UPS friendly green chips?

[AngledLuffa]

What is a UPS efficient layout for green chips? Either bots or belts would be welcome. I've been putting 2 wire assemblers and 2 green chip assemblers in a 4x8 rectangle of beacons. The wire assemblers are fast enough to keep up with the chips, and the chip assemblers are 5.5 speed (+340%). I think there must be a better layout, though, as a normal box of 12 beacons with 4 prod modules gives a speed of 8. I was guessing that a 12 beacon layout would actually be worse for UPS, though, as then the wire would either need bots or belts for transport.

[DaveMcW]

You can fit 11 beacons with direct insertion.

11-beacons.jpg (232.88 KiB) Viewed 6170 times

[AngledLuffa]

Very nice, thank you!

[AngledLuffa]

From a UPS perspective, do you know which of the following three are best?

- green chip assembler occasionally turns off because of not enough wire (the image shown)
- replace prod module with speed in the wire assembler
- remove one beacon near the green chip assembler

The green chip assembler has a speed of 7.375, but its effective speed is 6.883 because of how much wire it gets.

[Jap2.0]

From a UPS perspective, do you know which of the following three are best?

- green chip assembler occasionally turns off because of not enough wire (the image shown)
- replace prod module with speed in the wire assembler
- remove one beacon near the green chip assembler

The green chip assembler has a speed of 7.375, but its effective speed is 6.883 because of how much wire it gets.

The type of module should not affect UPS, I believe entities sleeping is pretty good (it's definitely not worse), and I think beacons have next to no UPS impact as well.

In conclusion, to the best of my knowledge, whichever produces the most green circuits.*

*But it's more complicated because replacing productivity with speed would increase your resource costs and hence the scale of your mining, smelting, and transportation infrastructure...

[AngledLuffa]

Indeed, I was wondering whether it is worth it to get full power out of the green chip assembler at the cost of flying around 20% more copper plates

[disentius]

Good excuse to finally try out factorio benchmarking...
Tested 2 setups. both built on the same empty map with the same settings.
They produce the same (12.5K per minute)

Setup 1:
-fully beaconed
-Copper wire with belts

12k green-hybrid-setup.png (2.23 MiB) Viewed 6091 times

Results:

Performed 600000 updates in 48127.782 ms
avg: 0.080 ms, min: 0.061 ms, max: 0.348 ms
checksum: 1649213351
59.183 Goodbye

Performed 600000 updates in 48264.212 ms
avg: 0.080 ms, min: 0.061 ms, max: 0.386 ms
checksum: 1649213351
59.116 Goodbye

Performed 600000 updates in 48209.047 ms
avg: 0.080 ms, min: 0.061 ms, max: 0.379 ms
checksum: 1649213351
60.901 Goodbye


Setup 2:
-fully beaconed
-full robot setup

12k green-robot-setup.png (2.91 MiB) Viewed 6090 times

Result:

Performed 600000 updates in 50863.923 ms
avg: 0.085 ms, min: 0.063 ms, max: 0.328 ms
checksum: 1216465678
61.699 Goodbye

Performed 600000 updates in 50546.559 ms
avg: 0.084 ms, min: 0.063 ms, max: 0.323 ms
checksum: 1216465678
61.607 Goodbye

Performed 600000 updates in 50693.630 ms
avg: 0.084 ms, min: 0.063 ms, max: 0.315 ms
checksum: 1216465678
61.564 Goodbye


System setup:
0.009 System info: [CPU: Intel(R) Core(TM) i5-9600K CPU @ 3.70GHz, 6 cores, RAM: 4830/16295 MB, page: 6845/18727 MB, virtual: 4252/134217727 MB, extended virtual: 0 MB]
0.010 Display options: [FullScreen: 0] [VSync: 0] [UIScale: system (100.0%)] [MultiSampling: OFF] [Screen: 255] [Lang: en]
0.016 Available display adapters: 3
0.016 [0]: \\.\DISPLAY1 - AMD Radeon R9 200 Series {0x80005, [0,0], 2560x1440, 32bit, 60Hz}

[AngledLuffa]

Neat! Can I ask what you use to do this testing? I would gladly stress test some of my own layouts if I had a good setup for it. I was thinking that god mode wouldn't quite work as it would easily do 60ups the entire way, but it seems you are not going that fast. How did you accomplish that without spamming the same layout over and over?

My initial suspicion is that both a direct insertion bot layout and a direct insertion belt layout will do better than the two layouts you showed, but I'd love to experiment on that myself.

[Jap2.0]

Neat! Can I ask what you use to do this testing? I would gladly stress test some of my own layouts if I had a good setup for it. I was thinking that god mode wouldn't quite work as it would easily do 60ups the entire way, but it seems you are not going that fast. How did you accomplish that without spamming the same layout over and over?

My initial suspicion is that both a direct insertion bot layout and a direct insertion belt layout will do better than the two layouts you showed, but I'd love to experiment on that myself.

I'm not entirely sure if it's what he used, but there's some command line stuff you can do.

[AngledLuffa]

So this should work?

- build a design in sandbox mode, save it
- run from the command line with -benchmark FILE -benchmark-ticks N

[disentius]

@Jap 2.0: yep.

Like this:(windows)

"D:\games\Factorio_16_standalone\bin\x64\Factorio.exe" --mod-directory "D:\games\Factorio_16_standalone\mods" --disable-audio --benchmark "bench_save.zip" --benchmark-ticks 600000 >> "D:\games\Factorio_16_standalone\benchmark.log" 2>&1

Last part generates output logfile in the main factorio directory.

[disentius]

direct insertion bot layout:
(Same throughput greens)

Performed 600000 updates in 40437.798 ms
avg: 0.067 ms, min: 0.047 ms, max: 0.283 ms
checksum: 4096971926
51.665 Goodbye

Performed 600000 updates in 40102.277 ms
avg: 0.067 ms, min: 0.047 ms, max: 0.353 ms
checksum: 4096971926
51.095 Goodbye

Performed 600000 updates in 40182.177 ms
avg: 0.067 ms, min: 0.047 ms, max: 0.320 ms
checksum: 4096971926
51.991 Goodbye

12k green-robot DI.png (925.02 KiB) Viewed 6033 times

Looks like your instinct is backed up by numbers, AngledLuffa

[disentius]

No need for sandbox mode. Edit config.ini in your factorio testinstallation freeplay directory. enables you to run a normal freeplay game with the tools you need.
You'll find config ini here:
You can also use Creative Mode mod, but that one has ups impact on its own.

[your factorio install directory]\data\base\scenarios\freeplay
Make a backup of the original before editing.
More info on the factorio api page and the Wiki.

I use this one to build test setups:

spoiler

Code: Select all

local silo_script = require("silo-script")
local version = 1

script.on_event(defines.events.on_player_created, function(event)
  local player = game.players[event.player_index]
  local character = player.character
  character.character_inventory_slots_bonus=150
  character.quickbar_count_bonus=2
  character.character_crafting_speed_modifier=100
  character.character_running_speed_modifier=2
  character.character_build_distance_bonus=500
  character.character_reach_distance_bonus=500
  character.character_item_drop_distance_bonus=500
  character.character_resource_reach_distance_bonus=500
  character.character_item_pickup_distance_bonus=5
  character.character_logistic_slot_count_bonus=20
  character.character_trash_slot_count_bonus=20
  player.force.research_all_technologies()
  player.cheat_mode=true
  player.surface.always_day=true 
  player.insert{name="power-armor-mk2"}
  player.insert{name="infinity-chest", count=20}
  player.insert{name="construction-robot", count=150}
  local armor = player.get_inventory(defines.inventory.player_armor).find_item_stack('power-armor-mk2')
  armor.grid.put{name='fusion-reactor-equipment'}
  armor.grid.put{name='fusion-reactor-equipment'}
  armor.grid.put{name='fusion-reactor-equipment'}
  armor.grid.put{name='fusion-reactor-equipment'}
  armor.grid.put{name='personal-roboport-mk2-equipment'}
  armor.grid.put{name='personal-roboport-mk2-equipment'}
  armor.grid.put{name='personal-roboport-mk2-equipment'}
  armor.grid.put{name='personal-roboport-mk2-equipment'}
  armor.grid.put{name='personal-roboport-mk2-equipment'}
  armor.grid.put{name="battery-mk2-equipment"}
  armor.grid.put{name="battery-mk2-equipment"}
  armor.grid.put{name="battery-mk2-equipment"}
  armor.grid.put{name="battery-mk2-equipment"}
  armor.grid.put{name="battery-mk2-equipment"}
  armor.grid.put{name="battery-mk2-equipment"}
  
  for name, recipe in pairs(player.force.recipes) do
	recipe.enabled = true
  end
   
  player.force.chart(player.surface, {{player.position.x - 1000, player.position.y - 1000}, {player.position.x + 1000, player.position.y + 1000}})
  if (#game.players <= 1) then
    game.show_message_dialog{text = {"msg-intro"}}
  else
    player.print({"msg-intro"})
  end
  silo_script.on_player_created(event)
end)

script.on_event(defines.events.on_player_respawned, function(event)
  local player = game.players[event.player_index]
  local character = player.character
  character.character_inventory_slots_bonus=150
  character.quickbar_count_bonus=1
  character.character_crafting_speed_modifier=10
  character.character_running_speed_modifier=2
  character.character_build_distance_bonus=50
  character.character_reach_distance_bonus=50
  character.character_item_drop_distance_bonus=50
  character.character_resource_reach_distance_bonus=50
  character.character_item_pickup_distance_bonus=50
  character.character_logistic_slot_count_bonus=20
  character.character_trash_slot_count_bonus=20
  
end)

script.on_event(defines.events.on_gui_click, function(event)
  silo_script.on_gui_click(event)
end)

script.on_init(function()
  global.version = version
  silo_script.on_init()
end)

script.on_event(defines.events.on_rocket_launched, function(event)
  silo_script.on_rocket_launched(event)
end)

script.on_configuration_changed(function(event)
  if global.version ~= version then
    global.version = version
  end
  silo_script.on_configuration_changed(event)
end)

silo_script.add_remote_interface()
silo_script.add_commands()

[Optera]

So as quick summary:
All bots: avg: 0.084 ms
Wire belts: avg: 0.080 ms
direct insertion: 0.067 ms

Direct Insertion having the best performance was to be expected, but seeing belted wires perform better than bots is a surprise.

PS you can use Creative Items, which only makes use of existing hidden items and should have no noticeable performance impact.

[disentius]

Yes, it looks that way.
However, the test was small scale, and i did not try different orientations. Still lots to be tested.
Optera means this Mod
It conveniently avoids the hassle of editing config.ini.

[AngledLuffa]

I've been trying some profiling experiments with all of these suggestions, and it's been very interesting so far. Without spamming too many pictures, I find that a simple change like putting a lane balancer at the end of a column of furnaces leads to a big improvement. Presumably the belt doesn't get those occasional gaps which throw off the compressed belt optimizations. I'll keep experimenting with other things (next up is testing my blue chip factories and structure factories, belt vs bot). If I come up with a green chip layout which blows away the layouts presented here, I'll report back, of course. In the meantime, thanks for all the help!

[AngledLuffa]

I built a couple blue chip factories to compare, and the results are pretty biased in favor of bots.

The bot based layout ran 1,000,000 ticks in 49s, and the belt based layout took 147s. I would say that there is probably some optimization to be done in both of the layouts, but in some ways they're about equal - they both have green chip assemblers that run at 5.5 speed, for example.

You can see that the resources come in via infinity chests, and the resulting chips go into an infinity chest that eats everything. Fluids are produced infinitely by putting infinite barrels of that fluid into an assembler which pipes directly into a tank. I figured this was close enough to simulating getting fluids brought in by train

How do I put these images in spoilers?

Edit: one thing I did do on the bot based version was put the inserters on timers, but I doubt that would have that large of an impact. I suppose I could test that as well.

clickme

clickme

[disentius]

Like this:

Code: Select all

[spoiler=clickme][attachment=0]12k green-robot.png[/attachment][/spoiler]

Result:

clickme

12k green-robot.png (377.29 KiB) Viewed 5932 times

Looks promising, your testing. Nice builds!