Factorio Forums

Here is my 2nd and last try but it's... still bad. Mainly trains, no bots, some belts, smelting off-patch. Smelting is done through a looping train with mixed wagons content.
I was told trains would magically improve UPS. Lies! Deception!

• Trains only

• No belts

• No cars

• On-patch smelting

A note: i was just so proud of myself when I figured out how to lose the internal buffer chests in my layout. Learned two things: (1) it's _slower_, and (2) long stretches of open belt seem are not negligibly cheap, replace them with neathie pairs for a noticeable bump. I won't make DaveMcW time the map, it just replaces the production cells with this: So it would appear the difference is actually in _train pathfinding_: that stacker? I take it out and the performance difference goes away.

56120.978 ms - Stevetrov trains 2 primed
63024.836 ms - flame_Sla v10 - train-sleeping cars
183328.220 ms - schmetterling train

Stevetrov is now the overall leader!
Stevetrov also takes the lead in "trains only" category with the same entry.

I will sadly be travelling during the next week, and thus not be able to take back the lead if not for this:
I actually forgot to send the cars to sleep in my map maybe it's enough to get back into leading position

54215.92 ms - Allaizn circuit.everything.3.sleepycars

Allaizn is back in the lead!

This thing went above my pay (and patience) grade SO fast!

But I'm (ab)using this competition so hard and gobble up all the techniques for my next 60+UPS base attempt - I have learned so much here! Thanks <3

I managed to build another variant - 70 miner/furnaces per patch and thus only 4.5 patches used. My personal benchmark showed that it was slightly worse than the other map, but I'm not quite sure why ¯\_(ツ)_/¯
A possible experiment here is to use 2 locomotives per train (or even twoheaded trains?), though I doube it will save much considering how super low the train update is in the first place.

50649.142 ms - Allaizn circuit.everything.4

Oh.my.god. They're _clocking the output inserter on a boosted greenchip assembler_. And it's working.

I trust at least some people besides just me will get a laugh at how chagrin-inducing this is for me. At least I know why I didn't think of it myself.

I'm not going to try for Allaizn's mark, I choose to believe circuit-controlled-everything is not the province of cowboys like me. So here's my tuned.2 map but with *clocked green chip unloads*. Still got that mind=blown feeling here.

57669.989 ms - quyxkh mostly.belts.tuned.3

I'll see if I decide to try and improve mine. At least I'll probably win the "longest trains" category .

Ok, I'm turning in my map so I can at least take second place while I can.

I tried to maximize direct insertion as much as possible for this build, which ended up being 6 beacons on the assembling machines. I didn't expect it to perform so well, but apparently it's pretty competitive. On my machine, it comes in second to only allaizn.circuit.everything.4. I haven't clocked everything yet as there are some inserters which seem to swing near peak efficiency already, though I'm sure there's something I've missed.

54617.979 ms - Syhn v1.13

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

As we are a few hours away from half way, could we have a leader board for the various categories? Maybe a Google doc that we can all view?

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.

We are at the halfway point, thanks to everyone who participated so far!

Here is the current leaderboard:

Overall - 50649.142 ms - Allaizn
No cars - 54617.979 ms - Shyn
Belts only - 54617.979 ms - Shyn
Trains only - 56120.978 ms - Stevetrov
Off-patch smelting - 66104.114 ms - Stevetrov

So, these aren't enough to vault me onto the leaderboards but I think they'll, uhh, carry their freight here.
So I was thinking, why chase after every scrap of production and then try to control delivery with production capacity? Why not just dispatch trains to pick up exactly as much as you need? That's how real maps do it.

These clock the pickup trains from a central dispatcher to maintain a pick up rate of 100000 green chips/minute.

The pickup.trains is a straight conversion of my tuned.2 map, dispatching a train every 2304 ticks, that's one every 64/100th of a minute. Any performance difference from tuned.2 is down in the noise on my rig.

The all.trains map is … well, here's a movie. It gets a bit better performance than tuned.2 on my rig, and it does it with cs12 smelters. I could clock the copper-pickup dispatch as well but haven't bothered as this seems to be working well. It's dispatching trains to pick up 4000 iron, that's enough to produce 5600 chips, 5.6*8/100*3600 is 1612.8 ticks, so with three dispatch stops that's a 4838 tick dispatch clock firing three trains at intervals. I'm using stevetrov's lovely production cells with some minor tweaks. There are only two train schedules, +cu -cu and +fe -fe+gc -gc.

Nice work quyxkh! That last trains one looks really good in the video. Clever idea to clock the trains to pickup at the right time instead of doing the factories.

I wonder if there's potential additional UPS/train savings to be had by having the train loop go like this:

• Copper Pickup
• Copper Dropoff
• Iron Pickup
• Iron Dropoff/GC Pickup
• GC Dropoff

56404.169 ms - quyxkh clocked.dispatch.pickup.trains
57041.624 ms - quyxkh clocked.dispatch.all.trains