Modular tileable endgame science megabase building block

[NeuralParity]

What is this?

This is a modular, tileable UPS-friendly sci/min megafactory design blueprint. It takes in raw materials from 3-8 trains and researches science. Each building block unit is fully self-contained 62.5sci/min factory requiring 250MW of power.

Blueprint: https://pastebin.com/vxEEjZkA

Screenshots: http://imgur.com/gallery/YNAMA

What's the point?

Firstly, it allows you to scale from 1 science per second up to gigafactory size with exactly the same factory design.

Secondly, the modular design allows for transport (thus UPS) efficiencies not possible with some other designs.

Thirdly, it is my contribution into the religion war raging regarding where to place your smelters. Here, there is no smelting at the ore patch, nor is there either a centralised or a distributed smelting facility. They're all integrated into each modular factory.

How much science does each block produce?

It is designed for 62.5/sci and achieves a steady state 64-65 sci/min


Does this include military science?
Yes.

How UPS efficient is this?
Creative mode testing on my system (Ryzen 5 1600, 2666 Mhz memory, GTX1060) indicates that game updates take 12.7ms for 40 building blocks (i.e. 78 UPS @ 2.5k sci/min). Note that to actually run a 2.5k megabase you need to feed it with 7k mining drills (using speed modules), 1k pumpjacks, and a fleet of trains so 60 UPS is unlikely in a real game, but 45 isn't out of the question.

Why is this design UPS efficient?
It minimises the number of item movements required as well as the size of the logistics networks. A single small logistics networks keeps all bot trips short and I've laid out the assemblers to minimise the average distance travelled. By shipping ore I can maximise the number of direct (via chest) insertions thus minimising the total number of bot trips. Did you know creating green circuits uses 40% of the total item transport required? Using this design, I can direct insert not only all the copper wires required by the green circuit production, but also most of the copper and iron. The fact that the direct insert ratios are not perfect doesn't matter: the left over wires not used the green circuits get delivered (via bot) to my yellow science assembler. Downstream assembler requires more than you can produce? Output to a requestor chest. Produce more than the downsteam direct-insert assembler can use? Output to a passive provider chest.
Overall, I save around 40% on item transport through direct insertion and reduce the number of active bots required by the same amount. This more than makes up for the occasional sub-optimal ratio (18% more than perfect ratio).

How many bots per building block are required?
Steady state research requires 350 active (speed 9) bots (i.e. 5.6 bots per sci/min). Peak usage when you first bring a block online and all chests are empty is 700-750 bots. It should be stable by the second rocket launch (~30 minutes after startup - if you want it stable faster, disconnect the research labs or stop researching and wait for all the buffer to fill). By speed 12 you can get away with ~ 250 active bots.

What train system is this built for?
Each building block is designed for multiples of 3-8 ore/fluid trains. Train stations of the same raw material type should all the same name (e.g. IronUnload, CopperUnload, WaterUnload, ...). Train schedules for your 7th iron patch should look like: IronLoad7 (Train Full) -> IronUnload (Train Empty) You need a single giant stacker in front of your module stations. Trains pathing should look like ore -> stacker -> whichever unload station is free first -> back to ore

Why do you trains only go in one direction?
To prevent rail throughput between the stacker and the unload stations becoming a bottleneck at megafactory sizes.
Isn't running every train through a single track not viable?
It's working at 1k sci/min. If you're aiming for something bigger, then you can add another tile and run longer trains to reduce traffic.
Traffic-wide, you should budget for 1 iron ore every 60s, 1 copper every 120s, and >3min/train/module for the rest.

How is it both modular and tileable?
It is modular since each building block is fully self-contained, and it is tileable since building blocks can be stacked on top of each other to allow you to run longer trains. You can use a single building block to run 3-8 trains, or tile two blocks and run 3-8-3-8 trains, three blocks for 3-8-3-8-3-8 trains, and so on. By running longer trains, can either you reduce the amount of traffic on your rail network or increase your sci/min for the same level of traffic.
I found designing a rail network around 3-8-3-8-3-8-3-8 (4 building block) trains was unwieldly as it required at least 308 tiles between intersections and making new trains was just too tedious. I've found 3-8-3-8 to be the sweet spot and I can easily run a 1k sci/min factory with just a 2 track (one each direction) rail network.

How are you supposed to load a 3-8-3-8, or even a 3-8-3-8-3-8-3-8 train at an ore patch?
Use bots and circle the wagons around the ore patch. 3 engines is enough to turn 90 degrees so with 3-8-3-8 you can have an L shaped ore loading station, and with 3-8-3-8-3-8-3-8 you encircle the entire ore patch with wagons. This approach does not scale past 4*3-8. Is anyone seriously considering running longer trains?
Note that after a 90 degree turn, subsequent cars aren't perfectly aligned with the tiles so you can only load 5 inserters per side after your first 8 wagons and fluid wagons won't connect to the pumps at all. Unfortunately, water and oil loading stations must be straight.

Why are all your trains n*3-8?
Simplicity of rail network design.

Why are you using fluid wagons?
Because they're simple.

Aren't pipe bad for UPS?
I've heard that they are. Feel free to use barrels instead.

Why don't your fluid trains unload next to the refinery?
Because they need roboport coverage to refill their rocket fuel and I wanted to make sure it was impossible for roboport networks from adjacent modules to overlap.

Enjoy!

[Sicnarf]

Saw your post on reddit and almost made an account to comment on it, but then I found you here.
Love the idea and the implementation I have no idea how you got this running with so few bots that is amazing.

I am half way through building a base that is similar in a couple of respects though not nearly as well optimized or Tileable. An issue I ran into is that you cant use this type of build to produce the level 3 modules that you need so many off to make the base or any of the other pieces like mines and bots for that matter. I found a way around this (not to pretty) so that the factory can build it self.

Would you be able to reconfigure this build using the same space to say make it a one of everything you need to make the base? Also ones that just makes Modules? That way you could turn several tiles into factory component produces to help build the factory and then when the rail line is hitting max you can convert them back to Science. Quick note the most SPM you can fit across a single rail using just raw resources is about 2.3k SPM but that was with a 6:24 train.

Again this is a amazing piece of work you have conceived a type of base that achieves nearly all the things I wanted in a base including things that I did not know I wanted. My current build is too far along and to different to take advantage of your design but I have bookmarked this for when I start my next base and I will be "borrowing" some of your idea's.

[NeuralParity]

The modules are indeed a pain. My current project is exactly what you suggested: create a set of blueprints of the same size that make this base. Blueprints on my TODO list are:
- T3 module production
- 3-8-3-8 rail and output blueprints
- A sub-factory to make every needed in all the blueprints

The final part of the plan is a set of early/mid game blueprints so I don't have to keep working out what ratios I need before beacons and modules are available.

Module production is the real killer. I expect the best way to ramp up to megabase size is to make lots of module production sub-factories then rebuild them as science once the thousands of modules required have been made. Once this is all in place I might play around to see what a speed run to a 2k sci/min base looks like.

[Alkyen]

First, I wanna say that I love this blueprint, amazing work by op.

But I have a problem, how do you handle trains here? They seem to prefer the first few stations and not go to the "far" ones? I looked into some designs to fix it and it seems very messy
https://prnt.sc/jnwlhy

[NeuralParity]

The setup is intentionally designed with only 1 chest per wagon as I use the wagon itself is the buffer. During normal operation, an active factory module will have all 6 trains slowly unloading until they are finally empty. Only then will they leave the station. Given this design, having a train station preference is actually a good thing. You want all your trains to prefer the first station because then it'll actually do something. Having an iron train at the 5th station, a copper at the 1st, and a stone at the 3rd is pretty useless because the only run when all 6 trains are there.

The blueprint is stackable so you can run longer trains. In my megabases, I've been running a modulur stack of two factories which means 3-8-3-8 trains.

If you just run a single track from the stacker you'll eventually bottleneck on the track between the stacker and the stations when you scale up. I ended up running everything through a single stacker but split off a separate track for each 4 sets of stations. This minimises the length of track that served as a bottleneck and in some cases, allowed for multiple tracks to dispatch at the same time.

[pieppiep]

A solution to make bigger n*3-8 trains can be a seperate place where you unload smaller 3-8 trains and fill the bigger n*3-8 trains.

[ALT+F4]

This is a pretty interesting design, and testing it out on my new base has been lots of fun.

I ended up settling on the single 3-8 setup. In testing it I was finding that the 3-8-3-8 or larger setups would sometimes get hangups with certain portions of the trains being empty for one block, and not another, and in extreme cases having the trains get completely hung up where different trains would be empty for different blocks, and the whole block that was fed by that group of trains would stop, and I would have to manually fix it to get it running again. I'm sure this could be solved with some circuitry, or messing around with other stop conditions, but the single 3-8 setup is just easier to load and move around as well, and having the stop condition as empty cargo inventory is just easier.

I'm using setups that utilize 8 blocks of the single 3-8 module for a total of 500sp/m of research per "mega block". You could do more per "mega block", but you would have to get creative with having them fed from the same train stacker. I don't think a single rail will move enough trains to keep them all fed all the time.

I currently have 4 of these "mega blocks" for a total of 2k sp/m. This is surprisingly efficient, and even on my daughters Ryzen 1200 CPU is running at 60ups. My whole base is also using less than 12GW of power, so it's been pretty easy to just use solar as well, which really helps with the UPS.

Its also shockingly fast to expand once you have the pieces in place. I currently have 3 bases other than these research blocks. I have 1 to build my power (solar panels and accumulators), I have one building L3 modules, and one building the rest of the pieces I need to build these research blocks plus the other pieces I need for mining, etc. I'm confident I could add an additional 1k sp/m in under 10 hours of game play at this point, and continue to do that until I ran out of UPS or got bored.

Thanks a bunch for sharing these blueprints. Very good stuff.