Testing science- and energyefficiency of chained science lab

[Nich]

Oh wow I never thought about the tick limit.

I am assuming you can still produce 2 copper wire in 1 tick?

What happens as you approch this limit. I.e. 1.5 tick craft time? do you get 2 crafts every 3 ticks or 2 every 4 ticks?

[hoho]

I am assuming you can still produce 2 copper wire in 1 tick?

Yes. A machine can complete one craft per tick at most, doesn't matter how many products it generates.
I'm not sure how it works with productivity.

What happens as you approch this limit. I.e. 1.5 tick craft time? do you get 2 crafts every 3 ticks or 2 every 4 ticks?

It caps out at one craft per tick, no matter how fast you manage to make the crafting machine.

[MeduSalem]

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.

Yeah well.

The fastest Recipes are 0.5s... and with Speed Beacons I think it's barely possible to go below 0.08-0.1s? ... which is far away from the 1/60=0.01666s limit. At least in Vanilla... so yes, mods would be required to go further, but I rarely ever play with mods.

[Nich]

12 becons and 4 in the machine give 1600% speed

.5 / (1.25*17) = .0235

at 2300% bonus speed is the range where you get into the tick production

fixed*

[MeduSalem]

You can place 12 beacons around a single assembler... so even faster.

But that said that many beacons around a single assembler is not energy/space efficient. Like said on the previous page you want a beacon to affect as many machines as possible while at the same time a machine to be affected by as many beacons as possible.

And that is only really viable with 8:8 setups. That way the energy consumption of a single beacon is distributed among 8 machines, while 8 machines can benefit from the speed boost of a single beacon. It's the time-proven sweet spot.

I've done excessive research on the matter 2-3 years ago with a lot of excel spreadsheets and it doesn't get any better. If you are using Productivity Modules it's the way to go, everything else is inefficient. If you are not using Productivity Modules then the matter is different anyways.

Maybe it is time the devs spiced up the beacon mechanics a bit though, because in my opinion it's getting a bit boring already... because the optimal layout has been found a long time ago and there hasn't changed anything about it for years.

[Nich]

HAHA no idea how I miss counted lol fixed above

[Lav]

Getting back to the original topic.

I have tested the direct insertion setup against a sushi belt in a couple of different configurations.

No beacons, 25 labs in each setup, 6 packs for direct insertion setup vs 7 packs on the sushi belt (military science wasn't being inserted into labs, but it still took space on the belt). Research speed was doubled via the console. Both setups were researching Mining Productivity 17 tech (1600 x 60 seconds x 6 packs). Time was measured by self-auto-incrementing combinator. As soon as the research was done, I disconnected power supply to the lab complex and checked the combinator output.

Testing setup

Test results:

1. Direct insertion setup researched Mining 17 tech in 80,286 ticks (~79,500 if we ignore the warm-up period - see note below). There were frequent "hiccups" in lab performance when inserters start working.
2. Sushi belt setup researched Mining 17 tech in 77,250 ticks. I haven't noticed even a single hiccup on the graph. That may have been luck, but all 25 labs worked at 100% performance 100% of the time.
3. Experimental sushi belt setup (3 inserters per lab, and injection belts not checking previous belt tile) was a failure. 3 inserters apparently offer no benefit over one (but cause no harm either), while not checking previous tile at injection points caused military packs to take too much space, periodically starving the last labs on the loop of production, high-tech and/or space packs. I had to drain the sushi belt of military packs at ~70% of research, and it eventually completed the research in 77,581 ticks. Not sure how long it would take if I didn't drain the military packs, and there's no beating 100% performance of standard setup anyway.

Conclusions:

1. A looped red sushi belt handled 25 labs on 2x research speed without any issues - should be able to handle 50 labs without console hacks.
2. Direct insertion 5x5 setup results in ~3% research speed reduction long-term.
3. Stack inserters are very hungry - for direct insertion setup, inserters consumed as much power as labs themselves. Meanwhile, fast inserters for sushi belt setup only consumed ~10% of labs consumption. Essentially, direct insertion setup was almost twice as power-hungry. Maybe I went overboard on the number of stack inserters, but no matter how you reduce them, direct insertion setup is noticeably less power-efficient.

Note: sushi belt takes an extra ~2800 ticks to "warm up" from an empty state. However since it's unlikely to ever be in such state at the beginning of a research, I was pre-populating it before actually starting the research. Direct insertion setup is noticeably faster to "warm up" (about 780-800 ticks), so it's actual performance is closer to ~79,500 ticks, rather than 80,286 that was measured.

[MeduSalem]

And why not make it a belt braid if it has to be belts? Too cheaty?

research.jpg (269.05 KiB) Viewed 4392 times

Enough for 8 different science packs.

[Lav]

And why not make it a belt braid if it has to be belts? Too cheety?

Frankly I just don't see the point. A sushi belt is pretty simple, you can build it very early in the game, and it will never need any additional maintenance, ever.

Maybe if I played a mod with more than 8 science pack types.

[MeduSalem]

Well, I hate Sushi belts to be honest. Back when I tried them last year when belts could be hooked up to the CN for the first time I had a hard time to get the belt saturated in equal proportions. It mostly always resulted in something getting out of balance and the entire belt filling up with too much of one item type and then the entire thing got stuck and I had to make it unstuck manually.

That is why ever since I'm not a friend of sharing the same belt side for multiple items. Basically I play it safe where possible. If something is doable without CN I do it without. Chance to fail is 0 with my approach.

That said the initial implementation of controlling the belts last year has been quite buggy and showed quite weird behaviour altogether. That might have changed ever since, but well... you know as it is... bad experience renders you prejudiced.


Also later on I change to the logistic network anyways... Mostly because of performance reasons as my Computer is getting 9 years old. Otherwise I wouldn't be able to maintain a megabase if there were a lot of belts everywhere... especially with the rising throughput demand. Belts are fine if you never need to expand anything, but if you need to extend something then belts will eventually reach their limit and if the bus becomes the bottleneck then it's a lot of work to fix it.

[Distelzombie]

I was wondering if lab setups with beacons an productivity modules really are more energy efficient as some people here stated. So I tested it:

Three setups:
- My setup with 25 labs and no beacons
- A setup with one lab and 12 beacons
- A setup with 1 lab and 8 beacons (as its usually build)

I used a setup of many accumulators that I first load using a creative power source. The number of stored MJ/accumulators is equal in all setups. After I started the research "Follower-robot count 8" I disconnect the power source and use a timer to measure the time it takes to empty the accumulators.
I chose folloer-robot count 8 just because it is a 30s research and it uses all seven science packs.
In all Labs were productivity modules and in all beacons were speed modules.

As you can clearly see the setup with 25 labs uses less energy per second and gives the most research than any other setup.

Do you want me to test an certain setup?

[MeduSalem]

The above setups are not compareable.

You'd need setups that have roughly euqal Research/Second to make them comparable... and then see how much the energy usage is.

[Distelzombie]

I dont see a problem here. It is obvious that the setup with 25 labs uses less energy for more than double the amount of research than a SINGLE lab with beacons. It would just get worse if you extend the beaconed setups.

I someone want my battery-charge display:

Code: Select all

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

Connect power line to the switch to make it toggleable


----------------------
Ok I realized you're right (I probably forgot how productivity works') and I tested a setup with 25 labs and full beacon coverage.
But no change in the outcome.

[iceman_1212]

Beacons are generally not used without productivity modules. So need to compare a beaconed + productivity setup to an unbeaconed + productivity setup to do an apples-to-apples power usage comparison.

The increased power cost compared to a setup without productivity modules is simply the premium that we pay in exchange for getting a free science cycle every five cycles.

[MeduSalem]

I dont see a problem here. It is obvious that the setup with 25 labs uses less energy for more than double the amount of research than a SINGLE lab with beacons. It would just get worse if you extend the beaconed setups

Did you put Productivity Modules in all three setups?


Also you are wrong about that it would get worse if one extends the beaconized setups. The fact is that they improve their efficiency the bigger they are. In a true 8:8 setup the power consumption results to (1 Lab + 1 Beacon) * Amount Of Labs. And the speed increase is 720%.

And in a setup without Beacons it would take a lot more Labs to achieve the same throughput... which require a lot more energy.

[Distelzombie]

1. Did you put Productivity Modules in all three setups?

2. Also you are wrong about that it would get worse if one extends the beaconized setups. The fact is that they improve their efficiency the bigger they are. In a true 8:8 setup the power consumption results to (1 Lab + 1 Beacon) * Amount Of Labs. And the speed increase is 720%.

3. And in a setup without Beacons it would take a lot more Labs to achieve the same throughput... which require a lot more energy.

1. Yes, please actually read what i write.
2. I know now, please see edited post above
3. Efficiency will get better the more productivity modules one uses. Therefore the setup without beacons will be even more efficient if it has more labs = more prod modules.
You see that too if you compare the new setup with 25 labs and full beacon coverage to the old one with one lab and full beacon...

[Distelzombie]

Beacons are generally not used without productivity modules. So need to compare a beaconed + productivity setup to an unbeaconed + productivity setup to do an apples-to-apples power usage comparison.

The increased power cost compared to a setup without productivity modules is simply the premium that we pay in exchange for getting a free science cycle every five cycles.

I did that. WTH? Why is nobody READING WHAT I WRITE? This is the forth or fifth time IN THIS THREAD someone didnt really read the stuff i wrote and i had to repeat myself... Damn whats wrong? Are my sentences so boring?

[MeduSalem]

Sorry but you didn't write that ALL setups are using productivity modules and due to the many science pack icons it isn't clear from the screenshot. So don't say people are not reading what you write if you didn't write it. If 2 people are saying the same then it was obviously not clear.

[orzelek]

Sorry but you didn't write that ALL setups are using productivity modules and due to the many science pack icons it isn't clear from the screenshot. So don't say people are not reading what you write if you didn't write it. If 2 people are saying the same then it was obviously not clear.

A bit nitpicky... but he wrote it twice in one post - one time in very explicit manner:

... In all Labs were productivity modules and in all beacons were speed modules. ...

PS.
I understand the pain Distelzombie... reading is hard.

[Distelzombie]

I understand the pain Distelzombie... reading is hard.

Seriously! xD

Ok, just to be even more clear this time: My point with this second test was to show ENERGY efficiency. Apparently beacons do not increase energy efficiency. (As stated by someone in this thread earlier). Beacons just increase speed: Therefore a setup with 25 labs and full coverage has a lower energy-to-research ratio than one without beacons.