merit. Except I didn't, so the argument applies to both sides.

Plus your design won't work that well. There's a reason why OP's design has inserters converge on one corner lab instead of splitting into several sequences - doing otherwise will play havoc with multiple packs per lab.

[Aeternus]

7. Construction density when running an internal belt loop is ~15 tiles of space per lab - roughly the same as with presented chained lab complexes.

But you can do more space efficient then proposed:

And I'm pretty sure someone will come up with a bigger more compact setup.

If you'er going to do that, do it in a curl with a nice speed or production beacon in the middle. Me, I'll stick to a "+" grid with beacons at the corners and fed from requestor chests. Simple works, and that makes each research post affected by at least 4 beacons. I don't see what the benefit of daisy chaining them with inserters is to begin with (aside from that you can do it).

[iceman_1212]

Here's a belt + inserter hybrid design that doesn't rely on either requester chests or sushi belts. Beacon coverage is only slightly less efficient than that of a sushi-belt.

[leoch]

That looks very expensive to build and to run, iceman. Unless you're really short on space, using more labs should save quite a bit on power and construction costs I think — I haven't done the maths though.

[iceman_1212]

the idea behind the speed beacons is to actually save on power, i.e., as compared to achieving the same science research speed using productivity moduled labs but without any speed boost.

you are right that it's not the cheapest setup. it's good value though given the cost and rate of research.

[MeduSalem]

That looks very expensive to build and to run, iceman. Unless you're really short on space, using more labs should save quite a bit on power and construction costs I think — I haven't done the maths though.

Resource costs are one-time... so don't matter at all. Build once, profit forever.

And power is basically infinite as well.. especially with Solar Power or now Uranium Power.

So beaconize the hell out of everything and try to get 8 Beacons to affect 8 Labs/Assemblers and vice versa wherever posssible. (8:8 is the best ratio achievable imho)


But that said iceman is right... a Speed-beaconized setup using Productivity Modules is ALWAYS more energy efficient than a setup using Productivity Modules without Speed Beacons if the output rates/sec of both setups are comparable.

it's because the speed beaconized setup spends less time per item, hence wasting a lot less energy per item, which even outweighs the additional energy cost of the beacons.


The only thing you need to care about when using beacons in a setup is how much time the setup will be idle... and if it is going to be idle then use a power switch to turn off the idle power draw or otherwise the setup is not more energy efficient anymore.

[Lav]

That looks very expensive to build and to run, iceman. Unless you're really short on space, using more labs should save quite a bit on power and construction costs I think — I haven't done the maths though.

So beaconize the hell out of everything and try to get 8 Beacons to affect 8 Labs/Assemblers and vice versa wherever posssible. (8:8 is the best ratio achievable imho)

Perhaps I'm misunderstanding something, but what's the point of affecting a building with more than 2 beacons? The only thing that matters is the number of modules affecting a building, and there's a limit to how many speed modules you can apply before you hit the speed bonus limit. With 4 productivity module per building, 6 speed modules in beacons should be enough to hit that limit - that's 2 beacons with 3 modules each. Fourth slot can be spent on effectiveness to reduce power costs if necessary.

So essentially, for as long as each building in your setup is served with 2 beacons, you're golden.

Or am I wrong somewhere?

[solntcev]

There is no limit on speed bonuses.
Only efficiency effect capped at 80%.

[hoho]

There is no limit on speed bonuses.

Technically, there is. You can't complete more than one craft per tick. I assume in case of science labs, that means you can't consume more than one "bottle" per tick.

Of course, it takes mods to ever get to that point with labs.

[Lav]

There is no limit on speed bonuses.
Only efficiency effect capped at 80%.

Ah, I mixed them up then. Ok, thanks!