Testing science- and energyefficiency of chained science lab

[MeduSalem]

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. ...

Yeah, at the first page in the first post of the thread and I admit that I didn't read that one because I actually replied to the thread because of a post of someone else and the discussion moved on from there for me and I didn't read all the posts that where before it.

Also if he'd written that he updated OP, then I would probably have read it. So, sorry for that.



But about the Beacons... here's the math:

Labs (PM3) + Beacon (SM3):

Base Speed = 0.75
Speed Bonus = +370%
Productivity bonus = +20%

Total Speed = 0.75 * (1+3.7) * (1 + 0.2) = 4.23

Lab Base Consumption = 60kW
Lab Energy Consumption through modules/boost = +720%
Lab Power Draw = 60kW * (1+7.2) = 492kW
Beacon consumption = 480kW

In an optimal 8:8 setup each Machine goes with one beacon... so:

Total Power draw = 492kW + 480kW = 972kW


Labs (PM3), no beacons:

Base Speed = 0.75
Speed Bonus = -30%
Productivity Bonus = +20%

Total Speed = 0.75 * (1-0.3) * (1 + 0.2) = 0.63

Lab Base Consumption = 60kW
Lab Energy Consumption through modules = +160%
Lab Power Draw = 60kW * (1+1.6) = 156kW


Normalization to make the builds compareable in research output per same timeframe:

4.23 / 0.63 = 6.71

So the setup without Beacons needs 6.71 Labs for every lab of the beaconized setup to perform equally fast.


156kW * 6.71 = 1047kW.


1047kW > 972kW.


Technical K.O. for PM3 Setup without Speed Beacons. At least on the paper.


That the beaconized setup you tested doesn't perform as well as the hypothetical math says is just because the layout is not big enough yet and because of ill utiziliation of the edge/corner beacons. One shouldn't place a edge beacons that only cover 2 or 4 Labs. It's not worth it as they falsefy the result.


For other machines which allow 3 or 4 PMs (Chemplants, Assemblers) the difference becomes even more noticable.

[Distelzombie]

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. ...

Yeah, at the first page in the first post of the thread and I admit that I didn't read that one because I actually replied to the thread because of a post of someone else and the discussion moved on from there for me and I didn't read all the posts that where before it.

No... It was on the second page in the eleventh post. I just copied it to the first one for new people.
Anyway:

Ok, I see what i did wrong. I have beacons that didnt even reach labs. Uups. xD
Im gonna rebuild it in a row and remove beacons that only cover two or less labs. What do you mean with "dont place beacons that only cover 4 labs"? The range is not big enough to cover more than four or five. So youre saying that in the setup i had, I shouldnt build the left and right row of beacons? Could you show me a build to test?
And should i compare the build to an build with an equal amount of labs or 6.7 times more?

A lab with prod mods and 8 beacons around it has a speed bonus of 370%, energy increase of 720% and prod bonus of 20% according to tooltip.

[MeduSalem]

No... It was on the second page in the eleventh post. I just copied it to the first one for new people.

Ah well, probably before you edited it out, because when I took a look after you complained about us not being able to read it wasn't there anymore, if it would have been there I would have reacted differently and probably admitted my mistake earlier. I felt like I'm accused of something I didn't do, but probably did but couldn't comprehend in the aftermath due to your edits. So, really I'm sorry for the misunderstanding. It seems like it was my mistake.

Ok, I see what i did wrong. I have beacons that didnt even reach labs. Uups. xD
Im gonna rebuild it in a row and remove beacons that only cover two or less labs. What do you mean with "dont place beacons that only cover 4 labs"? The range is not big enough to cover more than four or five. So youre saying that in the setup i had, I shouldnt build the left and right row of beacons? Could you show me a build to test?

What I mean with that is the lowest row of beacons and the most upper row of beacons... they effectively don't contribute to the setup anymore other than just slapping their greedy 480kW onto the build, but the speed gained from them is negligible. So they only contribute negatively to the power demand.

To be honest, it's impossible to fully reach the 8:8 efficiency... You will always have either edge beacons or edge labs/assemblers because the setup isn't infinite (but it can be alleviated by making the setup huge... the amount of edge elements only grows linearly while the area grows quadratically, so eventually the edge items won't matter anymore, also the shape matters... it should be roughly square or so, rectangles become more inefficient in their circumfence). If it is better to have beacons on the edges or machines I'm not entirely convinced of as of now because of how the calculation becomes much more complex then (it's do-able but not as short as the above). I think that would take further testing or... more detailed math to find out.

Back when the topic of beaconizing stuff came up some people argued though that it would be better to have beacons on the edges because an additional Lab/Assembler is more costly than a Beacon, but I can't confirm that. Maybe it also depends on wether it is a lab, a chemplant or an assembler, since the amount of PMs in them influence their power draw immensely. In the labs the power draw with 492kW is almost equal to the draw of Beacons... so I don't know.

I think that one could be quite okay (if bigger of course):

test1.png (1007.73 KiB) Viewed 4772 times

or that one:

test.jpg (180.39 KiB) Viewed 4773 times

One of them will have a distinctively higher research/power rating... and that one is the better one.

[Distelzombie]

No... It was on the second page in the eleventh post. I just copied it to the first one for new people.

Ah well, probably before you edited it out, because when I took a look after you complained about us not being able to read it wasn't there anymore, if it would have been there I would have reacted differently and probably admitted my mistake earlier. I felt like I'm accused of something I didn't do, but probably did but couldn't comprehend in the aftermath due to your edits. So, really I'm sorry for the misunderstanding. It seems like it was my mistake.

AAAh! It is still there and it was always there! xD
You can see that I didnt even edit that post.

Ok I could start when you tell me:

And should i compare the build to an build with an equal amount of labs or 6.7 times more?

A lab with prod mods and 8 beacons around it has a speed bonus of 370%, energy increase of 720% and prod bonus of 20% according to tooltip.

[MeduSalem]

AAAh! It is still there and it was always there! xD

What the hell, damn it... I even entered the sentence in Firefox' Find-tool to look where it was located and it didn't find it on page 2, only on page 1 in the first post. I had to manually scroll there to find it. Weird. o_O

No matter what, I feel sorry, I really didn't mean to treat you badly.

Ok I could start when you tell me:

And should i compare the build to an build with an equal amount of labs or 6.7 times more?

I think it doesn't matter if you build 6.7 times more. In the end you can multiply the time it took in seconds with the total power demand you measured and you will know how much energy each setup requires on average to do it's job. Because even if the beaconized setup is much faster, but the power draw/sec will also be proportional to its size. The unbeaconized version will take much longer, but draw less power/second.

If the size of the beaconized setup is good enough then the beaconized setup should draw as much or LESS then the unbeaconized one, but not by much (only 7% or something).

But if you want to be absolutely sure you can try building a setup using 6.7 times more Labs.

A lab with prod mods and 8 beacons around it has a speed bonus of 370%, energy increase of 720% and prod bonus of 20% according to tooltip.

Yeah the numbers match up with my math obviously... otherwise I'd have done something wrong. Also the game never includes the base value... so it actually is like 470%, 820% or 120%. That's why I always made the (1+3.7), (1+7.2) and so on or you wouldn't get the correect speedup/power consumption. It's really weird that the tooltip doesn't show 100% when there are no modules/beacons boosting anything. xD

The same problem is actually currently a problem in Reactor Setups and their neighbour effects. A lot of people forget about the base value since the tooltip doesn't say that the base value isn't included in the percentage specified.

[iceman_1212]

@Distelzombie My bad, you did use productivity modules. I am indeed guilty of only reading the 3 bullets at the top of your post before jumping to the results section

I ran some tests also - here are my results, using the in-game production/consumption page and power summary page. Running creative mode mod to generate science packs and provide power. All cases assume productivity 3 modules in labs. The beacon case (i.e., case 2) uses speed 3 modules in beacons, in addition to prod mods in labs. All cases are run after completing all lab speed research and inserter stack size research. (All cases exclusively use stack inserters.)

Case 1: 5x5 square of science labs, prod 3, no beacons

Power consumption

Science pack consumption rate, trailing 1 minute

Case 2: Alternating rows of science labs and beacons, prod 3, speed 3 in beacons (total of 25 labs)

Power consumption

Science pack consumption rate, trailing 1 minute

@Distelzombie I'm not sure why your results had the beaconed setup as being slower than the unbeaconed setup.
Edit: Ah, I see that it was because some of the beacons were too far from some of the labs.

***

Here is where it gets interesting. In the third case (below), there are a total of 150 labs with prod 3 and without beacons - this is the number of labs needed to match the science pack consumption rate from case 2 above (i.e., 25 beaconed labs). The labs are arranged in a similar configuration as the beaconed setup for an apples-to-apples comparison.

Case 3:

Power consumption

Science pack consumption rate, trailing 1 minute

We see that the 150 labs have a power consumption is 23MW, which is lower than that of the beaconed setup which has a power consumption of 30MW.

It looks like productivity modules do not have a sufficient penalty to make beacons worthwhile from an energy-efficiency standpoint.

(That said, the difference in energy requirement between cases 2 and 3, i.e., ~7MW, is approximately ~1/1000th of the total energy required to power a factory that is capable of producing ~770 science packs per minute, so it's not unreasonable to prioritize UPS/space efficiency and go with the beaconed setup anyways.)

[MeduSalem]

@ iceman... Would be interesting to know if the gap becomes smaller the bigger the setup. (Even bigger what you already did)... The gap should become smaller... Somewhere there's got to be a breakeven point, because the math from above says there is one.

Probably the setup needs to be at least 32x32 big or so... Meaning 1024 Labs... affected by 1024 Beacons. Yeah ridicilous... but then again we have infinite research also growing exponentially so eventually you'd need it anyways.

The thing is that the circumfence becomes negligible then. It doesn't grow at the same rate as the area does (only linearly compared to quadratically) and that's why the edge beacons/labs which aren't fully covered become negligible as well:

circumfence per area.png (24.59 KiB) Viewed 4760 times

So basically the circumfence makes up only 12.5% of all the area at 32x32... while on smaller builds the edge beacons/labs would still be taking a huge amount of the area... and they are all influencing the energy consumption negatively.

Which is why your setup of 25 beaconized labs is performing that badly, iceman.

[Mehve]

Interesting. I would hAve placed my money on the unbeaconed setup using more power, but I suppose the base power consumption is so low that the beacon power consumption can skew things more than with smelters or Chem plants.

I suppose it's also worth comparing the total number of modules required, though. 300 for an unbeaconed setup vs 70-80 for a beacons? Definitely going to keep these numbers in mind for pre-megabase design.

[Distelzombie]

Here are my new Results (Got the grey right this time. xD)

Red: 12 Labs and 15 Beacons
Green: 12 Labs and 5 Beacons
Blue: 12 Labs

I like my test setup because there is only a limited amount of Energy available, equal for all three.

[MeduSalem]

Interesting. I would hAve placed my money on the unbeaconed setup using more power, but I suppose the base power consumption is so low that the beacon power consumption can skew things more than with smelters or Chem plants.

I suppose it's also worth comparing the total number of modules required, though. 300 for an unbeaconed setup vs 70-80 for a beacons? Definitely going to keep these numbers in mind for pre-megabase design.

Yeah, the setup of his is too small to take a benefit.

Like my table says 80% (actually only 64% since the corners have been counted twice in my table but there's definitely only 1 at each corner ) of the beacons are edge beacons.

So 64% of the beacons are performing only half as good (the corner ones even worse) as they could possibly do... and that's where the energy is wasted and why the beaconized setup can't close the gap on the unbeaconized one.

Here are my new Results (Got the grey right this time. xD)

Red: 12 Labs and 15 Beacons
Green: 12 Labs and 5 Beacons
Blue: 12 Labs

I like my test setup because there is only a limited amount of Energy available, equal for all three.

Ah well... so it is as following:

Red = 12MW * 66sec = 792MW
Green = 5.75MW * 137sec = 787.75MW
Blue = 1.8MW * 428sec = 770MW

So you see... the beaconized versions are actually drawing almost identical total power to do the same job. That they are a little bit worse is due to the edge effects, which becomes better if the setup is bigger.

[Distelzombie]

Ah well... so it is as following:

Red = 12MW * 66sec = 792MW
Green = 5.75MW * 137sec = 787.75MW
Blue = 1.8MW * 428sec = 770MW

So you see... the beaconiced versions are actually drawing almost identical total power to do the same job. That they are a little bit worse is due to the edge effects, which becomes better if the setup is bigger.

What do you mean? The green setup was the most efficient. The power draw in the power menu is not that accurate. Maybe thats why you're math is wrong in this case.
The Thing that is important in my tests is the green line from the research. xD

[MeduSalem]

What do you mean? The green setup was the most efficient.

Yupp... so enclosing the beacons with labs is better than enclosing the labs with beacons.

And if the real green setup is even bigger than the green test setup they will outperform the blue setup energywise.

[Distelzombie]

And if the real green setup is even bigger than the green test setup they will outperform the blue setup energywise.

Ah, what? Real green setup? The green setup is outperforming every other setup. The math you did is wrong because the power menu is not accurate enough. There is always a limited amount of MW - the same for every setup. So Red, blue and green can only use about 780MW. Thats what the accumulators are for. (800MJ)

I edited the post above too late for you to see.

[MeduSalem]

And if the real green setup is even bigger than the green test setup they will outperform the blue setup energywise.

Ah, what? Real green setup? The green setup is outperforming every other setup. The math you did is wrong because the power menu is not accurate enough. There is always a limited amount of MW - the same for every setup. So Red, blue and green can only use about 780MW.

I edited the post above too late for you to see.

D'oh. I thought you actually let the setup run with unlimited power, look what it is drawing in an average of 1min (divide it through 60 to get MW/s) or whatever and note how long it took to complete the research. Yeah well good. Foolish me.


Now the progress bars make sense... since the green one is the furthest it could do more work with the same amount of power than the blue and red one can.

[Distelzombie]

D'oh. I thought you actually let the setup run with unlimited power, look what it is drawing in an average of 1min (divide it through 60 to get MW/s) or whatever and note how long it took to complete the research. Yeah well good.


Now the progress bars make sense... since the green one is the furthest it could do more work with the same amount of power than the blue and red one can.

Yea, I want accuracy. Stupid power menu.
I even deleted the accumulator-indicator things I use before because the decreasing amount of lights over a variable timespan would increase power consumption asychronous between setups.

Would you say it makes more sense to read the science packs consumed instead of the progress bars? It appears to me as not as accurate though... hm I didnt entirely understand consumption of packs.

[MeduSalem]

I think that the progress bar is accurate enough... if after the same timespan one of them is more advanced than the other it is clearly performing more efficiently.

Do you mean reading the consumption off of the production screen or actually counting them with a combinator contraption? xD

If you read it off of the production screen then it's guaranteed to fluctuate as hell which makes it hard to tell which one is in the lead when they are close to each other.

With combinators it could actually work accurately, as accurate as it can possibly get.

[Distelzombie]

Do you mean reading the consumption off of the production screen or actually counting them with a combinator contraption? xD

I would use the 10min graph from prod screen and change game.speed to 2. But still .. it would only be a 1min average over 10 min
But its easy to count the amount of pack going into the labs as a single number if I use a single belt

Im trying to combine both setups: Include beacons in the Grid version ... still testing

[MeduSalem]

You could go ahead and try it and post the results...

[Distelzombie]

I just realized I HAVE to use combinator stuff to achieve good accuracy. The limited power source does not empty in exact 10min and the numbers shown are not accurate...
Hmpf combinators would need three different belts per lab...

[iceman_1212]

@Medusalem Tried going up to 1k beaconed by my comp couldn't take it. Here are the relevant screenshots for 768 vs. 4608 (multiple of 6, similar to 25 vs. 150 from my previous post).

Beaconed

Non-beaconed

It does look like they are converging as you say.

@Distelzombie Are your setups running at 100% electricity satisfaction for the duration of the test (and then immediately falling to zero when energy runs out)? Or are there prolonged periods of operation of the test setups below 100% electricity satisfaction. Production rates of beacon-based setups are more harshly penalized by <100% electricity satisfaction than those with no beacons. Also, can you elaborate on inaccuracies in the in-game power graph? Would be helpful to be aware of any issues generally speaking.