A Case for Balancing Nuclear Power

[PunPun]

Practically, how much RAM would one need to generate ALL map chunks ?

If each tile is an int(max 2147483648 diffrent tiles) describing what tile it is then 2m*2m*4bytes=14,6terabytes. Or if they are packed in one byte(max of 255 different types of tiles) then 3,6 terabytes. Just for the tiles. Add to that all the potential biters+fish+ores+cliffs+trees+bushes+whatnots. I would say somewhere around 25 terabytes or(15 terabytes if we limit to 255 different types of tiles). It isn't that astronomical of a number. But if we tried to utilize all that space for optimum science per minute production we would be limited by the fact that a processor can access only so much data at a given time.

But anyway if you build your reactors with minimun amount of pipes and only so many reactors as you need then I doubt they would ever take more than 1% of the overall cpu time from the whole game. You would be much better off optimizing your railnetworks by removing all unneeded tracks and your oilrefineries by minimizing the amount of pipes and tanks and so on. Propably much more gains there than in the nuclear reactors.

[Adamo]

I took the time to build a small model of the nuclear fuel cycle and recreate the nuclear values and nuclear system using this more-realistic model. It essentially uses real-world statistics except that they're sped up by about, oh, 1000 times, so that you're not waiting weeks for your fuel cells to burn down even by the 1% or so that gets burned per cycle. The full mod implements plutonium breeding and recycling, and includes a "light water reactor" boiler, while using the vanilla reactor as the liquid-metal-cooled reactor, with the potential to completely burn all uranium fuel, but, this would take a very, very long time.

To balance this, I had to reimplement the efficiencies in turbines and boilers, but I kept all the vanilla ratios (like 2 steam engines per boiler) under the new numbers. I had to change the temperatures to normalize those changes (and in fact the temperatures are more realistic), and increase fuel values accordingly (which also turned out to be more realistic when you start to compute things like how much mass one unit of solid fuel would need to be to normalize with the fuel capacity value of 1 unit of water). I reduced solar power to 5.1kW, since this is about (and really still more than) the most power anyone should ever be able to get from a solar panel this size.

I did this because I am a physicist, so I was uniquely qualified to model the nuclear reactions. I don't have these mods listed on the portal, but if you're interested in a balancing of nuclear power, you may want to take a look. Let me know!

[Oktokolo]

The big O notation does not indicate time at all. It describes how something scales with input. The first thing you learn in programming courses that talk about the big O notation is that it is not to be used to decide what algorithm to use. It can be used as a hint on what could be useful but the final decision needs to be made by actually benchmarking with a realistic use case for the project.

Players and devs have benched solar and nuclear excessively already and the devs also explicitly stated, that the O(1) complexity of solar is the very reason for it to always performing better than any working nuclear setup, because for nuclear N is the amount of fluid segments to simulate and simulating one nuclear setup worth of them seem still to be much more costly than simulating solar.
Complexity matters a lot if it differentiates between "works fine at megabase scale" and "eats tenths of UPS at megabase scale" - if you want a megabase at least.

[Selvek]

I ran my 600SPM "mini-mega-base" on nuclear with zero UPS issues. The thought of running around placing enough solar panels to compete was horrifying.

How is that possible? Do you have PC Gamer Magazine's large pixel collider?

Nope. I have a stock Dell that I bought maybe 4 years ago for around $700.

As for why I had no UPS issues... well, I have no idea. I know there are plenty of people with bases much "mega-er" than 600SPS, but I had an 8-reactor nuclear blueprint that I kept stamping down (at least a dozen times, although I haven't opened that save in a while), and it always worked fine.

Do you mean 600 science packs total per minute, 600 science packs each per minute or actual 600 science packs each per second? I don't think anyone has build a 600 SPS mega factory yet. Or at least not one that runs at 60 UPS. That would be 36000 SPM and a rocket every 3 seconds.

600 of each science pack per minute. Whoops, definitely not 600SPS

[PunPun]

Players and devs have benched solar and nuclear excessively already and the devs also explicitly stated, that the O(1) complexity of solar is the very reason for it to always performing better than any working nuclear setup, because for nuclear N is the amount of fluid segments to simulate and simulating one nuclear setup worth of them seem still to be much more costly than simulating solar.
Complexity matters a lot if it differentiates between "works fine at megabase scale" and "eats tenths of UPS at megabase scale" - if you want a megabase at least.

Does not change the fact that the big O notation is not a reliable indicator on what algorithm will actually be the best for project. I'm not saying that a nuclear reactor does not use more cpu resources than solar. What I'm trying to get across that even if nuclear does use more cpu resources because solar requires a lot bigger footprint and more objects that need to be placed in the world as well as the difference in building resources, depending on how long you are going to play the save in realtime after building them going nuclear might actually give you more total research.

I have seen a lot of benchmarks comparing ups between solar and nuclear but not a single one where they account for the time to mine resources and crafting components for building them and actually placing them to the world(If there is one then please link it as it would be really intresting to read). Because in a realistic use case aka actually playing the game those actually matter. The big O notation only gives us a clue. The longer a reactor or an equal power production worth of solar exists then the more likely solar is going to win but it does not give us any clue at all on how long it actually takes.

[BlueTemplar]

I guess that this kind of benchmark, once you try to go farther than computing time/energy to craft the components, would be highly subjective ?

[Oktokolo]

What I'm trying to get across that even if nuclear does use more cpu resources because solar requires a lot bigger footprint and more objects that need to be placed in the world as well as the difference in building resources, depending on how long you are going to play the save in realtime after building them going nuclear might actually give you more total research.

I seem to have completely missed the topic. If we are not discussing megabases here, complexity really does not matter at all as UPS will be at constant 60 whether you go nuclear or solar.
In that case, the metric to consider is what makes more fun to build, maintain, and especially look at. And for that well-defined and purely objective metric, nuclear obviously always wins over solar because of its pretty pipe works!

[mrvn]

Practically, how much RAM would one need to generate ALL map chunks ?

Try running this on a new map:

/c local radius=1000000
game.player.force.chart(game.player.surface, {{x = -radius, y = -radius}, {x = radius, y = radius}})

[mrvn]

Sigh. First off the map is nowhere near endless. The max size is infact 2000000x2000000 tiles.

And you (and the wiki) are wrong. The map coordinates is a 32bit signed integer and goes from -2147483648 to 2147483647. The default map generator only generates chunk within 1000000 tiles from the origin and everything outside is out-of-map tiles. So when you walk/drive in a direction you will stop at 1000000. If you enter a plane or teleport you can get much further. And mods can catch the on_chunk_generated event and change the tiles from out-of-map to anything else.

So yes, vanilla has max 2000000x2000000 tiles (plus some out-of-map around that) but mods can have more. The game engine supports up to 18446744073709551616 tiles. The game estimates needing 18939904 Exabyte for that. Or 75759616 times the memory an amd64 CPU can address (48bit virtual address). Given the limit on the virtual address the maximum map size would be 2877268 * 2877268 tiles or for a ribbon world a width of about 1600 tiles and full length. Not that you could build much on that anymore since all the memory would be used for map chunks already.

[PunPun]

And you (and the wiki) are wrong. The map coordinates is a 32bit signed integer and goes from -2147483648 to 2147483647. The default map generator only generates chunk within 1000000 tiles from the origin and everything outside is out-of-map tiles. So when you walk/drive in a direction you will stop at 1000000. If you enter a plane or teleport you can get much further. And mods can catch the on_chunk_generated event and change the tiles from out-of-map to anything else.

So yes, vanilla has max 2000000x2000000 tiles (plus some out-of-map around that) but mods can have more. The game engine supports up to 18446744073709551616 tiles. The game estimates needing 18939904 Exabyte for that. Or 75759616 times the memory an amd64 CPU can address (48bit virtual address). Given the limit on the virtual address the maximum map size would be 2877268 * 2877268 tiles or for a ribbon world a width of about 1600 tiles and full length. Not that you could build much on that anymore since all the memory would be used for map chunks already.

Oh we are including mods now. Then neither solar or nuclear wins as you can just mod everything to not use any electricity or fuel. And while we are at it lets just make all research happen instantly without costing anything.

I seem to have completely missed the topic. If we are not discussing megabases here, complexity really does not matter at all as UPS will be at constant 60 whether you go nuclear or solar.
In that case, the metric to consider is what makes more fun to build, maintain, and especially look at. And for that well-defined and purely objective metric, nuclear obviously always wins over solar because of its pretty pipe works!

No you keep missing my point. Lets pull some numbers from thin air as no one has actually benchmarked these numbers. If we have say only 200 realtime hours to use for a run. Getting neccessary power from nuclear takes lets say 2 hours to build and makes the game lets say 1% slower. Then nuclear costs us 2h+200h*0.01= 4h. Now if we get the power from solar we have to spend lets say 6 hours building the solar plant and since it does not slow the game it costs us 6h+200h*0=6h. In this case nuclear is the optimal choise.

[mrvn]

And you (and the wiki) are wrong. The map coordinates is a 32bit signed integer and goes from -2147483648 to 2147483647. The default map generator only generates chunk within 1000000 tiles from the origin and everything outside is out-of-map tiles. So when you walk/drive in a direction you will stop at 1000000. If you enter a plane or teleport you can get much further. And mods can catch the on_chunk_generated event and change the tiles from out-of-map to anything else.

So yes, vanilla has max 2000000x2000000 tiles (plus some out-of-map around that) but mods can have more. The game engine supports up to 18446744073709551616 tiles. The game estimates needing 18939904 Exabyte for that. Or 75759616 times the memory an amd64 CPU can address (48bit virtual address). Given the limit on the virtual address the maximum map size would be 2877268 * 2877268 tiles or for a ribbon world a width of about 1600 tiles and full length. Not that you could build much on that anymore since all the memory would be used for map chunks already.

Oh we are including mods now. Then neither solar or nuclear wins as you can just mod everything to not use any electricity or fuel. And while we are at it lets just make all research happen instantly without costing anything.

No. This was about limitations of the game. And there it has a soft and a hard limit.

But to answer your suggestion of modding all buildings to need 0 power. I fairly certain that will run exactly the same speed down to the cpu cycle as with solar. The game still computes all the available power, all the required power (0 in this case) and gives every entity a fraction of available power even if the amount is 0. I highly doubt they optimized it for entities needing 0 power because that simply doesn't happen normally.

I seem to have completely missed the topic. If we are not discussing megabases here, complexity really does not matter at all as UPS will be at constant 60 whether you go nuclear or solar.
In that case, the metric to consider is what makes more fun to build, maintain, and especially look at. And for that well-defined and purely objective metric, nuclear obviously always wins over solar because of its pretty pipe works!

No you keep missing my point. Lets pull some numbers from thin air as no one has actually benchmarked these numbers. If we have say only 200 realtime hours to use for a run. Getting neccessary power from nuclear takes lets say 2 hours to build and makes the game lets say 1% slower. Then nuclear costs us 2h+200h*0.01= 4h. Now if we get the power from solar we have to spend lets say 6 hours building the solar plant and since it does not slow the game it costs us 6h+200h*0=6h. In this case nuclear is the optimal choise.

If you spend 6h building a solar plant then you are doing factorio wrong. It's about automation. So you spend 1h setting up bot production and blueprinting and then the bots build the plant in the background in 6h at 5% UPS cost. So 1h + 5*0.05 = 1.25h. Solar wins.

Note: it used to be much more than 1%. And when the game drops from 60 UPS to 30UPS playing is simply half the fun.

[PunPun]

No. This was about limitations of the game. And there it has a soft and a hard limit.

But to answer your suggestion of modding all buildings to need 0 power. I fairly certain that will run exactly the same speed down to the cpu cycle as with solar. The game still computes all the available power, all the required power (0 in this case) and gives every entity a fraction of available power even if the amount is 0. I highly doubt they optimized it for entities needing 0 power because that simply doesn't happen normally.

There are plenty of things in factorio that neither consumes or produces electricity. Infact it is completely impossible to launch a rocket without using them in vanilla so I would argue it normally happens pretty often. Also that calculation is done per electric network. If there is no electric networks then it is not performed. Ergo it is not the same amount of cpu cycles.

If you spend 6h building a solar plant then you are doing factorio wrong. It's about automation. So you spend 1h setting up bot production and blueprinting and then the bots build the plant in the background in 6h at 5% UPS cost. So 1h + 5*0.05 = 1.25h. Solar wins.

Note: it used to be much more than 1%. And when the game drops from 60 UPS to 30UPS playing is simply half the fun.

And if you use 2 hours building nuclear plants you are doing it wrong. So you spend 30 seconds to plop down the blueprints and wait 2 minutes for the robots to put it together. Also if you build the nuclear plants correctly(with minimal pipes and zero tanks) they take less than .1%(especially with the new parallel fluid simulation introduced in 0.17) so 0.042h+200h*0.001=0.242h and nuclear wins.

Also during those 6 hours the bots and asemblers and whatnots could be doing something more productive other than planting those solar panels. And you could have used the resources that went to the panels themselves in making something else. And that is not even counting the time it takes to clear all the biters and nests and landfill all the water where you plan to put down the 2mil panels.
It's not just the amount of time you personally spent clicking things.

These are also completely speculative numbers that are ultimately completely meaningless. It's pretty impossible to guess what the numbers actually are and what the break even point actually is until someone benchmarks them.

[Oktokolo]

I seem to have completely missed the topic. If we are not discussing megabases here, complexity really does not matter at all as UPS will be at constant 60 whether you go nuclear or solar.

No you keep missing my point.

Well, if we indeed are discussing megabases here, then UPS cost is the metric to use.
Does not matter how long it will take to setup (wich is done by bots after plopping down blueprints anyway). The nuclear setup will lower your UPS below 60 while the solar setup will not.
Solar always wins for megabases - because of O(1) for solar versus O(N) for nuclear.

[PunPun]

Well, if we indeed are discussing megabases here, then UPS cost is the metric to use.
Does not matter how long it will take to setup (wich is done by bots after plopping down blueprints anyway). The nuclear setup will lower your UPS below 60 while the solar setup will not.
Solar always wins for megabases - because of O(1) for solar versus O(N) for nuclear.

Would you spend 10 dollars to get a 10 cent discount on your mothly electric bill for a year. Because that is what you are doing when you use ~10x more resources to build the solar compared to equal amount of nuclear power.

[mrvn]

No. This was about limitations of the game. And there it has a soft and a hard limit.

But to answer your suggestion of modding all buildings to need 0 power. I fairly certain that will run exactly the same speed down to the cpu cycle as with solar. The game still computes all the available power, all the required power (0 in this case) and gives every entity a fraction of available power even if the amount is 0. I highly doubt they optimized it for entities needing 0 power because that simply doesn't happen normally.

There are plenty of things in factorio that neither consumes or produces electricity. Infact it is completely impossible to launch a rocket without using them in vanilla so I would argue it normally happens pretty often. Also that calculation is done per electric network. If there is no electric networks then it is not performed. Ergo it is not the same amount of cpu cycles.

Obviously things like belt are not in an electrical network. I was only talking about things that consume energy. And they will consume energy even if you set that to 0. I don't think factorio vanilla has any entity that can consume energy but has it set to 0. Even for mods I would think that strange.

And if you refer to things like a burner mining drill then that still consumes energy. It's not electrical, true. But there will be a check for the required energy level every tick. So setting things to be burner entities but use 0 energy would still check for the energy every time. So I stand by my words. I doubt setting energy consumption to 0 makes any difference.

You are right that without electrical networks the cost per network of solar cells would be gone. But simply setting the energy consumption to 0 won't have that affect. Try it out, mod an assembler to use 0 energy and place it without a power pole. Does the "no energy" icon blink? My guess is yes.

Changing every electrical entity to burner entity and setting them to 0 might though. The question then would be: Are burner entities more CPU efficient than electrical? They might be.

[Oktokolo]

Well, if we indeed are discussing megabases here, then UPS cost is the metric to use.
Does not matter how long it will take to setup (wich is done by bots after plopping down blueprints anyway). The nuclear setup will lower your UPS below 60 while the solar setup will not.
Solar always wins for megabases - because of O(1) for solar versus O(N) for nuclear.

Would you spend 10 dollars to get a 10 cent discount on your mothly electric bill for a year. Because that is what you are doing when you use ~10x more resources to build the solar compared to equal amount of nuclear power.

Would you spend it for beeing able to draw more power from the grid? Because that is more like what you get with Factorio: More SPM with solar than nuclear before UPS start dropping.

[mrvn]

Well, if we indeed are discussing megabases here, then UPS cost is the metric to use.
Does not matter how long it will take to setup (wich is done by bots after plopping down blueprints anyway). The nuclear setup will lower your UPS below 60 while the solar setup will not.
Solar always wins for megabases - because of O(1) for solar versus O(N) for nuclear.

Would you spend 10 dollars to get a 10 cent discount on your mothly electric bill for a year. Because that is what you are doing when you use ~10x more resources to build the solar compared to equal amount of nuclear power.

Would you spend it for beeing able to draw more power from the grid? Because that is more like what you get with Factorio: More SPM with solar than nuclear before UPS start dropping.

A 10 cent discount per kWh? Totally. That's like like 40% cheaper power.

[PunPun]

A 10 cent discount per kWh? Totally. That's like like 40% cheaper power.

No not per kWh. The entire bill. So if you paid $500 this month then with the discount it would have been $499.90.

Would you spend it for beeing able to draw more power from the grid? Because that is more like what you get with Factorio:

Not if it means I don't have enough money to buy things that would use that power. I can run my bandsaw a little slower if it means I can afford more wood to cut with it.

More SPM with solar than nuclear before UPS start dropping.

Well I guess if you only plan to play until the ups goes below 60 then solar might get you further. But if you play until a set amount of hours then the answer might be different. The game does not suddenly stop when ups starts to drop. If you can't play after it does then that is your personal problem. Also when you use solar you need to spend a lot more resources to get the same amount of watts. Meaning you will have to go to further away resources sooner and that is going to drop you ups. So in the end even the claim that using solar does not cause ups to drop is false.

that affect. Try it out, mod an assembler to use 0 energy and place it without a power pole. Does the "no energy" icon blink? My guess is yes.
Changing every electrical entity to burner entity and setting them to 0 might though. The question then would be: Are burner entities more CPU efficient than electrical? They might be.

Tried it. Changed it to use void energy source and it worked just fine without having a powerpole with no blinking icons. Did it for mining drills,labs and inserters too. And while I was at it made their crafting cost be 1 stone. I also made a recipe for one stone to give 10000 of every science and made it have .1s craft time. And also changed the max stacksize of all science to be 2mil and made inserters to move 2mil stacks and labs to have 2mil research speed. I very easily got a 10mil science per second factory that was running at 60ups with zero solar panels and longest part on the entire project was waiting for the game to load. And I could have gone much further if I wanted to but that would have been a waste of time.

[BlueTemplar]

Now try it with electric entities and a single GW(?) solar panel ?

[mrvn]

that affect. Try it out, mod an assembler to use 0 energy and place it without a power pole. Does the "no energy" icon blink? My guess is yes.
Changing every electrical entity to burner entity and setting them to 0 might though. The question then would be: Are burner entities more CPU efficient than electrical? They might be.

Tried it. Changed it to use void energy source and it worked just fine without having a powerpole with no blinking icons. Did it for mining drills,labs and inserters too. And while I was at it made their crafting cost be 1 stone. I also made a recipe for one stone to give 10000 of every science and made it have .1s craft time. And also changed the max stacksize of all science to be 2mil and made inserters to move 2mil stacks and labs to have 2mil research speed. I very easily got a 10mil science per second factory that was running at 60ups with zero solar panels and longest part on the entire project was waiting for the game to load. And I could have gone much further if I wanted to but that would have been a waste of time.

But was it any faster than the same with solar?