Factorio Forums

I was attempting to do a belt-only direct insertion build, crafting the circuits on top of the copper patches. I figured it probably wouldn't be very good but I gave it a try. I optimized a single cell and copypasted it across the patch, but I figured I wouldn't have to make it compact and tileable, and I fit 6 per patch. Turns out 13 patches of this was 48-49k/min so I would've needed considerably more. In retrospect, I really should have tested what I got with just one patch.

I got around to running all the 1M scale maps, here's the results.

50 runs, 1k ticks per run (train maps beware)

While I'm happy to see my map up there like that, and your ratios are much closer matches to the ones I get, I have to caution that the all.trains map in particular is bursty, and its first couple of minutes are in one of the ebbs, it's producing about 95-96k/min chips for that interval.

Stevetrov wrote: ↑

Tue Nov 05, 2019 6:51 am

mrvn wrote: ↑

Mon Nov 04, 2019 2:30 pm

Has anyone considered having a train long enough to be loaded by multiple patches? Load e.g. 4 wagons per patch then advance the train by 4 wagons and repeat till it's all full. Or have lots of locomotives to fill the gap between patches or simply leave cars empty.

I do not think this would help. This technique would reduce the total number of trains at the expense of more rolling stock. As the biggest UPS hit from trains is for the collision checks on the rolling stock this is likely to be a bad trade off.

Although it would look fairly cool!

To bad a train collides with itself or you could reduce the number of trains to one and avoid it.

mrvn wrote: ↑

Thu Nov 07, 2019 10:41 am

Stevetrov wrote: ↑

Tue Nov 05, 2019 6:51 am

mrvn wrote: ↑

Mon Nov 04, 2019 2:30 pm

Has anyone considered having a train long enough to be loaded by multiple patches? Load e.g. 4 wagons per patch then advance the train by 4 wagons and repeat till it's all full. Or have lots of locomotives to fill the gap between patches or simply leave cars empty.

I do not think this would help. This technique would reduce the total number of trains at the expense of more rolling stock. As the biggest UPS hit from trains is for the collision checks on the rolling stock this is likely to be a bad trade off.

Although it would look fairly cool!

To bad a train collides with itself or you could reduce the number of trains to one and avoid it.

That would be a nice optimization. You can at least have your trains move mostly along cardinal directions to make the collision checks less expensive.

mulark wrote: ↑

Thu Nov 07, 2019 1:09 am

I got around to running all the 1M scale maps, here's the results.

50 runs, 1k ticks per run (train maps beware)

26x.png

Woohoo! I only made it to second-worst!

I have learned so damn much from this competition, and am in the slowest of progressions towards my next mega base built on those discoveries, but I just don't have the time to also fit more iterations in here despite having a few ideas. This kills me!! (But plesae keep it coming. There is so much value in this!)

MEOWMI wrote: ↑

Wed Nov 06, 2019 10:47 pm

I was attempting to do a belt-only direct insertion build, crafting the circuits on top of the copper patches. I figured it probably wouldn't be very good but I gave it a try. I optimized a single cell and copypasted it across the patch, but I figured I wouldn't have to make it compact and tileable, and I fit 6 per patch. Turns out 13 patches of this was 48-49k/min so I would've needed considerably more. In retrospect, I really should have tested what I got with just one patch.

This is actually I good concept to use and its the basis of AssemblyStorms (Slyns) map. The trick is to balance the number of patches required versus the efficiency of the build.

I have added 1M variants of the two submissions since my initial posting. Testing TBA.

https://drive.google.com/drive/u/0/fold ... kXmbMQPZ4b

Belts only no cars no trains no bots.

BTW when does the competition end exactly?

50555.739 ms - Stevetrov submission 4

Stevetrov is now the leader in "belts only" category!

For announcing winners, I will only use saves submitted before 21:00 UTC on November 11.

61252.578 ms - DaveMcW belts only

I tried an 11-beacon assembling machine design, but the results were not great.

DaveMcW you were wondering whether anyone could beat your map?? Well technically I think this does, but as its 95% your map, it doesnt really feel like it should count!!!

I reduced the beacon count on the iron smelters slightly by shifting everything up 2 tiles (crashing lots of trains) and shifting around some beacons and miners.

39582.092 ms - Stevetrov trains only less bacon
40852.904 ms - DaveMcW trains only

Stevetrov (& DaveMcW) is now the overall leader!
The same entry is also the leader in "trains only" category.

I benchmarked my original save at the same time for comparison. I'm not sure why the numbers are slightly higher, but Stevetrov's improvements consistently save 1.3 seconds.

This one is my take on the hybrid design that quyxkh and Allaizn have used. On site smelting with plates belted to central production cell and GC trained to drop off but NO CARS!!

Nb for those that are paying attention, I will try and finish submission 5 later, it needs a tweek.

This one is based on the concept from Dave's train only map, but the bps are my own design. I went for a higher beacon count on my copper smelters than Dave, that means that my smelters / assemblers have slightly better UPS, but that I need more patches and more locos, unfortunately this seems to out-weigh the gains on the smelters / assemblers.

I used a signalled layout that means I have a higher train path finding cost but it still accounts for ~ 0.1% of total update.

Dave's copper smelters / assemblers and my iron smelters produces the best result, see above.

40822.041 ms - Stevetrov submission 5
46697.022 ms - Stevetrov submission 6

I re-ran the benchmark for the top two in each category to confirm the results. Here are the winning scores:

Trains only
39497 ms - Stevetrov & DaveMcW
40822 ms - Stevetrov

Belts only
50828 ms - Stevetrov
51754 ms - AssemblyStorm (Shyn)

Off patch smelting
66223 ms - Stevetrov
70821 ms - flame_Sla

Thanks to everyone who participated!
I will not close the thread, feel free to keep posting new designs and improvements.

DaveMcW wrote: ↑

Tue Nov 05, 2019 9:31 am

39991.665 ms - DaveMcW trains only

I don't have access to factorio for the coming weeks, so I can't check the save - but I did have a question. How do you control the trains? Regular signals, or circuit timing?

And did you consider having a dedicated rail for each production line so you don't need any junctions / signals / etc?

vanatteveldt wrote: ↑

Tue Nov 12, 2019 9:24 am

DaveMcW wrote: ↑

Tue Nov 05, 2019 9:31 am

39991.665 ms - DaveMcW trains only

I don't have access to factorio for the coming weeks, so I can't check the save - but I did have a question. How do you control the trains? Regular signals, or circuit timing?

And did you consider having a dedicated rail for each production line so you don't need any junctions / signals / etc?

One could set it up with no signals at all and not circuit controlled. "Simply" place the trains and fill them to different amount so they drive around at different times. It only has to work for 100k -1M circuit boards so shouldn't be too hard.

I got around to updating the 1M maps.

3 runs, 3k ticks per run.