I just want to point out two things - It's obvious when you sit down and think about this, but I think a lot of people don't realize this for quite a while. Besides splitters and mergers, the game also has a hybrid of the two. The Industrial Storage. Has two inputs *AND* two outputs. Only thing you have to keep in mind is that its behavior can be kind of odd if you're using a mixed resource belt. Storage units only output from the last stack. When choosing between a manifold and a balancer, there's a third option - You can hybridize the two. Suppose you're feeding an array of 12 machines. There's nothing saying that you can't split the feed into 4 mini manifolds of 3 machines. Pros and cons sit right between the two options.
There's also hybrid setups where you have a manifold that feeds into several mini-load-balancers. They are frequently suitable for many low-tier stuff - iron/copper smelters take 30/m, so you can just have a single splitter between two smelters and its input comes from a mk1 belt coming from a manifold. Similar thing happens with iron plate, steel pipe and cable constructors (2 constructors for 60/m input), iron rod and copper wire constructors (4 constructors for 60/m, slightly more complex but not by much), steel beams (no splitter, just make the input belt a mk1). Heck, there's even some late-game recipes that can use it like the heat sink alternate recipe that uses 30/m of both casings and rubber. They show up all over the place if you look, and you get the modularity of manifolds with the rapid startup of load balancing for a very minor amount of extra work.
Re mixed resource belts ("sushi belts"): I'm not going to say that sushi belts are a bad idea. But I will say that they need to be used with caution and you need to be aware of their limitations. The basic rules: * The sushi belt must not stop moving under any circumstances. That means you must either terminate it at a sink, or have a smart splitter overflow into a sink at the end of the line (the latter is usually used in conjunction with storage or an item sorter setup). * Do not loop back, unless you are 100% positive that the input ratios are exactly the same as the output ratios. Do not loop back as part of an item sorter (if the sorter refused to pick something up the first time, that's because it's full and can't hold any more of that item, so looping it back is just a waste of belt space). * Don't try to load balance a sushi line directly. Sort it first and then balance it. Otherwise, you'll have many separate sushi lines which all need to be protected against backup, and that's wildly impractical. You can instead use a sushi manifold, which is fine because it only uses one sushi line (all the other lines are single item and don't need backup protection), but it's a manifold so you'll need to live with items filtering through slowly at first. * Don't try to include any high-volume items like screws or other tier 1 basics. In most cases, these items are better handled through dedicated lines.
I appreciate the detailed information on load balancer theory. I'm not a Math Person myself, so when a video is like "here's an example of a load balancer" and shows off some pre-made assemblage of splitters and conveyors with little explanation of why it is built the way that it is, it doesn't help me understand how to actually implement one. This video helped tremendously. Thank you!
That is exactly why I made these videos about balancers. I never understood it from other explanations, so I really dove into how they worked and tried to explain it in a way that would have made sense to me. I'm glad that is resonating with other people!
I've been playing the game since inception and there's a couple things here I've even been overlooking in my maths of committing to load balancing. Couple of my latest load balancing builds just were not running up to snuff when I knew they should have been so I switched them to Manifold. Now, that worked and I won't be going back but I could have possibly tried some of your fixes and kept load balancing. I will keep these fixes in mind for future when I get back to the game when 1.0 is released. TY!!
This was so helpful - I see a lot of people willing to tackle how manifolds work but not as many who are able to get into the weeds of load balancers. This video helped me get a better understanding and I'll certainly be implementing these methods into my game. It feels like the missing piece to keep my factories tidy.
I think the most correct way to tell whether you'll need loopback belts is that if *every* prime factor is 2 or 3 you can skip them, rather than if *one* prime factor is 2 or 3. Your text restatement around 5:40 says the same as your original statement :)
I think my intention was to communicate that, but perhaps I didn't. I tried to use simple language to avoid being wrong about something but maybe it was too simple lol.
literally all of the downsides for manifold are offset by 1) prefilling them 2) just waiting. And I PROMISE the time you wait will be less than the extra time it took you to calculate and setup the load balancing.
Underclocking (7:55) is a totally underrated solution in my opinion. Of course you need to have the space and the resources for the extra machines, but if its "just" some smelters - i really dont care. But Underclocking has a nice side effect: The machine takes less energy and i dont have to care about belt speeds, feeding back and yadda. especially working with multiple players on the same factory/world we endet up just building load balancers with enough machines in the most cases. only there the machines are too pricy we have those with loopback and a clear warning to ask the person X before doing changes here.
In my opinion, there are three reasons to use load balancing over manifold. One, a load balanced system is more fun to build than a manifold (not by a particularly massive margin, but it is more fun) two, it looks better because items are always flowing with no stutter, and three, you don't have hundreds of machines filled input and output with maxed stacks of items, which becomes a major mess if you want to dismantle the factory. All in all, the biggest crime of manifolds is that they are boring.
Fantastic explanation and will be used by me as my maths is not great! I also stage my factories: build a section, prime it with materials Then I use a throughput mod and dump the output into a sink to see how that section performs Once sorted, I move onto the next, leaving space for an overflow smart splitter should my calculations be wrong (they usually are... I love overflow splitters)
4:57 There may be a mistake here, or maybe a miscommunication. The loop back belt returns 60 materials a minute, and the total input stabilizes to 360 to compensate. I know this after applying an infinite geometric series formula to the problem (learned it in calculus 2). The formula is S = a/(1-r). r is the common ratio which is 1/6. a is 50 at the first iteration of our 300's division. Plugging it in and doing the math, we find that 50 becomes 350, then divides, and adds to the next 300 to become 358.33. This continues, slightly larger every time, 359.72, and again infinitely, until you approach the value of S, which is =60. Therefore your topmost conveyors need to take another 60, not 50, which is equal to your non-feedback output members.
That's beyond my pay grade, so I'll trust you. As far as I understand, if you require X materials, you input X materials into a loopback system, and for a while it not deliver the right amount to each machine, but after several iterations, it will balance out and each machine will get what it needs.
@@SatisfactoryNews In reality I suppose thats all you need to know xD. Im just saying that the feedback loop in the video gives back 60 not 50. Resulting in 360, not 350. Overall I used your video for in depth reference and took notes on everything. Which is how I found that issue. You did great regardless!
@@SatisfactoryNews for this same reason is why some of your bigger builds with loop backs require that overflow solution, because more circulation is present than expected. I believe its because the infinite division changes when the outputs are not uniform.
Another way to calculate this is to assume the system is at steady-state. Label the input belt rate as A. After the loopback, label it B. And let C be the total loopback. In the case of a 1:5 splitter, B gets evenly split 6 ways, so C = B/6. And because C is the loopback, we know A + C = B. Therefore, A + B/6 = B. Multiply by six and rearrange terms: 6A + B = 6B 6A = 5B 6A/5 = B So if A = 300 then B = 6×300/5 = 360, and C = B/6 = 60. Consider a second example, with a 1:13 balancer, which will have 16 outputs, with 3 of those being loopbacks. In this case, we still have A + C = B, but because of the different layout, now C = 3(B/16). A + 3B/16 = B 16A = 13B B = 16A/13 And by substituting, C = 3A/13. If our initial input A is 130 items/min, then B = 160 items/min, and C = 30 items/min.
So I might be crazy, but I don't try to perfectly balance things. I've always done load balancing, but since we need to use many of these items for crafting, I just...end it in a storage box, label the storage box with my input (amount crafted) and my output (amount used by machines further up the crafting chain) so I know my crafts/min of say iron plates. Since I need a surplus of iron plates for crafting machines/conveyor belts/etc. I want a surplus. I figured my options are make a system that ends at iron plates, or just make a surplus of plates and run them up. It still takes up inSANE amounts of space to do it this way (sometimes 20+ constructors with several layers of splitting and remerging on the output), but I never had to care much "excess" or waste. in the end I smart split before the box that holds my precious goods and I just send the excess of the storage into a SINK for juicy tickets. It works for me, and it lets me expand to the higher tier stuff quite nicely, cuz I already know what I'm making from the labels and then I know if I need more to just slap some more machines down (usually this ends w/ several systems for one material as its easier to take a new ore and make some new constructors etc. to add to the crafted amount than it is to rebuild or OC something earlier on). It works for me and it makes it so I can still go grab some random mats I may need from their box instead of having to hunt them while my system fills entirely
Ya, nah, i'll stick with the manifold. ot enough problems with math to use the loadbalancing type lol Great video. followed you until the load balancer
bro has these crazy machines while im just pumping everything into one smelter, one constructor, and one storage, and just overclocking everything 3 times
this was good, thank you very much. I've been playing for a long time but never actually got into load balancing too hard core due to the obvious complexities. I did make an over-flow once to get around an issue I ran into trying to take a couple of input belts and making n number of machines based off of that to maximize the production from those belts. I can't remember the example any more but I was feeding in 2 mk2 belts because I didn't have the faster belts I needed but it was feeding a weird number of machines. The over-flow helped this out. Thank you again.
About load balancing, you can stack mergers and splitters on top of each other and use the upward belts, it can be a bit tricky but by doing that, it takes up a lot less space you can also put the splitters right next to each others, you will still have to place a belt between them tho but it will seems like they are directly connected if done properly it's a bit hard to explain without showing it in action but i tried it myself and adding multiple looping belts is still as easy as when you do it the way you showed on the video although some might argue that it's way less satisfying, and i honestly can't disagree.
Very nice video. As an enthusiast of logistics, I find it both enjoyable and informative and appreciate you taking the time to (re)make it despite having covered the topic already. Well done. As a sidenote, most of the concepts covered could be extended to mixed belt (sushi) scenarios. Mixed-item manifolds (aka "sushifolds") can be commonly seen (especially at higher tiers and lower throughputs) and possibly confuse new players. And, even if with much more niche usage than mixed manifolds, even mixed belts can be load-balanced in different ways and even make use of loop-backs in some scenarios (just like normal balancers, but with more than one kind of item). If interested in covering the subject, I can provide more in-depth info (afaik, the only (rudimentary) video on the subject is on my channel, atm).
Thank you! I couldn't live with my errors anymore, haha. I haven't seen too many players using sushi belts but i will check out your video because that sounds interesting.
The only time I use slower belts for machines in manifolds is when the machine uses the exact rate that the slower belt can provide (60, 120, 270, etc.). In the general case it's better to fully max out your manifold belts with the highest tier of belts. To fully startup a manifold, the idea is that you're trying to fill up the internal buffers of the first X-2 machines ASAP. By using slower belts, you artificially limit how fast you can fill up the internal buffers. A very simple approach to tell if your manifold is filling up ASAP is to look at your input belt. If it's set up optimally it should only back up if the machines get clogged. If it's backing up before all the machines are filled, then your slower belts are artificially increasing your startup time. The input line will typically only get clogged like this if you have the last few machines that need to be filled but are limited by the lower belt speed.
ironically the opposite of what he suggested, using smart splitters and overflow on main and any on the machine on the side, would make the manifold fully start up fast
@skillfulfighter23 What using slower belts as manifold branches actually does is it decreases the number of machines you're trying to fill quickly. You gave the example of X-2 machines, which is correct in the case where the branches are the same speed as the main line. This is because the last two machines will fill at exactly the same rate once everything in front of them is full. But if the branches are slower than the main line, that changes. For example, if the main line is Mk5 (780) and the branches are Mk1 (60), then the math becomes "how fast can I fill the first X-13 machines", because the last thirteen machines will fill at exactly the same rate once everything in front of them is full. Which is "better" honestly depends on the number of machines in your manifold, compared to the ratio of manifold mainline speed to manifold branch speed. In one most extreme case (Mk6 belts vs Mk1), a manifold that contains twenty machines or fewer will fill up "instantly" (as soon as the material travels the length of the mainline). This means any manifold shorter than twenty will not benefit from a different approach (including balancing). In the other most extreme case (the branches are the same speed as or slower than the mainline), then any manifold longer than two machines will take longer to fill. I guess the overall point is that your statement of "you're trying to fill up the internal buffers of the first X-2 machines ASAP" isn't *quite* true. It's often true, but more specifically, it's a special case of the more general statement "you're trying to get enough internal buffers filled up that the last 2 machines are always getting their proportional share." The example of faster mainline slower branches helps lower the time to achieve *some* instances of this, but not others. It'll never make it worse though, as even if you're artificially limiting the speed at which you are filling up the first machine, you're filling up *other* machines by the same amount (unless the last machine is getting its equal share). Very situational-dependent. Your last point, however, is nearly completely valid. If the input belt is backing up on a fully-functional manifold, that means every single machine on the manifold either is maxed on the input *or is actively running*. No machine *needs* a full input buffer to run at 100% as long as it's running and is guaranteed enough input to keep it going.
One thing I do, on any type feed system, is to power all buildings, then switch them until the input(s) is full. I then turn on the buildings one at a time from closest to source to farthest.
In my opinion, a complex loop-back system is better than under-clocking because we have an abundance of space for the footprint but only a limited amount of power. Conveyors don't suck up any power (THANK YOU for that magic Coffee Stain!) so spamming them like a pile of spaghetti is only aesthetically problematic.
A note about the first two chapters, regarding how to reduce the "startup time" of manifolds. There's 2 preferences I'd like to highlight: wanting all machines to produce/reach 100% efficiency as soon as possible, or wanting to have output items come out of the system as soon as possible (these can lead to different results). Trying to have all machines in production, with a manifold, means that all machines (minus one or two) need to have an inventory full of input items. The QUICKEST ways to do so are (from the fastest): 1) Manual prefill. 2) Smart splitters in the manifold. 3) Normal manifold. 4) Manifold with low-tier input belts for machines (exception being when the belts' throughput matches the input requirement exactly, aka "balancefold", or the main belt has more input than the line of machines need so it can "overfeed" all inputs). What's a bit counterintuitive, is that method (4), while being the best at satisfying the second preference mentioned earlier, it's also the one that takes the longest to fill up all machines with input items. The reason for that is that since more machines start running while the whole system fills up, more input items will be consumed, thus needing more time for the main belt to provide the items to fill up machines AND feed the ones already consuming. On the opposite end, using smart splitters leads to the least items produced AND consumed by turning on machines one by one. Addition: if one dislikes manually filling machines but still wishes to speed up a normal manifold, they can add one or more container(s) and connect them to the manifold via (temporary) mergers. When provided with items, the containers will "refill" the manifold at the merging points, ideally after it has too few items/min on it as they're being used to fill up the first few machines of the production line.
One video addon to this that I would love for someone to do is explaining how machines also impact how many items fit on a belt. Far FAR too many people keep asking for ratio splitters because they 'only want 24 of x item down this line' as that's 'all that line needs' When if the machines on that line only need 24, then only 24 will fit on that belt line. There's no point nor need to ever tell a splitter how many items to allow on a belt. It may sound simple and common sense but many are not getting it. So something to show and explain how machines through put effects the amount of items on a belt would be great. Bonus, is also adding that just because a belt says 480ipm, don't mean every mixed item that gets added is also 480ipm each, then don't understand why their mixed belt isn't giving them the throughput they want. Some people just need to see thus stuff for themselves. I wish I could do it myself as MMMGH some of these people are too insistant... but my desktop only records gameplay footage at the rate of a slide show.
That's a really good point, and I have the same annoyance with people! I might make a Short about this. If I do, I will be sure to shout out your comment.
@@SatisfactoryNews Id do a full vid, just to nail the point. because holy cow... we have one person in the q&a site right now going and I quote: "if i split the 30 steel beamsPM thats go to send 15 a minute each way which is going to force me to overflow the 6 line and undeflow the 24 line, which with the idea i suggested wouldnt happen ever as it would split into the ratios set/required by this well thought out system" Then he ALSO add's. "how do you plan to split a conveyor belt of 120 between 9 coal generators (all underclocked at 88.888% which needs 13.3333> coalPM and 40 waterPM (makes it so i can use 3 coal gens per water extractor)) precision is key with a game like this." and I look at my 46 coal gens, my lines making 6000 quickwire... and everything else in my world and wonder what this kid is smoking.
Tbh, if you are doing a loop back load balancing, you don’t need to worry about the volume of items being sent back if you put an overload to the first belts after the first splitter.
I'm good with math but i really don't like the bunch of splitters, mergers and belts (and in some cases spaghettiness) needed for load balancing, i fell in love with "modular load balancers" mod, it's very aesthetic and easy to learn to use for everyone, a must have for lovers of order and minimalist design ❤
i dont understand what problem ppl are solving with load balancing.. whats the goal here? i always use manifolds and when machines further down the line run dry, i add injector lines. the best solution for most problems: push more resources into the system. i dont care if a input belt stops because its full/backing up, as long as the end product is flowing.
Appreciate the time you take to do make these videos. It's interesting to see others understanding of these methods, I agree you should use use whatever you are happy with, though given I always "load balance" first, it always seems that the disadvantages of "manifolds" are understated and those for load balancing exaggerated. Especially the space taken, which is usually more, but your 1 input to 10 output example, probably expanded to illustrate the loopback mechanism, can be built on the one 4m wide strip of tiles next to the machine inputs, so the same as allocated for a manifold. The power of load balancing also goes beyond the simple one input to multiple outputs shown in your videos, to distributing a common source in the exact amounts required by multiple processes. There is some math involved with this and needs to be done out of game, same as working out what splitter/merger assemply is required to do this, but can be used whenever the same ratios are present, whether you're making 10 parts per minute or 100. I unfortunately do have to remark on the Balancing: Loopback Overflow section. If overflow is required in the normal running of a process then balancing is not being done.
I think load balancing gets a bad reputation because it's most unnecessary. Manifold will get the job done 99% of the time. There are very few cases where a factory requires a load balancing system to function properly. You're right that size isn't really an excuse, especially in this massive game world, but it's something to keep in mind. As far as the loopback overflow section, there is some truth to your comment. When you have to do this, it becomes a less-than-perfect system. However, it gets pretty close. The system is not very efficient at the beginning, but the more iterations it goes through, the more balanced and up-to-speed it gets. It's also a "simple" solution that gets close enough for, again, like 99% of cases.
i notice that for 5 belts u can just make same thing u did with 10belts but merge two lines at the end so u can do loop even if main line is going at full speed
I hate to contradict your correction on the loopback belts from last video, but unless mergers work completely counter to how splitters work, then your original design is fine. True, you can't over-saturate the starting belt, but the merger will still inject an item into its output from a different input each operation if possible. Your machines only backed up in the first video because they weren't on. You won't end up with a deadlock of belts, because if the loopback locks (which it won't**), it would still just send the items into the machines like normal. It's actually better to have it sent back to line 1 and be "over-saturated" because then the machines at the end will always end up with a consistent amount of items, and you wont end up with a huge backlog of materials you aren't using, provided you throw a smart splitter before the merger of the loopback on line 1 to send the overflow somewhere else to be used. **Your loopback belt will never deadlock, even if you think it might, because you are thinking of the splitters and mergers the wrong way. You are not running a belt THROUGH a splitter and pulling another belt off of it, you are putting a belt INTO a splitter, and making it become more than one belt on the other side. Mergers are the same, you are not running line one THROUGH the merger, and only adding the loopback to line one, you are running both lines INTO the merger, and combining them into one output belt. If able, the merger will cycle which belt it pulls from every time it outputs an item. This may seem like a pointless distinction, but it's the only thing I can think of for why so many people thought you were wrong in the first video. Regardless of where you put it, having a loopback at all will effectively lower the efficiency of the amount of items you are inputting, vs. the amount you are using, because there will always be some amount of items on it. The fact that this would be "over-saturating" line 1 doesn't matter, because you weren't even using all the items on it anyways. You weren't wrong in the first video, but you also weren't technically wrong in this one either. Both are fine, but the way logistics work in this game, you can only really deadlock things that have more than 1 output (unless you are specifically trying to at least)
I think the point is, if your input belt is already at its maximum capacity, then a loopback belt will oversaturate the input belt. But, that depends on your numbers, and you can only guarantee that it applies to a fully utilized belt.
One thing you didn't mention that makes underclocking "too many" machines much more attractive: In addition to removing the need for loopback belts in your load balancing systems, "more machines" is a *one-time* cost that provides benefits for the lifetime of the factory... and best of all, underclocking has non-linear power usage benefits. Let's examine your example of needing 5 machines' output (at 7:55 in the video): 5 machines running at 100% speed uses 500% power, and requires a load balancing system with a loopback. In addition to simplifying your input by removing the need for a loopback, 6 machines running at 83.4% speed uses 471.5% power; that is, those 6 machines draw less power than the original 5 machines at full tilt. Stop here if you hate math, or you don't care about power draw. You can dramatically decrease the power requirements of the entire system (factory) by multiplying the number of buildings and underclocking them to produce the desired amount of product. 12 machines running at 41.67% speed would use less than 3.8 machines' worth of power, while still producing the output of 5 machines... but effectively giving you more than a full machine's output for free, power-wise. I'm well aware that this next example is going to look stupid at first glance, but bear with me; in all honesty, it's not even as silly as I have taken this concept. Let's say you still want the output of 5 machines at 100% power, but you're willing to build 36 machines, each underclocked to 13.89% speed. The power requirement for those machines is only 3.15% each, so you end up producing your 500% output for a power cost of only a hair over 113.5% power... which means you're getting nearly 4 machines' worth of output FOR FREE (in terms of power). As regards taking the concept to silly extremes: I have built an entire early-game production system making plates, rods, screws, wire, cable, and concrete... while using only the 40 power produced by the HUB's pair of (weak) biomass generators. It required pure nodes so I could underclock the miners (which were capped by belt speed anyway), and it took many hours to build hundreds of smelters and constructors clocked at (roughly) 3%... but while it was an absolutely ridiculous project, it was also absurdly satisfying. Yes, it uses more space to build more machines... but you *really* need to get over the idea that space is even a consideration. According to the world size listed on the wiki, there's a theoretical build volume of 105,578 cubic *kilometers* on the map; you would need 736,127 foundations just to cover a single layer. Unless you're trying to build inside a cave, stop thinking about how much space your factory will use; there's nearly always dozens of unused floors of space *above* your factory that you're simply wasting... and FICSIT does not waste. Apply verticality to your build, instead. Edited to add: There's yet another benefit to "extreme" (half or lower production speed) underclocking that I forgot to mention: later, when you have bigger miners and faster belts, you can simply throw more materials at your existing factory, replace belts, and clock everything up... and you've increased production without having to build a new factory. TL;DR: Stop thinking in twos and threes... start thinking in (at least) sixes and twelves. Underclock and overbuild; it's fun and efficient!
Question about the loop back belts - if when an output belt is full the splitter ignores it, why not just have the 6 belts at the end of the two splitters (if you are trying to achieve 1:5 for instance), but just not connect the 6th to anything? Then you wouldn’t need to worry about a loop back belt oversaturating the input and the load should be shared evenly among the 5 left over right?
not evenly. if the first splitter splits 1:2 and the subsequent ones split 1:3 but the one one the left has a saturated output it effectively also becomes 1:2. So on the left you get 1/2 * 1/2 = 1/4 on each output and on the right you get 1/2 * 1/3 = 1/6 on each output. All of this assumes free flowing belts, When the belts start backing up the resources will just flow to wherever they are missing (that's how splitters work in that situation) so you could just use a manifold instead.
In theory, that makes sense. But in reality, depending on how saturated the belt gets, it might not work. The first splitter will split the input in half. But one half is only serving 2 outputs, while the other is serving 3. Therefore the side with the dummy belt is getting more resources than it needs, and more than the other half. In some cases this will balance out, but at that point you would have been better off making a manifold.
Not for nothing but I've never felt the need to use balancers instead of manifolds. Takes time to saturate? And? you'll set it up once and walk away. All the funky math around loop backs on balancers? slap a manifold on it and walk away lol
Before I see the revised rule I'll try to make my own: (it is easy to load balance a number of machines divisible by n^2 or n^3), please correct me if I am wrong. Edit: I was close enough.
For the overflow, if you don't need the additional overflow, could you just sink it? Like, if you had 120 output, but only needed 115, instead of looping back the extra 5, it'd just get sunk.
Definitely could do that! Nothing wrong either leaving extra in the system to smooth out any irregularities but if the belts being backed up would cause an issue, you should definitely sink the extra.
I linked an imgur gallery of load balancer designs to the last vid, but it looks like my comment was removed before the last vid was taken down. Are links like that not allowed? Also, good vid. I've recently switched to using manifolds myself, but I find balancers to be more aesthetically pleasing.
7:20 Couldn't you do this for the 5 example too? Just split both of the input and the loopback belts in two and merge the respective halves before splitting each of those into 3.
I use load balancing and underclocking for inputs to all machines, it makes me feel wasteful to let all machines load up And use manifold only when produce has to be only stored and not sent to another machine
Why do you need two loopbacks on the example with 12 storage containers? Why not just send those two onward like you did on the right side? Doesn't 12 follow the "rule of 2 or 3" such that it doesn't need loopback belts?
Hmm I think maybe you misinterpreted something? The example you're referring to that has 2 loopback belts is 10 containers, as I was showing how to do a loopback system that doesn't follow the "rule of 2 or 3". But you need to round up to the nearest number that does follow the rule, which is 12 belts, then loopback those 2 extra belts.
Great... I'm ahead of progress then UGH and last belt is on last 2 elevator upgrade, and I just spend hours on mini factories setup, luckily not too detailed, but still. edit: figured it out, will do extra steps, but need shift the 1/5 later part.of the chain, phew
I see a lot of people asking themselves why anyone would use a load balancing system instead of a manifold. I use it for my biofuel factory. I have 12 biofuel burners and with a manifold, the last machine runs out of materials while the first one still has a ton in storage. This means I get a power outage, while still having enough biofuel.
It matters in some very particular cases, and it's a way to nerd out about math which some people like. I have exactly one load balancer in my game, it's really not needed in most cases.
i am here at this video because i just made 6 nuclear power plants. and the first plant is filled up with materials but the last plant is still at 0. due to not all 6 plants receiving a 1 to 1 input. the machines previously in the chain are backing up on sulfuric acid because i set it all up to be a perfect 1:1 ratio. load balancing would have fixed my issue but i went manifold
Just use a manifold, period. Set it up, and by the time you turn around do almost any other task, the manifold will be full and primed and all of your machines will be doing their jobs.
Excellent Vid, thanx for the update / corrections! Now I have another wrench to toss into your machine! (please don't hate me!) Your Load Balancer does NOT have to take a 'country mile' to set up! ruclips.net/video/ILEY8Z46A4c/видео.html or this one; ruclips.net/video/MaFM0Dlp3nA/видео.html Try these, they are a little complicated, but it is very tight and neat, and fortunately doesn't need to be rebuilt over and over, else a blueprint would be best.
you did the ONE thing you must not do wit ha factory. create a belt swastika. i will NEVER take your advice on anything EVER. youve broken the cardinal rule. the one rule that must not be bent, let alone broken. i am severely disappointed.
![Wiz Khalifa - Khalifa's Home [Official Music Video]](http://i.ytimg.com/vi/wnrLCTe0Nz4/mqdefault.jpg)
I just want to point out two things -
It's obvious when you sit down and think about this, but I think a lot of people don't realize this for quite a while. Besides splitters and mergers, the game also has a hybrid of the two. The Industrial Storage. Has two inputs *AND* two outputs. Only thing you have to keep in mind is that its behavior can be kind of odd if you're using a mixed resource belt. Storage units only output from the last stack.
When choosing between a manifold and a balancer, there's a third option - You can hybridize the two. Suppose you're feeding an array of 12 machines. There's nothing saying that you can't split the feed into 4 mini manifolds of 3 machines. Pros and cons sit right between the two options.
There's also hybrid setups where you have a manifold that feeds into several mini-load-balancers. They are frequently suitable for many low-tier stuff - iron/copper smelters take 30/m, so you can just have a single splitter between two smelters and its input comes from a mk1 belt coming from a manifold.
Similar thing happens with iron plate, steel pipe and cable constructors (2 constructors for 60/m input), iron rod and copper wire constructors (4 constructors for 60/m, slightly more complex but not by much), steel beams (no splitter, just make the input belt a mk1). Heck, there's even some late-game recipes that can use it like the heat sink alternate recipe that uses 30/m of both casings and rubber.
They show up all over the place if you look, and you get the modularity of manifolds with the rapid startup of load balancing for a very minor amount of extra work.
Re mixed resource belts ("sushi belts"): I'm not going to say that sushi belts are a bad idea. But I will say that they need to be used with caution and you need to be aware of their limitations. The basic rules:
* The sushi belt must not stop moving under any circumstances. That means you must either terminate it at a sink, or have a smart splitter overflow into a sink at the end of the line (the latter is usually used in conjunction with storage or an item sorter setup).
* Do not loop back, unless you are 100% positive that the input ratios are exactly the same as the output ratios. Do not loop back as part of an item sorter (if the sorter refused to pick something up the first time, that's because it's full and can't hold any more of that item, so looping it back is just a waste of belt space).
* Don't try to load balance a sushi line directly. Sort it first and then balance it. Otherwise, you'll have many separate sushi lines which all need to be protected against backup, and that's wildly impractical. You can instead use a sushi manifold, which is fine because it only uses one sushi line (all the other lines are single item and don't need backup protection), but it's a manifold so you'll need to live with items filtering through slowly at first.
* Don't try to include any high-volume items like screws or other tier 1 basics. In most cases, these items are better handled through dedicated lines.
This doesn’t work because it doesn’t split items evenly.
I appreciate the detailed information on load balancer theory. I'm not a Math Person myself, so when a video is like "here's an example of a load balancer" and shows off some pre-made assemblage of splitters and conveyors with little explanation of why it is built the way that it is, it doesn't help me understand how to actually implement one. This video helped tremendously. Thank you!
That is exactly why I made these videos about balancers. I never understood it from other explanations, so I really dove into how they worked and tried to explain it in a way that would have made sense to me. I'm glad that is resonating with other people!
I've been playing the game since inception and there's a couple things here I've even been overlooking in my maths of committing to load balancing. Couple of my latest load balancing builds just were not running up to snuff when I knew they should have been so I switched them to Manifold. Now, that worked and I won't be going back but I could have possibly tried some of your fixes and kept load balancing.
I will keep these fixes in mind for future when I get back to the game when 1.0 is released.
TY!!
This was so helpful - I see a lot of people willing to tackle how manifolds work but not as many who are able to get into the weeds of load balancers. This video helped me get a better understanding and I'll certainly be implementing these methods into my game. It feels like the missing piece to keep my factories tidy.
I think the most correct way to tell whether you'll need loopback belts is that if *every* prime factor is 2 or 3 you can skip them, rather than if *one* prime factor is 2 or 3. Your text restatement around 5:40 says the same as your original statement :)
I think my intention was to communicate that, but perhaps I didn't. I tried to use simple language to avoid being wrong about something but maybe it was too simple lol.
It's hard to be specific and succinct without using terms that most people forget as soon as they leave school lol.
Glad to see this uploaded again! It'll get better every time. 🤠
literally all of the downsides for manifold are offset by 1) prefilling them 2) just waiting. And I PROMISE the time you wait will be less than the extra time it took you to calculate and setup the load balancing.
Exactly
Yep.
But load balancing looks dope af
Underclocking (7:55) is a totally underrated solution in my opinion. Of course you need to have the space and the resources for the extra machines, but if its "just" some smelters - i really dont care. But Underclocking has a nice side effect: The machine takes less energy and i dont have to care about belt speeds, feeding back and yadda. especially working with multiple players on the same factory/world we endet up just building load balancers with enough machines in the most cases.
only there the machines are too pricy we have those with loopback and a clear warning to ask the person X before doing changes here.
In my opinion, there are three reasons to use load balancing over manifold. One, a load balanced system is more fun to build than a manifold (not by a particularly massive margin, but it is more fun) two, it looks better because items are always flowing with no stutter, and three, you don't have hundreds of machines filled input and output with maxed stacks of items, which becomes a major mess if you want to dismantle the factory.
All in all, the biggest crime of manifolds is that they are boring.
Fantastic explanation and will be used by me as my maths is not great!
I also stage my factories:
build a section, prime it with materials
Then I use a throughput mod and dump the output into a sink to see how that section performs
Once sorted, I move onto the next, leaving space for an overflow smart splitter should my calculations be wrong (they usually are... I love overflow splitters)
I'm with you on overflow splitters, I have more than a few of those 😂
never knew about manifold, just instantly went to load splitting as soon as i started the game
4:57 There may be a mistake here, or maybe a miscommunication. The loop back belt returns 60 materials a minute, and the total input stabilizes to 360 to compensate. I know this after applying an infinite geometric series formula to the problem (learned it in calculus 2).
The formula is S = a/(1-r).
r is the common ratio which is 1/6.
a is 50 at the first iteration of our 300's division.
Plugging it in and doing the math, we find that 50 becomes 350, then divides, and adds to the next 300 to become 358.33.
This continues, slightly larger every time, 359.72, and again infinitely, until you approach the value of S, which is =60.
Therefore your topmost conveyors need to take another 60, not 50, which is equal to your non-feedback output members.
That's beyond my pay grade, so I'll trust you. As far as I understand, if you require X materials, you input X materials into a loopback system, and for a while it not deliver the right amount to each machine, but after several iterations, it will balance out and each machine will get what it needs.
@@SatisfactoryNews In reality I suppose thats all you need to know xD. Im just saying that the feedback loop in the video gives back 60 not 50. Resulting in 360, not 350. Overall I used your video for in depth reference and took notes on everything. Which is how I found that issue. You did great regardless!
@@SatisfactoryNews for this same reason is why some of your bigger builds with loop backs require that overflow solution, because more circulation is present than expected.
I believe its because the infinite division changes when the outputs are not uniform.
Another way to calculate this is to assume the system is at steady-state. Label the input belt rate as A. After the loopback, label it B. And let C be the total loopback.
In the case of a 1:5 splitter, B gets evenly split 6 ways, so C = B/6. And because C is the loopback, we know A + C = B.
Therefore, A + B/6 = B. Multiply by six and rearrange terms:
6A + B = 6B
6A = 5B
6A/5 = B
So if A = 300 then B = 6×300/5 = 360, and C = B/6 = 60.
Consider a second example, with a 1:13 balancer, which will have 16 outputs, with 3 of those being loopbacks. In this case, we still have A + C = B, but because of the different layout, now C = 3(B/16).
A + 3B/16 = B
16A = 13B
B = 16A/13
And by substituting, C = 3A/13.
If our initial input A is 130 items/min, then B = 160 items/min, and C = 30 items/min.
@@coolman5001 not gonna lie that seems easier (good job)
So I might be crazy, but I don't try to perfectly balance things. I've always done load balancing, but since we need to use many of these items for crafting, I just...end it in a storage box, label the storage box with my input (amount crafted) and my output (amount used by machines further up the crafting chain) so I know my crafts/min of say iron plates. Since I need a surplus of iron plates for crafting machines/conveyor belts/etc. I want a surplus. I figured my options are make a system that ends at iron plates, or just make a surplus of plates and run them up. It still takes up inSANE amounts of space to do it this way (sometimes 20+ constructors with several layers of splitting and remerging on the output), but I never had to care much "excess" or waste. in the end I smart split before the box that holds my precious goods and I just send the excess of the storage into a SINK for juicy tickets. It works for me, and it lets me expand to the higher tier stuff quite nicely, cuz I already know what I'm making from the labels and then I know if I need more to just slap some more machines down (usually this ends w/ several systems for one material as its easier to take a new ore and make some new constructors etc. to add to the crafted amount than it is to rebuild or OC something earlier on). It works for me and it makes it so I can still go grab some random mats I may need from their box instead of having to hunt them while my system fills entirely
Top level balancers and 4 splitters manifold, works reasonably well
Ya, nah, i'll stick with the manifold. ot enough problems with math to use the loadbalancing type lol Great video. followed you until the load balancer
Yeah this load balancing stuff isn't for me, either 😂 I am generally not a fan.
bro has these crazy machines while im just pumping everything into one smelter, one constructor, and one storage, and just overclocking everything 3 times
8:18 that plus 1 machine isnt taking up more space than all those splitters and mergers for the loopback
this was good, thank you very much. I've been playing for a long time but never actually got into load balancing too hard core due to the obvious complexities. I did make an over-flow once to get around an issue I ran into trying to take a couple of input belts and making n number of machines based off of that to maximize the production from those belts. I can't remember the example any more but I was feeding in 2 mk2 belts because I didn't have the faster belts I needed but it was feeding a weird number of machines. The over-flow helped this out.
Thank you again.
About load balancing, you can stack mergers and splitters on top of each other and use the upward belts, it can be a bit tricky but by doing that, it takes up a lot less space
you can also put the splitters right next to each others, you will still have to place a belt between them tho but it will seems like they are directly connected if done properly
it's a bit hard to explain without showing it in action but i tried it myself and adding multiple looping belts is still as easy as when you do it the way you showed on the video
although some might argue that it's way less satisfying, and i honestly can't disagree.
solid vid man, motivating me to do some similar verse "Let's play series"
Very nice video. As an enthusiast of logistics, I find it both enjoyable and informative and appreciate you taking the time to (re)make it despite having covered the topic already. Well done.
As a sidenote, most of the concepts covered could be extended to mixed belt (sushi) scenarios. Mixed-item manifolds (aka "sushifolds") can be commonly seen (especially at higher tiers and lower throughputs) and possibly confuse new players. And, even if with much more niche usage than mixed manifolds, even mixed belts can be load-balanced in different ways and even make use of loop-backs in some scenarios (just like normal balancers, but with more than one kind of item). If interested in covering the subject, I can provide more in-depth info (afaik, the only (rudimentary) video on the subject is on my channel, atm).
Thank you! I couldn't live with my errors anymore, haha.
I haven't seen too many players using sushi belts but i will check out your video because that sounds interesting.
The only time I use slower belts for machines in manifolds is when the machine uses the exact rate that the slower belt can provide (60, 120, 270, etc.). In the general case it's better to fully max out your manifold belts with the highest tier of belts.
To fully startup a manifold, the idea is that you're trying to fill up the internal buffers of the first X-2 machines ASAP. By using slower belts, you artificially limit how fast you can fill up the internal buffers.
A very simple approach to tell if your manifold is filling up ASAP is to look at your input belt. If it's set up optimally it should only back up if the machines get clogged. If it's backing up before all the machines are filled, then your slower belts are artificially increasing your startup time. The input line will typically only get clogged like this if you have the last few machines that need to be filled but are limited by the lower belt speed.
ironically the opposite of what he suggested, using smart splitters and overflow on main and any on the machine on the side, would make the manifold fully start up fast
@skillfulfighter23 What using slower belts as manifold branches actually does is it decreases the number of machines you're trying to fill quickly. You gave the example of X-2 machines, which is correct in the case where the branches are the same speed as the main line. This is because the last two machines will fill at exactly the same rate once everything in front of them is full.
But if the branches are slower than the main line, that changes. For example, if the main line is Mk5 (780) and the branches are Mk1 (60), then the math becomes "how fast can I fill the first X-13 machines", because the last thirteen machines will fill at exactly the same rate once everything in front of them is full.
Which is "better" honestly depends on the number of machines in your manifold, compared to the ratio of manifold mainline speed to manifold branch speed. In one most extreme case (Mk6 belts vs Mk1), a manifold that contains twenty machines or fewer will fill up "instantly" (as soon as the material travels the length of the mainline). This means any manifold shorter than twenty will not benefit from a different approach (including balancing). In the other most extreme case (the branches are the same speed as or slower than the mainline), then any manifold longer than two machines will take longer to fill.
I guess the overall point is that your statement of "you're trying to fill up the internal buffers of the first X-2 machines ASAP" isn't *quite* true. It's often true, but more specifically, it's a special case of the more general statement "you're trying to get enough internal buffers filled up that the last 2 machines are always getting their proportional share." The example of faster mainline slower branches helps lower the time to achieve *some* instances of this, but not others. It'll never make it worse though, as even if you're artificially limiting the speed at which you are filling up the first machine, you're filling up *other* machines by the same amount (unless the last machine is getting its equal share). Very situational-dependent.
Your last point, however, is nearly completely valid. If the input belt is backing up on a fully-functional manifold, that means every single machine on the manifold either is maxed on the input *or is actively running*. No machine *needs* a full input buffer to run at 100% as long as it's running and is guaranteed enough input to keep it going.
One thing I do, on any type feed system, is to power all buildings, then switch them until the input(s) is full. I then turn on the buildings one at a time from closest to source to farthest.
In my opinion, a complex loop-back system is better than under-clocking because we have an abundance of space for the footprint but only a limited amount of power. Conveyors don't suck up any power (THANK YOU for that magic Coffee Stain!) so spamming them like a pile of spaghetti is only aesthetically problematic.
yeah? you can underclock the buildings/machines so that they dont consume extra power than needed for the desired amount though
Underclocking saves power per item
A note about the first two chapters, regarding how to reduce the "startup time" of manifolds.
There's 2 preferences I'd like to highlight: wanting all machines to produce/reach 100% efficiency as soon as possible, or wanting to have output items come out of the system as soon as possible (these can lead to different results).
Trying to have all machines in production, with a manifold, means that all machines (minus one or two) need to have an inventory full of input items. The QUICKEST ways to do so are (from the fastest):
1) Manual prefill.
2) Smart splitters in the manifold.
3) Normal manifold.
4) Manifold with low-tier input belts for machines (exception being when the belts' throughput matches the input requirement exactly, aka "balancefold", or the main belt has more input than the line of machines need so it can "overfeed" all inputs).
What's a bit counterintuitive, is that method (4), while being the best at satisfying the second preference mentioned earlier, it's also the one that takes the longest to fill up all machines with input items. The reason for that is that since more machines start running while the whole system fills up, more input items will be consumed, thus needing more time for the main belt to provide the items to fill up machines AND feed the ones already consuming. On the opposite end, using smart splitters leads to the least items produced AND consumed by turning on machines one by one.
Addition: if one dislikes manually filling machines but still wishes to speed up a normal manifold, they can add one or more container(s) and connect them to the manifold via (temporary) mergers. When provided with items, the containers will "refill" the manifold at the merging points, ideally after it has too few items/min on it as they're being used to fill up the first few machines of the production line.
Thank you for the comment and all the additional information! I'm hearting this comment so it appears up at the top so people can see it.
One video addon to this that I would love for someone to do is explaining how machines also impact how many items fit on a belt. Far FAR too many people keep asking for ratio splitters because they 'only want 24 of x item down this line' as that's 'all that line needs' When if the machines on that line only need 24, then only 24 will fit on that belt line. There's no point nor need to ever tell a splitter how many items to allow on a belt. It may sound simple and common sense but many are not getting it. So something to show and explain how machines through put effects the amount of items on a belt would be great.
Bonus, is also adding that just because a belt says 480ipm, don't mean every mixed item that gets added is also 480ipm each, then don't understand why their mixed belt isn't giving them the throughput they want.
Some people just need to see thus stuff for themselves. I wish I could do it myself as MMMGH some of these people are too insistant... but my desktop only records gameplay footage at the rate of a slide show.
That's a really good point, and I have the same annoyance with people! I might make a Short about this. If I do, I will be sure to shout out your comment.
@@SatisfactoryNews Id do a full vid, just to nail the point. because holy cow... we have one person in the q&a site right now going and I quote:
"if i split the 30 steel beamsPM thats go to send 15 a minute each way which is going to force me to overflow the 6 line and undeflow the 24 line, which with the idea i suggested wouldnt happen ever as it would split into the ratios set/required by this well thought out system"
Then he ALSO add's.
"how do you plan to split a conveyor belt of 120 between 9 coal generators (all underclocked at 88.888% which needs 13.3333> coalPM and 40 waterPM (makes it so i can use 3 coal gens per water extractor)) precision is key with a game like this."
and I look at my 46 coal gens, my lines making 6000 quickwire... and everything else in my world and wonder what this kid is smoking.
Tbh, if you are doing a loop back load balancing, you don’t need to worry about the volume of items being sent back if you put an overload to the first belts after the first splitter.
I'm good with math but i really don't like the bunch of splitters, mergers and belts (and in some cases spaghettiness) needed for load balancing, i fell in love with "modular load balancers" mod, it's very aesthetic and easy to learn to use for everyone, a must have for lovers of order and minimalist design ❤
"Multi Splitter/MergerBalancer" mod also has balancers. Made everything look pretty, functional, and consume less space.
I have only ever used manifolds, gotta think about the space!
Loadbalancing sounds like a solution that problem that shouldn't exist. Manifold has only 1 issues, the lack of player's patience. :D
i dont understand what problem ppl are solving with load balancing.. whats the goal here? i always use manifolds and when machines further down the line run dry, i add injector lines. the best solution for most problems: push more resources into the system. i dont care if a input belt stops because its full/backing up, as long as the end product is flowing.
@@FalconDS9 Not having to wait 2 minutes, mostly. Maybe at later stage in the game it shines, I just started mid game lol
Appreciate the time you take to do make these videos. It's interesting to see others understanding of these methods, I agree you should use use whatever you are happy with, though given I always "load balance" first, it always seems that the disadvantages of "manifolds" are understated and those for load balancing exaggerated. Especially the space taken, which is usually more, but your 1 input to 10 output example, probably expanded to illustrate the loopback mechanism, can be built on the one 4m wide strip of tiles next to the machine inputs, so the same as allocated for a manifold.
The power of load balancing also goes beyond the simple one input to multiple outputs shown in your videos, to distributing a common source in the exact amounts required by multiple processes. There is some math involved with this and needs to be done out of game, same as working out what splitter/merger assemply is required to do this, but can be used whenever the same ratios are present, whether you're making 10 parts per minute or 100.
I unfortunately do have to remark on the Balancing: Loopback Overflow section. If overflow is required in the normal running of a process then balancing is not being done.
I think load balancing gets a bad reputation because it's most unnecessary. Manifold will get the job done 99% of the time. There are very few cases where a factory requires a load balancing system to function properly. You're right that size isn't really an excuse, especially in this massive game world, but it's something to keep in mind.
As far as the loopback overflow section, there is some truth to your comment. When you have to do this, it becomes a less-than-perfect system. However, it gets pretty close. The system is not very efficient at the beginning, but the more iterations it goes through, the more balanced and up-to-speed it gets. It's also a "simple" solution that gets close enough for, again, like 99% of cases.
Big advantage of manifold is you can add/remove machines easily.
i notice that for 5 belts u can just make same thing u did with 10belts but merge two lines at the end so u can do loop even if main line is going at full speed
If want you can use manifold input also center so example 4 smelters your input is between 2 and 3 working maybe little bit better 🤔
Yeah I cover manifold injection in the original video
I hate to contradict your correction on the loopback belts from last video, but unless mergers work completely counter to how splitters work, then your original design is fine. True, you can't over-saturate the starting belt, but the merger will still inject an item into its output from a different input each operation if possible. Your machines only backed up in the first video because they weren't on. You won't end up with a deadlock of belts, because if the loopback locks (which it won't**), it would still just send the items into the machines like normal. It's actually better to have it sent back to line 1 and be "over-saturated" because then the machines at the end will always end up with a consistent amount of items, and you wont end up with a huge backlog of materials you aren't using, provided you throw a smart splitter before the merger of the loopback on line 1 to send the overflow somewhere else to be used.
**Your loopback belt will never deadlock, even if you think it might, because you are thinking of the splitters and mergers the wrong way. You are not running a belt THROUGH a splitter and pulling another belt off of it, you are putting a belt INTO a splitter, and making it become more than one belt on the other side. Mergers are the same, you are not running line one THROUGH the merger, and only adding the loopback to line one, you are running both lines INTO the merger, and combining them into one output belt. If able, the merger will cycle which belt it pulls from every time it outputs an item. This may seem like a pointless distinction, but it's the only thing I can think of for why so many people thought you were wrong in the first video.
Regardless of where you put it, having a loopback at all will effectively lower the efficiency of the amount of items you are inputting, vs. the amount you are using, because there will always be some amount of items on it. The fact that this would be "over-saturating" line 1 doesn't matter, because you weren't even using all the items on it anyways.
You weren't wrong in the first video, but you also weren't technically wrong in this one either. Both are fine, but the way logistics work in this game, you can only really deadlock things that have more than 1 output (unless you are specifically trying to at least)
I think the point is, if your input belt is already at its maximum capacity, then a loopback belt will oversaturate the input belt. But, that depends on your numbers, and you can only guarantee that it applies to a fully utilized belt.
One thing you didn't mention that makes underclocking "too many" machines much more attractive:
In addition to removing the need for loopback belts in your load balancing systems, "more machines" is a *one-time* cost that provides benefits for the lifetime of the factory... and best of all, underclocking has non-linear power usage benefits.
Let's examine your example of needing 5 machines' output (at 7:55 in the video):
5 machines running at 100% speed uses 500% power, and requires a load balancing system with a loopback.
In addition to simplifying your input by removing the need for a loopback, 6 machines running at 83.4% speed uses 471.5% power; that is, those 6 machines draw less power than the original 5 machines at full tilt.
Stop here if you hate math, or you don't care about power draw.
You can dramatically decrease the power requirements of the entire system (factory) by multiplying the number of buildings and underclocking them to produce the desired amount of product. 12 machines running at 41.67% speed would use less than 3.8 machines' worth of power, while still producing the output of 5 machines... but effectively giving you more than a full machine's output for free, power-wise.
I'm well aware that this next example is going to look stupid at first glance, but bear with me; in all honesty, it's not even as silly as I have taken this concept. Let's say you still want the output of 5 machines at 100% power, but you're willing to build 36 machines, each underclocked to 13.89% speed. The power requirement for those machines is only 3.15% each, so you end up producing your 500% output for a power cost of only a hair over 113.5% power... which means you're getting nearly 4 machines' worth of output FOR FREE (in terms of power).
As regards taking the concept to silly extremes: I have built an entire early-game production system making plates, rods, screws, wire, cable, and concrete... while using only the 40 power produced by the HUB's pair of (weak) biomass generators. It required pure nodes so I could underclock the miners (which were capped by belt speed anyway), and it took many hours to build hundreds of smelters and constructors clocked at (roughly) 3%... but while it was an absolutely ridiculous project, it was also absurdly satisfying.
Yes, it uses more space to build more machines... but you *really* need to get over the idea that space is even a consideration. According to the world size listed on the wiki, there's a theoretical build volume of 105,578 cubic *kilometers* on the map; you would need 736,127 foundations just to cover a single layer. Unless you're trying to build inside a cave, stop thinking about how much space your factory will use; there's nearly always dozens of unused floors of space *above* your factory that you're simply wasting... and FICSIT does not waste. Apply verticality to your build, instead.
Edited to add:
There's yet another benefit to "extreme" (half or lower production speed) underclocking that I forgot to mention: later, when you have bigger miners and faster belts, you can simply throw more materials at your existing factory, replace belts, and clock everything up... and you've increased production without having to build a new factory.
TL;DR: Stop thinking in twos and threes... start thinking in (at least) sixes and twelves. Underclock and overbuild; it's fun and efficient!
Question about the loop back belts - if when an output belt is full the splitter ignores it, why not just have the 6 belts at the end of the two splitters (if you are trying to achieve 1:5 for instance), but just not connect the 6th to anything? Then you wouldn’t need to worry about a loop back belt oversaturating the input and the load should be shared evenly among the 5 left over right?
not evenly. if the first splitter splits 1:2 and the subsequent ones split 1:3 but the one one the left has a saturated output it effectively also becomes 1:2.
So on the left you get
1/2 * 1/2 = 1/4 on each output and on the right you get
1/2 * 1/3 = 1/6 on each output.
All of this assumes free flowing belts, When the belts start backing up the resources will just flow to wherever they are missing (that's how splitters work in that situation) so you could just use a manifold instead.
In theory, that makes sense. But in reality, depending on how saturated the belt gets, it might not work. The first splitter will split the input in half. But one half is only serving 2 outputs, while the other is serving 3. Therefore the side with the dummy belt is getting more resources than it needs, and more than the other half. In some cases this will balance out, but at that point you would have been better off making a manifold.
Not for nothing but I've never felt the need to use balancers instead of manifolds. Takes time to saturate? And? you'll set it up once and walk away.
All the funky math around loop backs on balancers? slap a manifold on it and walk away lol
Before I see the revised rule I'll try to make my own: (it is easy to load balance a number of machines divisible by n^2 or n^3), please correct me if I am wrong.
Edit: I was close enough.
I thought for a moment that you had Drew Scanlon guesting on your video from that thumbnail.
For the overflow, if you don't need the additional overflow, could you just sink it? Like, if you had 120 output, but only needed 115, instead of looping back the extra 5, it'd just get sunk.
Definitely could do that! Nothing wrong either leaving extra in the system to smooth out any irregularities but if the belts being backed up would cause an issue, you should definitely sink the extra.
man i cant get my belts to work properly. i do exactly this, but the feeding belts keep stuttering still, and everything seems alright
I linked an imgur gallery of load balancer designs to the last vid, but it looks like my comment was removed before the last vid was taken down.
Are links like that not allowed?
Also, good vid. I've recently switched to using manifolds myself, but I find balancers to be more aesthetically pleasing.
RUclips probably flagged it if it had a link, so I did not see it.
7:20 Couldn't you do this for the 5 example too? Just split both of the input and the loopback belts in two and merge the respective halves before splitting each of those into 3.
my tip when load balancing is to use a notebook. I love math but i still write everything down to ensure i'm mathing right.
I use load balancing and underclocking for inputs to all machines, it makes me feel wasteful to let all machines load up
And use manifold only when produce has to be only stored and not sent to another machine
I just work with 2^n×3^m (usually 12) outputs and downclock my machines.
How come you had to reupload this one?
Someone was quick to point out that my math was STILL wrong about load balancing , so I re-recorded some information and re-uploaded it.
Why do you need two loopbacks on the example with 12 storage containers? Why not just send those two onward like you did on the right side? Doesn't 12 follow the "rule of 2 or 3" such that it doesn't need loopback belts?
Hmm I think maybe you misinterpreted something? The example you're referring to that has 2 loopback belts is 10 containers, as I was showing how to do a loopback system that doesn't follow the "rule of 2 or 3". But you need to round up to the nearest number that does follow the rule, which is 12 belts, then loopback those 2 extra belts.
@@SatisfactoryNews Sounds like the sort of silly mistake I'd make. This is why I suck at Satisfactory :P
Thanks for the clarification!
Don't worry about it! Lots of information was on screen and this game is confusing.
Great... I'm ahead of progress then UGH and last belt is on last 2 elevator upgrade, and I just spend hours on mini factories setup, luckily not too detailed, but still.
edit: figured it out, will do extra steps, but need shift the 1/5 later part.of the chain, phew
The one place load balancers are essential is nuclear power.
I see a lot of people asking themselves why anyone would use a load balancing system instead of a manifold.
I use it for my biofuel factory. I have 12 biofuel burners and with a manifold, the last machine runs out of materials while the first one still has a ton in storage. This means I get a power outage, while still having enough biofuel.
Great use for a load balancer, now that biomass burners have inputs!
One additional way to remove the manifold warm up is to just manually put a full stack in each machine.
Load balancers are for perverts imho.
That's an... interesting description.
Moral of the video, just manifold
Is this a re-upload?
Yeah, I had to fix a few errors with the math that were pointed out
load balancer fans jumping through an insane amount of hoops just to not make a manifold lol
omg, I love you
I just use Manifolds... they are so much simpler
Heres the best tip youll get:
Every machine in Satisfactory has a huge buffer. You don't ever need to balance anything.
Why does anyone even care to use a load balancing system?
It only really matters for power set ups without buffers.
It matters in some very particular cases, and it's a way to nerd out about math which some people like. I have exactly one load balancer in my game, it's really not needed in most cases.
Because they make for a satisfying factory.
@@RainbowGin But they look ugly and manifold is definition of conveyor factory - belt in, belt out
i am here at this video because i just made 6 nuclear power plants. and the first plant is filled up with materials but the last plant is still at 0. due to not all 6 plants receiving a 1 to 1 input. the machines previously in the chain are backing up on sulfuric acid because i set it all up to be a perfect 1:1 ratio.
load balancing would have fixed my issue but i went manifold
Just use a manifold, period. Set it up, and by the time you turn around do almost any other task, the manifold will be full and primed and all of your machines will be doing their jobs.
omg. the math.
Excellent Vid, thanx for the update / corrections! Now I have another wrench to toss into your machine! (please don't hate me!)
Your Load Balancer does NOT have to take a 'country mile' to set up!
ruclips.net/video/ILEY8Z46A4c/видео.html
or this one;
ruclips.net/video/MaFM0Dlp3nA/видео.html
Try these, they are a little complicated, but it is very tight and neat, and fortunately doesn't need to be rebuilt over and over, else a blueprint would be best.
2:29 Sinon tu met 6 Stock et finito
you did the ONE thing you must not do wit ha factory. create a belt swastika. i will NEVER take your advice on anything EVER. youve broken the cardinal rule. the one rule that must not be bent, let alone broken. i am severely disappointed.
Whoops! If it makes you feel any better, I think so little about nazi imagery that I literally didn't even realize I did that.