Is this the end of my playthrough?

[wobbycarly]

I had a lofty target of 5k SPM, did some planning, and figured out how much infrastructure I'd need. I prefer belts over robots, and figured that with all the recent optimisations, I should have a pretty good chance. I had planned to have central smelters for copper, iron and steel, then a couple of separate builds for plastic and rocket fuel/lube. Each science build would take in refined resources (copper/iron/steel/plastic/stone/etc), build circuits and other intermediates on site, then ship out by train to a centralised science area.

I knew I couldn't build it all in one go, so I planned out the smelters and built a column or 2, then would put together a science build before going back to the smelters. I ended up with all of the science built, and have it running at a reasonable rate, but with only about half the smelting, of course I can't feed my monster yet.

I know I didn't build specifically with UPS in mind, so I'm using nuclear to generate about 30GW (solar just takes sooooo long to build in huge amounts, so I tend to not bother with it) and I''m not maximising direct insertion options, neither am I doing belt->box->machine tricks. So of course, now I'm experiencing a slow death by UPS, and it swings between 40 and 50 with occasional drops even lower than that. I don't feel like redesigning my builds, but I don't want to abandon this base.

1) Without analysing my base - I wouldn't expect anyone to go through that painful task - does the community think that switching to solar may be enough to save this base?

2) Can anyone provide insight into where most of the processing is going, from the debug screenshot?

3) Is my computer just too old to get any more out of it?

4) Any other broad suggestions?

debug

2020-04-17 16_53_03-debug.png (229.75 KiB) Viewed 3602 times

science production

2020-04-17 16_54_16-Production.png (81.78 KiB) Viewed 3602 times

power

2020-04-17 17_27_31-PowerUsage.png (73.64 KiB) Viewed 3602 times

nuclear setup

2020-04-17 17_31_42-Starter Base and Nuclear.png (168.42 KiB) Viewed 3602 times

base overview

2020-04-17 17_28_49-Base Overview.png (280.84 KiB) Viewed 3602 times

system specs

2020-04-17 16_57_38-System.png (21.02 KiB) Viewed 3602 times

2020-04-17 16_52_38-Windows Task Manager1.png (64.39 KiB) Viewed 3602 times

2020-04-17 16_52_38-Windows Task Manager2.png (76.82 KiB) Viewed 3602 times

[JimBarracus]

how fast is your ram?
do you use dual channel?

can you post your savegame?

I guess its time to update to something with ddr4.

[wobbycarly]

Hi @JimBarracus, I can't answer the questions about the RAM - no idea.

I have a suspicion that the limitations are my system, though, not the game, as I also turned off steam/smoke, turned down the graphics detail, deleted all visible biters and cleared the map so that only areas covered by radar were redrawn.

Savegame: https://drive.google.com/open?id=1bPxKt ... i0YLTB0t5a

[JimBarracus]

Hi @JimBarracus, I can't answer the questions about the RAM - no idea.

I have a suspicion that the limitations are my system, though, not the game, as I also turned off steam/smoke, turned down the graphics detail, deleted all visible biters and cleared the map so that only areas covered by radar were redrawn.

Savegame: https://drive.google.com/open?id=1bPxKt ... i0YLTB0t5a

biggest bottleneck seems to be ram speed.
next thing is cpu single core performance
almost any graphic card can handle factorio and barely has an influence on performance
a ssd could also help

I know its a little cheaty but you might try an infinite power source.
In the end it doesnt matter if you plop down bazillions of solar pannels or simply use infinite power.

Nuclear power needs a lot of fluid calculations.

[mmmPI]

From the debug screenshot and the production tab i am under the impression that the differences between producing/not producing, is quite high.

transport line 8.9 / 7.1 / 15.6
entity updates 10 / 8 / 19
trains 2.0 / 1.4/ 9.2.

those seems to be what is taking the more time (in ms) for the game to process.
( which is good it's what is supposed to take long for the game)

if i remember correctly, those are average/minimum/maximum time in ms that the game need to process one particular aspect oof the game, can't remember how much time is taken in account for the average though.

Fluid/ heat / and electric network, all combined are not representing a whole lot in comparaison, which would be the case if the nuclear setup was the problem, i don't think you would gain that much switching to solar.

The quite significant spread of those number however is showing that during a period of time sometimes the games takes long to process, sometimes way less, the ups "spike".

The ratio Transport lines / entity is a bit too much favouring the transport lines to my personnal taste, that would be a consequences of having long stretch of belts and complicated balancer (semi-educated guess) vs more compact/efficient design for moving material around.

the train path finder stat is also a bit weird, from 0 to 6 ms for the game to process, reading this i expect long trains that don't move often, but when they do it's all at once, and goes through a "complicated" path finding ( with many different path to explore including "painful-for-ups loops").

to keep the 60 ups, one would require to have the whole update taking less than 16.6 ms versus +/-24 ms to 37 ms at the moment.

One thing that might reduce frustration would be to have a constant UPS, by reducing game speed based on the time period when the game require the most time to process, in that case 37ms would correspond to 27 UPS or so.

Using /c game.speed=0.5 would halve game speed and be close to those 27 ups, so the spike wouldn't be felt so much. You can double walking and/or mining speed using command to restore a decent character speed. it would make a somewhat "bullet-time" where everything is slowed except your character.

I would use this to modify the base in order to make it so that the production is more constant over time however if the spikes are due to the base not being finished yet, and is expected, when fully functionning, to always require 30+ms for a tick to be processed then there is no need for fiddling with the base design, instead you'd have to continue reducing game speed over and over.

One time i finished a base i had like 12 UPS; constantly, it wasn't great, but i was reducing game speed by small amount regularly thoughout the game, so it was manageable, balanced by the fact that i really wanted to finish the base i had spent so much time building.

I was relieved to start over a fresh map keeping in mind the limitations of my computer afterwards

Edit : https://mods.factorio.com/mod/DeleteEmptyChunks
This mod has a button to remove empty chunks, it will freeze the game badly at first, depending on the harsheness of the settings when considering what is to be deleted. But if you have revealed lots of areas that you don't plan to utilize and that aren't occupied by any player-placed-entity it could reduce significantly the size of savefile and the amount of RAM required to load/play.

[DerGraue]

1) Without analysing my base - I wouldn't expect anyone to go through that painful task - does the community think that switching to solar may be enough to save this base?

2) Can anyone provide insight into where most of the processing is going, from the debug screenshot?

3) Is my computer just too old to get any more out of it?

4) Any other broad suggestions?

I took a quick look at your save without mods and highest graphic settings. I have the same CPU, 8 GB Ram, 1,333 MHz.
I get around 40ish UPS, so your PC system in itself is not the problem.

Based on my personal experience with large maps and other ppls systems, I would estimate that if you would buy the fastest CPU and Ram that is out there, you can expect to approximately double your UPS. A faster PC is faster, sure. Upgrading stuff for your pc is probably not worth it, it is just too old.


But the main problem is that you build not UPS efficient, like you stated yourself. Especially all the splitters and balancers (the transportlines in the screenshot) take their toll, there are so many gaps on the belts, inserters don't go to sleep, those are some of the obvious things. There are many little things that you can do to improve stuff. But sometimes it is very hard to see them and even test them.

You can definitely "save your base", but you will have to redesign a lot. That's the challenge of the lategame.

Will switching to solar increase UPS? Sure, but that alone won't be enough. You can test it by cheating and replacing it with an infinite power source to test it.
You can also disable pollution, that will help a bit.

[wobbycarly]

Thanks everyone for your input and "care" over my poorly constructed base. I didn't think about pollution, I will try that. I didn't want to be too cheaty with infinity power, etc, and I'm sure it's inefficient builds (from a UPS point of view) that is the biggest problem. I'll keep fiddling with it and see how I go.

Cheers all!

[Serenity]

Somewhat lower UPS isn't the end of the world at that scale. If it dips a bit that's ok. It should just be consistent and not jump around all the time

But since you said you didn't build all of it, it's going to get pretty bad at the end

[wobbycarly]

Removing pollution stabilised UPS at around 48-50. That's playable. Thanks for the input. Guess it's a combination of older hardware and pushing the game limits.

[disentius]

Very organized base, I like!
When going that big, you have to do some UPS related optimizing of your builds:) first candidates are Smelting, green circuits, red circuits. ( mmmPI Had quite a few good remarks about that)
Some sidenotes:
- Traingrids are known to be bad for UPS, pathfinder is not optimized for choosing efficiently between equal(ish) paths.
- Minimize the use of splitters. They are very ups intensive. You can use an unload configuration like this saving 5 splitters per wagon

2 belt per wagon with buffer equal unload.png (722.41 KiB) Viewed 3520 times

- As mentioned: try to fully fill belt segments with stuff and make them as long as possible.
- Try to balance your science production to equal amounts. I see they "burst" in the prod history

On a side note: Windows 7 is not safe anymore, microsoft has stopped support. https://www.microsoft.com/en-gb/windows ... nformation
You can update to windows 10, your windows 7 licence is valid for upgrading. https://www.microsoft.com/en-us/softwar ... /windows10


Good luck

[wobbycarly]

I've had a chance to reread the feedback in a bit more detail. Really appreciate the time and effort people have gone to to help me out with suggestions.

From the debug screenshot and the production tab i am under the impression that the differences between producing/not producing, is quite high.

Correct - this is because the smelters are not yet complete.

Fluid/ heat / and electric network, all combined are not representing a whole lot in comparaison, which would be the case if the nuclear setup was the problem, i don't think you would gain that much switching to solar.

Good to know.

The ratio Transport lines / entity is a bit too much favouring the transport lines to my personnal taste, that would be a consequences of having long stretch of belts and complicated balancer (semi-educated guess) vs more compact/efficient design for moving material around.

Yes, with a couple of 32-32 balancers, plus an abundance of 16-16 (3-8 trains with 2 blue lanes out of each wagon) is probably overworking the game engine a bit.

the train path finder stat is also a bit weird, from 0 to 6 ms for the game to process, reading this i expect long trains that don't move often, but when they do it's all at once, and goes through a "complicated" path finding ( with many different path to explore including "painful-for-ups loops").

This is also true - the long pauses without production are partly due to insufficient throughput from the incomplete copper smelter. The train system is on a grid... no roundabouts, but plenty (too many!?) 4-way intersections. I had started with a one-way system, but then realised that the time spent to travel was going to cause a problem with rhroughput.

Edit : https://mods.factorio.com/mod/DeleteEmptyChunks
This mod has a button to remove empty chunks, it will freeze the game badly at first, depending on the harsheness of the settings when considering what is to be deleted. But if you have revealed lots of areas that you don't plan to utilize and that aren't occupied by any player-placed-entity it could reduce significantly the size of savefile and the amount of RAM required to load/play.

I did this with a command, and so the game only redrew the chunks that were covered by radar. I had made the "mistake" of "going West" to look for some larger ore sources, but turned out I didn't really need them... yet. However, this made minimal improvement.

Very organized base, I like!

Why, thank you kind sir/madam.

When going that big, you have to do some UPS related optimizing of your builds:) first candidates are Smelting, green circuits, red circuits. ( mmmPI Had quite a few good remarks about that)
Some sidenotes:
- Traingrids are known to be bad for UPS, pathfinder is not optimized for choosing efficiently between equal(ish) paths.

I thought roundabouts were "the worst" - after 2500 hours, it seems I still have quite a bit to learn! I have seen plenty of other players using a grid-like system. I usually have rail spaghetti, and wanted something a little more structured.

- Minimize the use of splitters. They are very ups intensive. You can use an unload configuration like this saving 5 splitters per wagon

2 belt per wagon with buffer equal unload.png

That's a cool layout that should still give a similar throughput to what I have - I'll give it a go. Given my number of stations, that might indeed make a difference. I still don't know a better way of ensuring even consumption without using balancers, though.

On a side note: Windows 7 is not safe anymore, microsoft has stopped support. https://www.microsoft.com/en-gb/windows ... nformation
You can update to windows 10, your windows 7 licence is valid for upgrading. https://www.microsoft.com/en-us/softwar ... /windows10


Good luck

I realise this, but I have a few devices that I know don't work on Win10, so I'm putting it off as long as possible so that I don't need to upgrade the devices, too.

[GrumpyJoe]

- Minimize the use of splitters. They are very ups intensive. You can use an unload configuration like this saving 5 splitters per wagon 2 belt per wagon with buffer equal unload.png

That's a cool layout that should still give a similar throughput to what I have - I'll give it a go. Given my number of stations, that might indeed make a difference. I still don't know a better way of ensuring even consumption without using balancers, though.

You might wanna have a look at Tuplex´s grid layout in his 2,5k +10k SPM base. Not for the grid, but for the non existing balancers
https://www.youtube.com/watch?v=9laKxRE ... slCNO8qRms

Basicly what he does is, that he takes in 8 cargo wagon trains, unloads into 8 seperate smelter/production lanes and loads them back into 8 cargo wagons. They never split, because the production is balanced.
There are two things you need to consider here:
1) you need overproduction. If you can´t equally devide the production into 8 lanes, you need to epxand a little. For example if you calculate 60 smelters, you put down 64, 6 per lane. Or if you have 4 wagon trains, devide by 4.
2) it would need some redesign of blocks, but if you devide them down to manageable blueprint sizes, it shouldn´t be that hard in the late game, with decent bot numbers and speed research.

Also, about the train grid: you dont really need ALL of it. You can use it to space out your production blocks so they dont get in the way of a rail, IF you´d need that piece of rail connection, but you dont have to build all of them.
When i first switched to bigger bases, i liked grids, because you dont need to think about what goes where, but for the last year or so i began to dislike them more and more.
I began to think about it the same way i think about bus builds. As every speedrunner (which im not) will happily tell you, a bus is easy, but inefficient.
A train grid to me is like a 100% bus base. Plus, since i watch alot of Youtube, ive seen it one too many times

I dont really like bots, but for mall production, i will always use them. Material request->assembler->provider chest. Since i play modded, i wouldn´t wanna belt a mall, even with Bob´s crazy inserter bends, because there are so many things to build, 6 different belt speeds and inserter colors, 4 different machine tiers, etc..

So for the next base, im going back to some "backbone" rail lines and split off where needed, i want train spagetti! (ofc, with LTN)

[gGeorg]

On a side note: Windows 7 is not safe anymore, microsoft has stopped support. https://www.microsoft.com/en-gb/windows ... nformation
You can update to windows 10, your windows 7 licence is valid for upgrading. https://www.microsoft.com/en-us/softwar ... /windows10

Win10 sends "metrics" to HQ. Win7 dont. You are probably kidding about safety.
More over, Win10 are regulary sent to unnconsions state after each bigger update, or might even eat your data (as it already did).
Win7 actualy are way safer and reliable.

[disentius]

Win10 sends "metrics" to HQ. Win7 dont. You are probably kidding about safety.
More over, Win10 are regulary sent to unnconsions state after each bigger update, or might even eat your data (as it already did).
Win7 actualy are way safer and reliable.

No security updates => no safety
No software/driver updates => no reliability

Yes, I hate the "Phone Home" features of Windows 10 too. So use something else (linux comes to mind) or disable the phone home features.