Challenge: maximize research density

Post all other topics which do not belong to any other category.
Post Reply
pichutarius
Fast Inserter
Fast Inserter
Posts: 158
Joined: Mon Feb 27, 2017 12:47 am
Contact:
Contact pichutarius

Challenge: maximize research density

Post by pichutarius »

Rules:
- creative mode: you get ore, water, oil, electricity for free
- goal: design a rectangular (other shape not allowed) factory, takes in resources, turn it into 6 type of science packs (excluding military), and consumed by labs.
- scoring is science consumption per hour per meter^2 (more on that later)
- you're allowed to use any number of buildings, modules, etc. The only constant supply are ore, water, oil and electric

Example
Math and stats
Story
QnA


Top
User avatar
DaveMcW
Smart Inserter
Smart Inserter
Posts: 3700
Joined: Tue May 13, 2014 11:06 am
Contact:
Contact DaveMcW

Re: Challenge: maximize research density

Post by DaveMcW »

0.41 sph/m^2

I like to play zoomed in.
Top
pichutarius
Fast Inserter
Fast Inserter
Posts: 158
Joined: Mon Feb 27, 2017 12:47 am
Contact:
Contact pichutarius

Re: Challenge: maximize research density

Post by pichutarius »

wow.... that's impressively small!
Top
Koub
Global Moderator
Global Moderator
Posts: 7199
Joined: Fri May 30, 2014 8:54 am
Contact:
Contact Koub

Re: Challenge: maximize research density

Post by Koub »

We're talking about DaveMcW. He uses dark magic to do impossible things :lol:.
Koub - Please consider English is not my native language.
Top
Jap2.0
Smart Inserter
Smart Inserter
Posts: 2339
Joined: Tue Jun 20, 2017 12:02 am
Contact:
Contact Jap2.0

Re: Challenge: maximize research density

Post by Jap2.0 »

pichutarius wrote: ↑
Sun Apr 12, 2020 6:19 am
 wow.... that's impressively small!
Just wait until they make recursive blueprints vanilla...
There are 10 types of people: those who get this joke and those who don't.
Top
pichutarius
Fast Inserter
Fast Inserter
Posts: 158
Joined: Mon Feb 27, 2017 12:47 am
Contact:
Contact pichutarius

Re: Challenge: maximize research density

Post by pichutarius »

So i manage to double sph, uses less than twice the area, result in higher research density.

Science rate = 264 spm = 15840 sph
Area = 162 m x 91 m = 14742 m^2
Research density = 1.074 sph/m^2 (and it even includes military at the unused corner)
4.4.png
4.4.png (2.07 MiB) Viewed 7652 times
stats.png
stats.png (167.3 KiB) Viewed 7652 times


We can plop 4 of these and reach 1 rpm

Top
GrumpyJoe
Filter Inserter
Filter Inserter
Posts: 443
Joined: Fri Apr 06, 2018 7:10 pm
Contact:
Contact GrumpyJoe

Re: Challenge: maximize research density

Post by GrumpyJoe »

pichutarius wrote: ↑
Sun Apr 12, 2020 5:49 am
 Rules:
turn it into 6 type of science packs (excluding military), and consumed by labs.
before i attempt this, can we clarify the wording so that it has to be a reasonable consumption?
You know, the way you wanna put in prod modules into your labs to get the most bang for your buck?

Simply consuming them, I would put speed modules into the labs, saving space
Top
User avatar
Khagan
Fast Inserter
Fast Inserter
Posts: 233
Joined: Mon Mar 25, 2019 9:40 pm
Contact:
Contact Khagan

Re: Challenge: maximize research density

Post by Khagan »

As the TV chefs say, here is one I made earlier. 200 spm (including military) in 148x156 for a density of 0.52 sci/h/m2. (I'm still tinkering with the expanded version which will be 1200spm in significantly less than 6 times the space.)
Top
pichutarius
Fast Inserter
Fast Inserter
Posts: 158
Joined: Mon Feb 27, 2017 12:47 am
Contact:
Contact pichutarius

Re: Challenge: maximize research density

Post by pichutarius »

GrumpyJoe wrote: ↑
Tue Apr 14, 2020 11:21 am
pichutarius wrote: ↑
Sun Apr 12, 2020 5:49 am
 Rules:
turn it into 6 type of science packs (excluding military), and consumed by labs.
before i attempt this, can we clarify the wording so that it has to be a reasonable consumption?
You know, the way you wanna put in prod modules into your labs to get the most bang for your buck?

Simply consuming them, I would put speed modules into the labs, saving space
Yes you can, i cover that in QnA
Q: You should insert speed3 in labs, not prod3.
A: If all we care about is science consumption, of course speed3 in labs makes more sense, but lets face it, nobody does that. Its more reasonable to put prod3 inside labs. Consider this as an optional challenge.
So you are allowed to put speed3, but remember the base must be rectanglular, you wont be surprised there will be remain unused space at the corner. In this case plop some labs with prod3 doesnt cost extra area.

The reason i use science consumption (not research production) is because most megabase use spm (without the x1.2 bonus from lab prod3)
Top
pichutarius
Fast Inserter
Fast Inserter
Posts: 158
Joined: Mon Feb 27, 2017 12:47 am
Contact:
Contact pichutarius

Re: Challenge: maximize research density

Post by pichutarius »

Khagan wrote: ↑
Tue Apr 14, 2020 10:48 pm
 As the TV chefs say, here is one I made earlier. 200 spm (including military) in 148x156 for a density of 0.52 sci/h/m2. (I'm still tinkering with the expanded version which will be 1200spm in significantly less than 6 times the space.)
The base can be further shrink down using blue belts. But maybe its just your intermediate base, not end game base.

if you going over 1000 spm , you would need at least 27 blue belts just for iron ore alone. sadly, bots are more space efficient than belts.
Top
quyxkh
Smart Inserter
Smart Inserter
Posts: 1028
Joined: Sun May 08, 2016 9:01 am
Contact:
Contact quyxkh

Re: Challenge: maximize research density

Post by quyxkh »

I'm thinking you can save at least 10% on the space with buffer chests, four-wide production lanes instead of five.
Top
User avatar
Khagan
Fast Inserter
Fast Inserter
Posts: 233
Joined: Mon Mar 25, 2019 9:40 pm
Contact:
Contact Khagan

Re: Challenge: maximize research density

Post by Khagan »

pichutarius wrote: ↑
Wed Apr 15, 2020 12:41 am
Khagan wrote: ↑
Tue Apr 14, 2020 10:48 pm
 200 spm (including military) in 148x156 for a density of 0.52 sci/h/m2.
The base can be further shrink down using blue belts. But maybe its just your intermediate base, not end game base.

if you going over 1000 spm , you would need at least 27 blue belts just for iron ore alone. sadly, bots are more space efficient than belts.
Yes, it's just intermediate. But even my real endgame one doesn't use blue belts for ore; for low-value items they're just not cost-effective compared to red ones. It would be a different matter if I were deliberately designing for the challenge in this thread, where cost is not a consideration.
Top
quyxkh
Smart Inserter
Smart Inserter
Posts: 1028
Joined: Sun May 08, 2016 9:01 am
Contact:
Contact quyxkh

Re: Challenge: maximize research density

Post by quyxkh »

Okay, I'm not going to build a finished stampable one because bin packing on this scale is not my cup of insanitea, but here's as far as I got, there's some big wins there. I have to admire your oil and bluechip/battery setups up front though, I tried to improve the oil but mine came out worse and that bluechip/battery setup I'm'a just plain steal.

Anyway, the wins I found:
  • Four-wide lanes. Pack almost everything at 18*3×(5*8+3)/(14*5-4)~30.1m²/crafter instead of the five-wide 18*3×(2*8+3)/30=34.2m² per.
  • Bot management. I ran the entirety of purple science production on eight roboports total. Only green chips really need a full inner lane of 'ports, and I suspect they could be run along the edges to eliminate those. Besides making the high-traffic routes short, an even bigger trick is clocking unloads to get multiples of four. From your four full inner lanes of ports bot management can get it down to about one, and I think even that could be reduced.
  • Speed modules in primary processing (iron/copper/oil), late game this is pure win since once mining productivity gets past ~300% the ~7% reduction in furnace count is larger than the 20% increase in (much lower) miner count, and before then the loss, the more extra miners spent than furnaces saved, is not large.
Four-wide lanes: Your five-wide-lane setup's tileable chunk gets 34.2m²/crafter
snap@T23898446=3536x1264+163.25+83.5,z2.jpg
snap@T23898446=3536x1264+163.25+83.5,z2.jpg (413.71 KiB) Viewed 7347 times
Even in this worst case, giving up seventhree assemblers to get full rate out of the green-chip assemblers, the four-wide chunk gets 18*3*38/(14*5-7)~32.6m²/crafter, with full packing it gets ~30.1, about a 13.5% improvement:
snap@T267548=3568x2528-77.25+45,z2.jpg
snap@T267548=3568x2528-77.25+45,z2.jpg (825.22 KiB) Viewed 7347 times
Smelter unloads are clocked every 192 ticks for iron and copper, 1022 for steel that gets a full 12 since why not. The full ticks math is recipe_time/#output/crafting_speed/productivity*handsize*60 ticks.

Using the buffer chests for smelter chains means you have to be aware of the `request from buffer chests` flag on your requesters, and the priorities (buffer chests can't request from buffer chests) can make things weird. But they can also help, for instance in the bluechip factory I just didn't want to do the math on clocking the outputs, so I left the unload chests red and put a buffer chest at the end of the row to collect everything for longer shipping. Since bluechips and batteries aren't the bottleneck all seven of the factories supply the a single buffer so the partial loads from uneven drain only hurt the short trip to the buffers.

edit: switched assemblers-per-row accounting (14 per full row somehow became 14 per every row in my head) in the middle of the arithmetic, duh. Fixed the numbers. edit 2: uneven drain is a thing, but you can limit its impact on bot traffic with buffer chests.
Top
pichutarius
Fast Inserter
Fast Inserter
Posts: 158
Joined: Mon Feb 27, 2017 12:47 am
Contact:
Contact pichutarius

Re: Challenge: maximize research density

Post by pichutarius »

quyxkh wrote: ↑
Sat Apr 18, 2020 10:18 pm
 Okay, I'm not going to ...
an improvement is "4.5-wide lane" , which means rows with substation uses 5 tiles, and without that uses 4 tiles.
not sure if your calculation m²/crafter is correct, we should exclude one row of beacon, consider infinite tiling, crafter:beacon = 1:1 in term of lanes.
so consider 2 lanes,
4-wide lane : 54*16 / 30 = 28.8 m²/crafter
5-wide lane : 54*14 / 27 = 28.0 m²/crafter
4.5-wide lane : 54*15 / 30 = 27.0 m²/crafter

However 4-wide layout remove alot of "inserter connections" so that will put more pressure on roboport. Also the corner crafter doesnt have enough space for both input/output so u need to push roboport outside, which cause problem in tiling horizontally, reducing density. Might need to consider that.

The 2 other point u made is great. Now u plant the idea that 4 tile is denser, i might want to try it, but it might take another week off mine. Am i that insane enough?
Top
quyxkh
Smart Inserter
Smart Inserter
Posts: 1028
Joined: Sun May 08, 2016 9:01 am
Contact:
Contact quyxkh

Re: Challenge: maximize research density

Post by quyxkh »

pichutarius wrote: ↑
Mon Apr 20, 2020 2:29 am
 However 4-wide layout remove alot of "inserter connections" so that will put more pressure on roboport.
Ah but thats the good part, logistics chests handle that fine. Steel for instance: speed-moduled iron smelter feeds passive provider chest feeds prod-moduled steel smelter. Excess iron is available to the logistic network but the steel runs nonstop. You can even see the I guess negative space of the idea in the picture above, where I had to put stack filter inserters on sequential iron smelters to *stop* the inserter connections.

The LDS layout for instance is cute, check it out: two pairs of copper smelters feeding buffer chests requesting the full LDS recipe, followed by one extra copper smelter, and so on. Since one lds factory takes 1.46+ S3'd-smelters' worth of copper (3.75/s output/smelter, factory needs 5.5/s (20s recipe @ 20 copper is 1 copper/s * crafting speed 5.5) all the copper from the two direct feeds goes straight to the lds, and 92% of the copper from the spare takes a very short hop, only about 3% of the copper takes a long trip. You *could* do it with the two-P3'd smelters-per, do them inline going in different directions, but about the same fraction of the copper produced would be taking a long trip and the S3'd way takes ~7% fewer iron/copper smelters overall, saving space is the point, right?

Check it out,

the green wire's fed from a combined iron/copper filter setter that staggers the unload signals for simpler wiring:
snap@T26069055=3360x608-70.75+208.25,z2.jpg
snap@T26069055=3360x608-70.75+208.25,z2.jpg (201.33 KiB) Viewed 7249 times
edit: p.s. when I did quickie space math in my head I screwed it up too, see the corrections to my earlier post, but you don't have tio trust it, you can count up the space and smelters from the pictures above.
Top
pichutarius
Fast Inserter
Fast Inserter
Posts: 158
Joined: Mon Feb 27, 2017 12:47 am
Contact:
Contact pichutarius

Re: Challenge: maximize research density

Post by pichutarius »

still it doesnt tile nicely horizontally, might need to tinker about how to fit the puzzle pieces.

my next goal is to build 462 spm (from 5:6:12:7:7) with these tips in mind. but it might take weeks to design one. (or maybe i wont, because its kinda insane)
Top
pichutarius
Fast Inserter
Fast Inserter
Posts: 158
Joined: Mon Feb 27, 2017 12:47 am
Contact:
Contact pichutarius

Re: Challenge: maximize research density

Post by pichutarius »

after a week of melting my brain, i finally come out with higher research density.
i thank quyxkh for his suggestion for 4-tile wide lane, and speed module on furnace. (i lost one week of my free time over this :cry: )

this factory has 462 spm and area = 162 * 132 = 21384 m^2
so the research density is 462 * 60 / 21384 = 1.296 sph / m^2
which is 20.64% denser than the last factory.

here it is:
screen shot.png
screen shot.png (2.24 MiB) Viewed 7112 times
1hr.png
1hr.png (68.01 KiB) Viewed 7112 times
layout.png
layout.png (23.27 KiB) Viewed 7112 times
at least 4000 bots, bot-speed +500% is needed to run the factory.

462spm come from science 5:6:12:7:7 all prod3 and fully beacon-ed. hence the bottleneck is science. labs has capacity of 462.35 spm so it comes second on the list.
every assembly is exactly the number needed, hence it takes forever to converge to 462 spm. After 2 hours it still fluctuates around 458 spm.

the solution is to pause the research, let the chest refilled and bots finish their job, then resume the research.
from the 10-hour graph, it shows before pausing, the spm is unstable, after resuming it can run 462 spm for more than 7 hours.
10hr.png
10hr.png (67.83 KiB) Viewed 7112 times
i'm pretty satisfied with the result, just plonk down and plug in the resources and electricity.
Top
Niko8
Manual Inserter
Manual Inserter
Posts: 1
Joined: Mon Sep 21, 2020 10:40 pm
Contact:
Contact Niko8

Re: Challenge: maximize research density

Post by Niko8 »

Hi. Why is there no military research?
Top
User avatar
DaveMcW
Smart Inserter
Smart Inserter
Posts: 3700
Joined: Tue May 13, 2014 11:06 am
Contact:
Contact DaveMcW

Re: Challenge: maximize research density

Post by DaveMcW »

Because "Follower robot count" is not a very useful technology.
Top
User avatar
Khagan
Fast Inserter
Fast Inserter
Posts: 233
Joined: Mon Mar 25, 2019 9:40 pm
Contact:
Contact Khagan

Re: Challenge: maximize research density

Post by Khagan »

DaveMcW wrote: ↑
Tue Sep 22, 2020 2:16 am
 Because "Follower robot count" is not a very useful technology.
True. But the other infinite military techs are. Though "Follower robot count" is (for hysterical raisins?) the only one to require military tech as well as production tech; the others are all military instead of production.
Top
Post Reply

Return to “General discussion”