Performance optimization - post your saves

[Rseding91]

Ah that makes sense. Missed that one. Thx for the fast response and help

The issue is made worse by the most extreme case of over-signaling I have ever seen in your save. If you removed 90% of the signals it would probably drastically improve the performance.

[zer0t]

Ah that makes sense. Missed that one. Thx for the fast response and help

The issue is made worse by the most extreme case of over-signaling I have ever seen in your save. If you removed 90% of the signals it would probably drastically improve the performance.

Did I just win the over-signaling competition? 1st place
Anyway, thx again. And yes I noticed a performance drain previously when the trains were actually running. I used that many signals to improve throughput in/around factory and had the feeling that it worked. Ofc it is a nightmare regarding performance.

[boskid]

Did I just win the over-signaling competition? 1st place
Anyway, thx again. And yes I noticed a performance drain previously when the trains were actually running. I used that many signals to improve throughput in/around factory and had the feeling that it worked. Ofc it is a nightmare regarding performance.

I think i saw this save file already winning a similar competition when 1.1.0 was released, it was about transport belt groups performance because this save file had one insanely large transport line group inside.... So that would be another winning price in a category of "devs did not think someone would ever do that"

[mrvn]

Ah that makes sense. Missed that one. Thx for the fast response and help

The issue is made worse by the most extreme case of over-signaling I have ever seen in your save. If you removed 90% of the signals it would probably drastically improve the performance.

Did I just win the over-signaling competition? 1st place
Anyway, thx again. And yes I noticed a performance drain previously when the trains were actually running. I used that many signals to improve throughput in/around factory and had the feeling that it worked. Ofc it is a nightmare regarding performance.

Trains have a breaking distance that you can observe by signals switching from green to yellow in front of the train. That sets the minimum separation between trains. If your signals are that distance apart then throughput isn't limited by signals. If you have more than 2 yellow signals (outside of crossings obviously) then that's too many signals. If you have 2 and the train isn't about to cross the first one that's still more than you need.


Another way you can reduce train load is to make bigger trains. Make big ore haulers for example.

[zer0t]

I think i saw this save file already winning a similar competition when 1.1.0 was released, it was about transport belt groups performance because this save file had one insanely large transport line group inside.... So that would be another winning price in a category of "devs did not think someone would ever do that"

Yeah it was the same save file (older version). I'll try not to create another edge case with this save file

Trains have a breaking distance that you can observe by signals switching from green to yellow in front of the train. That sets the minimum separation between trains. If your signals are that distance apart then throughput isn't limited by signals. If you have more than 2 yellow signals (outside of crossings obviously) then that's too many signals. If you have 2 and the train isn't about to cross the first one that's still more than you need.


Another way you can reduce train load is to make bigger trains. Make big ore haulers for example.

Thx for input. I'll check this as soon as the factory is running again

[ganz_entspannt]

My first post about my favorite game. I love it, but ... is my PC too slow?

My endless Map with many Robots, Rail Signals and Trains ... one try without a Belt:
https://drive.google.com/file/d/1ju-PMU ... sp=sharing
How many fps do you have?

[Rseding91]

My first post about my favorite game. I love it, but ... is my PC too slow?

My endless Map with many Robots, Rail Signals and Trains ... one try without a Belt:
https://drive.google.com/file/d/1ju-PMU ... sp=sharing
How many fps do you have?

Just barely 60 FPS/UPS.

[ganz_entspannt]

OK thanks. I have 60fps too, but for me it regularly drops to 35fps. Don't you have any break-ins?

It bothers me. I'm afraid I've reached a limit. I wanted to expand even further.

Is a faster CPU worth it (currently AMD 7 5800X3D)? Is it a fast Pentium or are there planned optimizations to increase performance. This score is hardly any fun anymore.

Sorry for my bad English, Google Translate.

[quyxkh]

Bots, trains and biters is what's killing your performance. You're using bots to carry ores long (for bots) distances to trains that run on congested routes to depots that use bots to carry them even longer distances to the smelters. You're averaging well over forty thousand bots in the air. Cut the bot route lengths and train traffic *dramatically* with longer trains serving closer miners and smelters either on site or anyway well outside the base core. Ship the plates in. You're in memory-constrained territory, this base is pushing your memory bus as hard as it'll go. In this territory a faster cpu won't help. Faster memory would help some, going from 3200 to 3600 with the spendy low-latency ram if you haven't already.

[ganz_entspannt]

Maybe a peaceful run without biters would have longer performance. I didn't even think about the memory. I have F4-3600c16 G.Skill. Switch to DDR5? Then this idea of a base is at its limit.

Thanks for this idea. I was really wondering about the performance without having anything to compare it to. That was the reason for coming to the forum. Thank you.

[Atraps003]

Here is a map of landmines and combat robots belonging to opposite forces. Nothing happens but the lag is intense.

viewtopic.php?f=23&t=111259

[boskid]

Here is a map of landmines and combat robots belonging to opposite forces. Nothing happens but the lag is intense.

viewtopic.php?f=23&t=111259

Such synthetic save files are not interesting, they do not represent a real use cases. It just says "land mines are slow" and i know that land mines are slow. Land mines are updating when there are enemies around and in this case for obvious reasons you have enemies around (the combat robots), so basically all land mines are kept active. Every active landmine does an entity search around itself every 10 ticks to find if the enemy (that made it active or anything else) is close enough to cause land mine to blow off. This save file is literally a lot of entities doing slow things each. I cannot reduce amount of entities since that is controlled by a player and i cannot make it not do its slow thing because it would not blow off causing them useless.

The only possible optimizations possible to apply around "when enemy is around" logic were already applied for 2.0 but your save file is unlikely to benefit from that because there are combat robots almost everywhere. This case is just bad because those combat robots make land mines active while not being able to trigger them to blow off because of collision mask condition.

[derpumu]

Here is an older 4k SPM save of mine - I had huge performance problems with my old potato, but even with a new high end PC it's not too great.

https://derpumu.de/factorio/megabase202 ... 1.1_EE.zip

- idling, it is just about 60 fps, but seems close to dropping below that.
- research artillery range (I added Editor Extensions to the save to easily test that), watch the fireworks, and fps drops to 40 over the next minutes as biters pack their stuff to start the long treck towards my walls