10K Monolithic Train Base with extra pictures.

[Stevetrov]

Firstly

Google style map of base thanks to /u/MathWizz94.

0.17 Savefile can be found here

0.18 Savefile can be found here with fixes for some stuck trains

After I posted my last large megabase on reddit, a friend commented that monolithic bases were more satisfying than modular ones, so I decided to build a monolithic train base.

So this is the top level design:

• Loads of trains

• Sub-factories dedicated to single output, connected with a “drive anywhere” train network.

• 3 factories for the 3 different plates (iron, copper & steel)

• 3 factories for the circuits (green, red & blue)

• Factories for each of the science packs (except military)

• Oil refinery inc plastic, sulfur & rocket fuel.

• 10K SPM

• Reasonable UPS

Stats

So here are some stats about the base:

• 40-50 UPS on my laptop

• 82 UPS on friends 8700K (headless)

• 600 trains

• 50K+ rockets launched

• 12,000 rolling stock (locos & wagons)

• 1.5M solar panels

• 7,400 ASMs

• 8600 furnaces

• 48,000 inserters

• 475,000 rail pieces

• 9.9K SPM sustained for 50h

• 265h played on this map, some of that was unattended to test everything was working.

Rail network

I knew I would need scalable rail junctions, and a quick search around found that the only junctions with the required throughput were some of the bigger junctions in the 4-way junction megathread. Unfortunately these didn’t seem to scale well to the size of the trains I was planning to use (20 wagon lengths).



After playing around with a number of different designs I settled on this design. Initially all the junctions were going to be 8 lanes each way but I upgraded some of them to 16 lanes each way to avoid trains backing up. Most of the junctions are T junctions but I couldn't resist using one of these behemoths for the circuits. I did experiment with using CN controlled signals (as in the gif I shared) but unfortunately I had a lot of issues with them not working as expected and trains getting stuck in the middle of junctions.

The rail network is a bit over-engineered to reduce the chance that a train has to slow down and wait at a junction, but not by a massive amount.

I also had to build some interesting junctions to get the ore balanced across all the smelters.

UPS

Although optimising UPS wasn't the primary concern for this base, I wanted it to run 10K at reasonable UPS, so I used some direct insert rail based designs that I had developed in conjunction with others on various public servers over the last year or so. I didn't have BPs for all the necessary recipes so I just built the first thing that came into my head along the same lines as the optimised ones without actually benchmarking anything. IIRC the core of the smelter was designed by /u/Phoenix27833.



I am sure that if I were to refine the builds, test variant and get more DI I could improve performance, but thats my next project.

Oil Refinery

I have been testing oil builds that use basic oil processing to avoid having to crack light -> gas and they have been performing very well so I used that approach for this refinery.




Adv processing is used to produce enough light for rocket fuel, the gas is made into plastic, the heavy is made into lube and any excess is cracked to light. I had a fair amount of trouble with keeping fluids balanced and ended up adding some CN controlled filter inserters to disable production when gas pressure was low.

As this only produces about half the required plastic and none of the sulfur produces these are produced separately using basic oil processing.

Labs

The build was one that I designed in conjunction with Darklich14. The setup is fairly simple but its not immediately obvious how it works.

There is a steel chest on each side of the track aligned with each of the cargo wagons.

First of all, the chests are “primed” with the following

• 333 red science

• 667 green science

• 1000 blue science

• 1333 purple science

• 1667 yellow science

• 2000 white science

Once red science is below 200 the red science station activates and a red science train unloads 2000 red science from each wagon into its chests. Then once green science is below 200 the green train unloads 2000 science into the chests etc..

As between each train we process 333*6 ~= 2000 packs the chests never overflow. As a cargo wagon can hold 8000 packs the wagons are limited to only using 20 slots. This also helps to smooth out production at the science pack factories that mostly have a single steel chest for the output buffer.

16 row of 32 labs is sufficient to get us to 10K

In reality the rows were grouped in fours and each group was given a different primed state so they would all require red packs (or any other color) at the same time.

Note about how this base was built.

This base was not built in the traditional manor. Instead it was completely built using the editor. Starting on a map with no biters, no water and no resources.

So all resources were placed manually using the editor tools and to keep in the spirit of one of everything I made one mine for each resource: copper, iron, stone, coal, water & crude oil & uranium (although they are all rich effectively never run out.

During the building of the base I used infinity chests to provide intermediates that I hadn't been built yet and as sinks to test that the factory could perform at the required throughput, I think they have all been removed, but if the prod stats look a bit weird that's probably why.

I also started with all base research complete and some MP research although I cant remember exactly how much.

Finally

If you made it this far well done, thanks for your time and feel free to ask any questions

[coppercoil]

Thanks for sharing. It's unusual and very compact megabase. It looks like there's no single belt

I saw stations that are always blocked. What is its purpose?

UnavailableStations.jpg (295.04 KiB) Viewed 10349 times

[azesmbog]

• 40-50 UPS on my laptop

• 82 UPS on friends 8700K (headless)

Thanks for the next interesting megabase, but still I'm more interested in the performance of UPS in it.
on my rather weak processor i5-6400 ($ 200), I got the following results:

UPS avg - 81.73, max - 96.58, min - 23.53

and you write that at 8700K ($ 500) (scalped?, as I understand it, and therefore overclocked?) - the same 82 ups?
the question is, am I still going to buy an i7-9700K processor, and it doesn’t differ much from 8700K, and in the end I won’t get any increase in UPS?
or we test performance differently, or something is wrong with the numbers (for me).
How to check?))

[Stevetrov]

Thanks for sharing. It's unusual and very compact megabase. It looks like there's no single belt

There are actually belts in the blue science build but I think thats it, there are probably some places where belts might be better than bots, like the rocket silo setup.

I saw stations that are always blocked. What is its purpose?

Yes the main stations that the trains actually visit are disabled when they do not require a visit from a train (chests too full or too empty). So these stations act as dummy stations to give the trains something to path too, that way the trains stay in their stackers and dont no path or skip the station.

[Stevetrov]

UPS avg - 81.73, max - 96.58, min - 23.53

Couple of questions,

1. how many ticks did you run the benchmark for? I think the original test was done for 10,000 ticks.
2. How fast is the memory in your desktop? As memory is probably more significant than CPU.

[azesmbog]

factorio --benchmark "Stevetrov_train_10K_megabase.zip" --benchmark-ticks 1000 --disable-audio --mod-directory /dev/null --benchmark-runs 5 >test.txt

That's how I measured performance.
There are 1000 ticks.
Do I need to put 10,000 ticks ??
What does this affect ??

The memory frequency of 2946 MHz, at present, is not very high.

upd:
ran tests with 10,000 ticks
The average and maximum values of UPS have practically not changed; the minimum value has significantly fallen.
But we focus on the average, right ??

[Stevetrov]

upd:
ran tests with 10,000 ticks
The average and maximum values of UPS have practically not changed; the minimum value has significantly fallen.
But we focus on the average, right ??

yea its the average values we care about, the underlying data is actually very noisy.

BTW are you on windows? apparently windows has significantly better performance than linux and the tests on the 8700K were on linux.

[azesmbog]

Yes, I'm at the windows. Win7
I would be very pleased with the test results of this base on processors i7-9700k versus i9-9900k. I myself am inclined to purchase i7-9700k, but suddenly I would be convinced by the results Or vs Ryzen 3600

[DustFireSky]

For all people who wondering about: How can a i5-6400 get so much UPS. He overclocked the CPU to 4,6Ghz with 1,4V and a bus speed of 169Mhz. He beat the crap out of it. A normal i5-6400 NEVER gets this UPS values with stock settings like he said!

He is just hiding the important information!

The picture is from himself in a conversation with him via PM.

i5-6400.jpg (435.01 KiB) Viewed 10084 times

Don't believe all what people are saying in the internet guys. Let them proof what they are saying.

[azesmbog]

Well then, in your opinion, the “correct” result is exactly at 3300 MHz:

3300.jpg (368.49 KiB) Viewed 10065 times

If you count the script, then not everything is so bad for the processor, which was released 5 years ago, and which, officially, is not accelerated by either a multiplier or a bus. Learn the materiel.

UPS avg - 69,00, max -81,90, min - 19,75
You can compare with the results above.
And this is the result at the standard turbo (?) Frequency.
And the result is higher - at my DAILY working frequency,
otherwise it can be overclocked to 5 GHz, but there will be no stability.
You can also show your results, I will only be happy
It is possible at the standard frequency, but it can be honest - exactly at 3300 MHz. For the correct comparison of processors for $ 180 and $ 600 (?)

Well, then, I don’t understand what the problem is ??
I showed how this megabase works on my current computer. And I'm going to buy a new processor, instead of an outdated one, so I choose between 9700K and 9900K. And I would like to see other results, all just. What kind of bloating holivar?
Everyone understands that the larger the base, the greater the load on the processor. With 10-20 FPS it is simply uncomfortable to play.
Now I can only look at my base, FPS dropped to 45-47, it’s already uncomfortable for me, so I choose a processor. Yes, the top processor for this one game.
For everything else, I have quite a few other computers.
I hope I cooled your ardor ??

[DustFireSky]

Don't get me wrong. I am glad for you with these values. My simple problem was, that u never said, that your machine is overclocked. That's all.

That would be the same if I say, i get 100UPS with an Ryzen 9 3900x but don't mention, that i have overclocked it to 6 GHZ. Just saying! All people would think, I use the cpu with stock settings. And that is misleading.

Enjoy the game.

Have a nice day.

[azesmbog]

OK. Accelerate your processor to 6 GHz, show the result, and then I will go to the store for the same.
And yes, it’s not forbidden to disperse, even the non-accelerated one ??
So there will be a result from you, well, at least a screenshot of this base?))
I already did two)

[hitzu]

What does this thing do?

Screenshot_756.jpg (348.38 KiB) Viewed 10061 times

[DustFireSky]

You know, that the I9 9900K beats the Ryzen 9 3900X in single core games. Enough techsites in the internet have proof that. If u only play singlecore games, buy this CPU. Only for just "gaming" u don't need a CPU with 12! Cores. And all things with the subject: "Multithreading", the I9 9900 would be a bad choice in comparison. Also this proofed enough techsites.

A little bit disappointing. I think the problem are the low CPU Clocks. It goes up and down. Always round about 4,2GHz+-. In good situations sometimes 4,4GHz. An interesting fact is the CPU Vcore! If I let the mainboard decide with the VCore auto setting, I get the same values as with my own fixed 1,2V settings before... So 0,2V are totally wasted energy for nothing.

Pic.png (6.22 MiB) Viewed 10055 times

[azesmbog]

Thanks for the screenshot.
As I understand it, if the frequency is fixed exactly at 3300 MHz - then the result will be even sadder ??
The forum here is about one single game, and it is for her that I choose a processor. Over the past many years, this is the only game I play. Well, even in the Spider Solitaire
You can also try to download my database (not the latest, but still)
https://drive.google.com/open?id=1XQ84O ... DmQFK_EZiu
without a screenshot, only numbers.
And think for yourself what choice I have left.

[DustFireSky]

"As I understand it, if the frequency is fixed exactly at 3300 MHz - then the result will be even sadder ??"

What u mean? Are u talking about your processor and the stock clocks? But yes, normally the game should run slower with fewer MHZ.

Edit:// Now I know what u are talking about... U have compared two different savegames man! U get more UPS with lower clocks, because u choosed a different savegame. Choose the right one and do it again.

"The forum here is about one single game, and it is for her that I choose a processor"

Yes, this single Forum. But not all people here just playing one game. I have many other games and I think many other people here too.

[azesmbog]

I suggested that you fix your processor at a frequency of 3300 and show the result.
And my result is for the current topic.
Well, I’ll put two results side by side, if so more clearly
On my current overclocked processor
UPS avg - 81.73, max - 96.58, min - 23.53
On my processor, but with a standard frequency of 3300 MHz
UPS avg - 69.00, max -81.90, min - 19.75
I emphasize - these are the results for the same base of the starter topic, but with different frequencies, with everything else being the same.
I see no reason to double-check your base, the result is predictable.
The line of the script for verification is higher. Believe it or not, if you prefer.
And yes, I quite believe and know that in many other games the result of AMD processors will be decent, but we are now talking about Factorio, right?)

[disentius]

The google link to your base is not working, I really wanted to benchmark it on my I5 9700:)

[azesmbog]

The google link to your base is not working, I really wanted to benchmark it on my I5 9700:)

The link is correct, now I tried to download - it downloads normally.
I am very, very, very interested in testing my base on the i7-9700, preferably with the letter "K" unlocked.
but it is possible and simply on 9700. maybe still i7 and not i5? typo?
If it’s interesting, I’ll upload a fresh database, but again on Google Drive, alas

and if the link of the author of the topic - then just copy all the letters )
или так
https://drive.google.com/open?id=1cAUBT ... kecKH-4MsC

[Stevetrov]

What does this thing do?

I assume you are asking about the line of rail signals and the combinator, its my smart stacker design, a way to load share trains across multiple stackers that maybe 1000s tiles apart.

Here is a post on reddit about them.