Friday Facts #376 - Research and Technology

[FuryoftheStars]

It was a waste of time (for improving the system), because:

• The 35-page discussion was very low quality, mostly talking about unimportant things like the name and useless critique based on misunderstanding of the quality system. And it was mostly redundant. It wasn't really 35 pages of discussion, it was 35 pages of the same post redundantly posted in copies over and over from people who hadn't even read the thread (or blog post properly) to see that their view had already been shared by others.

Sooo... we should have one person that posts in favor of the system and one that posts against and just leave it at that? That doesn't seem helpful. I mean, there's no poll or anything, so even if several people share the same exact view, how else do you measure the quantity of likes vs dislikes? I can also understand not wanting to read through several pages to see if someone else posted the exact same thing that you're thinking.

Are you missing the point on purpose?
Yes not everyone reads all 35 pages before making their post. But you don't have to be that extreme to see that naming is commonly talked about.
Also you don't have to be that extreme about limiting the copies to one single time. You are just exaggerating because you have no argument.
Just read a few posts or pages, if something trivial has been talked about 7 times already then consider if you adding an 8th post on it really adds much value to the discussion or not.

It doesn't matter how many times it's already been said. If you also agree/disagree, then you should (be allowed) to voice your opinion on it. If you have a question about something, absolutely agree they should read through some of it first. But if it's simply a statement of agree/disagree, maybe throwing in their 2 cents of an alternative, then that's perfectly fine and they shouldn't be expected to read through it and silence themselves if it's already been mentioned several times.

Whatever arbitrary number you want to pick for the point where people shouldn't post any further (1, 7, or 100) simply cuts off further discussion/expressing of opinion and thus is not helpful to gauge how much something is actually liked or disliked.

• Wube already knows how to design game systems better than more than 99.9999% of the player population (they made Factorio after all, the best game).

This isn't a valid point. Not an attack on Wube, this is just a generalized statement, but the ability to program (and do it well) or lack there of has no correlation to the ability to have a good idea or not.

I'm not saying that everything they do is perfect, I'm not saying being better at games design always makes you get the 100% superior design idea every single time. I'm just saying that they are better at it (overall) and have actually tested it and also understand the system so 35 pages of mostly low quality posts don't indicate that Wube should change anything. The argument made by coppercoil was that the quantity of pages of responses should change Wube's opinion on their design.

And yet they can still get it wrong and the fact that we have 35 pages of discussion, regardless of "quality", should be an indicator to review to be sure. And there have been changes they've made to the game before that they thought would be good but due to subsequent discussions they reverted or changed further. So this "point" you were trying to make about how the discussions were a waste of time "because the devs know what they're doing better than the players" is invalid.

[dead-duck]

Grade 1
Grade 2
Grade 3
Grade 4
Grade 5

Kovarex said the criteria is feels good and is clear. I think the flaw with many proposals has been lack of clarity where exactly some terms like "improved" and "enhanced" would fall. Should be self evident. The numerical grades are self evidently clear and I think their impersonal, no-marketing-frills names feels at home in the practical world of Factorio. Nothing in factorio is made to look nice or like something for sale. It's all function.

[gGeorg]

No. You need just one splitter to peel off standard production of the higher quality production. Then direct the higher quality belt into the High-quality mall. And of course you need divert standard quality parts from High quality mall recycler back into mass pro. Which makes nice spaggeti challenge, with possible sushi on top. :=)

No, it's not. There's five levels of quality, and combining quality in a recipe essentially voids the higher quality item. Therefore in any recipe chain, products of each recipe should be sorted by quality and combined with other ingredients of the same quality. It's not something solvable by "a single splitter".

There are two tasks:
1. peel off standard quality from higher quality
- it is done on the mass-pro line by one splitter
2. sort different quality products
- it is done inside High-quality mall by 4 splitters ( optionally filter inserters & combinators ? )
- this task is done the same way in current design as in the proposed design >> in the High-Q-mMall

If you introduce quality in a production chain anywhere aside from the final product itself, you have to use it everywhere in that production chain. There's also a pitfall of mismatching quality proportions between production chains of different lengths, so naive approach would also fail. That's probably why devs imply that everyone just does recycle loops for the final product and not implements it everywhere.

No. When you recycle only the final products, as I reccomend.You get base and high-q parts inside the High-Q-mall. Then you re-use higher Q parts on spot (in the mall). Also, need transport base quality parts from recycler back to mass pro line.
This way we get higher complexity of production process, becouse you need mass production line to create the higher level of mid products. You can not make endless closed grind of single part. Which do not rise complexity of production flow.

[gGeorg]

Grade 1
Grade 2
Grade 3
Grade 4
Grade 5

- many proposals has been lack of clarity where exactly some terms like "improved" and "enhanced" would fall. Should be self evident.

- It's all function.

hit the bulls eye

[Panzerknacker]

Grade 1
Grade 2
Grade 3
Grade 4
Grade 5

- many proposals has been lack of clarity where exactly some terms like "improved" and "enhanced" would fall. Should be self evident.

- It's all function.

hit the bulls eye

Disagree, immersion plays a role for me, also in Factorio. Amongst many things, the graphics contribute to that, so do unique names for items and quality levels.

[FuryoftheStars]

You know, with semiconductor/microchip manufacturing being pulled out so much, does anyone have some source numbers showing high failure rates? Because a limited, cursory Google search is coming up with results that suggest the failure rates are often measured in (single digit) parts per million.

You’re Googling the wrong key words. That’s the defect rate of products delivered to customers, where the product doesn’t conform to its listed specs, and it indicates a failure in their QA process to detect defects. The rate of defects in manufacturing is much higher, but they usually catch it rather than shipping defective chips. A cursory search shows an article about Samsung reducing the defect rate on their 4nm process to below 25%. 25% is a far cry from 90%, much less 99%, but conceptually it’s reminiscent of what they seem to be going for with this mechanic. They will often stamp out a whole bunch of chipsets on the same wafer, and statistically some of them just won’t work, it’s just a question of how many, detecting which ones, and if they are salvageable or have to be trashed.

Ok, so changing my keywords from "failure" to "defect" helped. But I'm still getting some pretty good numbers? Not 100% sure as I don't typically look at this kind of stuff, so can't properly convert defect density into an estimate of final product failure.

• AMD 93.5% success (final product fully functional) with a defect density of 0.09: https://www.reddit.com/r/Amd/comments/e ... _and_full/
• TSMC’s N5 process has a defect density of 0.10 - 0.11, and they expect it to go lower once full scale manufacturing ramps up: https://www.anandtech.com/show/16028/be ... tes-for-n5

Both links are from a few years ago, too.

I didn't find the Samsung one you referenced, but off the bat, this tells me that Samsung may be producing some lower quality chips compared to these other two, and thus why the disparity in failure/defect rates. I'd liken it to Samsung making Factorio's "normal" quality stuff vs AMD/TSMC making "epic" or so.

----------------------------------------------

I feel like "modern day" space travel manufacturing is a bit different, simply because everything about it is freakishly expensive and there are often times actual lives on the line, so you always want to make sure you get it right the first time. But honestly, without reading the article and just going on your synopsis, this sounds more like the design phase and handcrafting the final product because you don't need them massed produced. If it ever becomes a more mainstream thing likes cars of today, I can totally see the attention to "only the best of the best" going to the wayside.

Ah maybe you would have had a different impression if you had read the article ! But if you are still ok to argue on side topics, then i could say than cars in factorio are not mass produced ! That would be more similar to F1 car, and they also take "only the best of the best".

Is it realistic to expect the highest quality everywhere as a standard ? That's never the case in real life where for most things you have more of "normal quality" or even "low quality" than you have of "high" quality. And in real life there exist some process where the result is not fully predictible and tests are done to select the "highest quality" amongst the production. ( semi conductors , rocket engines )

--snip--

(I am sorry, but that's way too much text for me to deal with right now. )

Yes, cars in Factorio are not mass produced, but seems the Quality system is being applied to (almost) everything across the board, I'm not sure we should be pointing to one offs for justification?

No, irl, it is not realistic to expect highest quality everywhere, but that's because of finances, not because the manufacturing lines of the highest quality stuff can't actually output them that reliably. Finances are not a concern in Factorio seems you're building it all yourself with resources you mine/gather yourself. As for semiconductors, please see my findings above in this same post.

[mmmPI]

(I am sorry, but that's way too much text for me to deal with right now. )

Yes, cars in Factorio are not mass produced, but seems the Quality system is being applied to (almost) everything across the board, I'm not sure we should be pointing to one offs for justification?

No, irl, it is not realistic to expect highest quality everywhere, but that's because of finances, not because the manufacturing lines of the highest quality stuff can't actually output them that reliably. Finances are not a concern in Factorio seems you're building it all yourself with resources you mine/gather yourself. As for semiconductors, please see my findings above in this same post.

It's ok i can wait till you read the message before you answer it, otherwise it's going nowhere.

[Khagan]

• AMD 93.5% success (final product fully functional) with a defect density of 0.09: https://www.reddit.com/r/Amd/comments/e ... _and_full/

Look further down that discussion. 93.5% are 'functional', but nowhere near all of those meet the highest spec; many are only usable at lower specs.

[Khagan]

I'm not looking to get in to a naming discussion right now, but I'll just point out that "Refined" is not an ideal quality term because there are other refined substances in the game.

Correct me if I am wrong but isn't it only 3 items that would cause an issue there: Refined Concrete, Refined Hazard Concrete and Oil Refinery.

I've always assumed that 'Refined Concrete' was just a Czech-to-English translation error for 'Reinforced Concrete'. The iron bars are a bit of a giveaway.

(This isn't a voice in support of "Refined' as a quality level. It might make some sense as one for raw materials, but not for finished products.)

[FuryoftheStars]

• AMD 93.5% success (final product fully functional) with a defect density of 0.09: https://www.reddit.com/r/Amd/comments/e ... _and_full/

Look further down that discussion. 93.5% are 'functional', but nowhere near all of those meet the highest spec; many are only usable at lower specs.

Ok, yup. I missed that. It would be good to have some actual numbers, though, with links.

Reading through that more, though, it does feel like in many ways semiconductors are a bit of a gamble. That said, these things I've found still make me feel like they're worlds better than what the devs are doing for Quality. But, that still only really applies to semiconductors, which the green/red/blue circuits are the closest analogies of. Plates, wires, gears, pipes, even belts, inserters, assemblers, and the other dozens of products in the game, however, are not like that irl (or at least when compared to their closest equivalents).

------------------------------------------

(I am sorry, but that's way too much text for me to deal with right now. )

Yes, cars in Factorio are not mass produced, but seems the Quality system is being applied to (almost) everything across the board, I'm not sure we should be pointing to one offs for justification?

No, irl, it is not realistic to expect highest quality everywhere, but that's because of finances, not because the manufacturing lines of the highest quality stuff can't actually output them that reliably. Finances are not a concern in Factorio seems you're building it all yourself with resources you mine/gather yourself. As for semiconductors, please see my findings above in this same post.

It's ok i can wait till you read the message before you answer it, otherwise it's going nowhere.

Yeah, sorry, it's still going no where. I have no idea what you were trying to say in that wall of text. Aside from what I've already answered, you talk about specialized application testing processes for concrete (which is well above and beyond the Quality system the devs are introducing), different game mechanics other than what we're talking about, and something about forum post sample sizes that isn't relevant to the point I was trying to make to Qon. I pretty much stopped reading at that point, sorry.

[mmmPI]

Yeah, sorry, it's still going no where. I have no idea what you were trying to say in that wall of text. Aside from what I've already answered, you talk about specialized application testing processes for concrete (which is well above and beyond the Quality system the devs are introducing), different game mechanics other than what we're talking about, and something about forum post sample sizes that isn't relevant to the point I was trying to make to Qon. I pretty much stopped reading at that point, sorry.

TL DR: you keep arguing about how it's done real life which doesn't matter since it's a game, it's symbolic, abstract, it's not made to model the reality perfectly and maybe its just a lack of knowledge that prevent you from seeing the parralels because you don't read the articles others or even yourself posted while still being the one complaining that the discussion is called "of poor level".

It's much more rationnal to trust people that worked for years on video game than someone who post his own idea he got while reading the FFF. Because they had time to try ideas, and discard those that leaded to no-fun gameplay. ( haha high quality is random in video game too, but you can influence it by repeating a lot the random process of "getting an idea",testing, and repeating which none of the readers of the FFF did).

( Khagan spotted a thing, but if you read the whole discussion you'd realize that intel's rate of success for 10nm are mentionned as "probably abysmal", and why do they even call this a chiplet ? the very concept of which seem to me to be a modular engraving of 8 core so that it can still be sold under a different name if only 6 or 4 are functionnal which cannot be more explicit in showing the whole process is difficult to predict... Random)

[Zabronax]

Thanks for the section about the quality feature
And I am looking forward to what mod authors will do with the research event triggers

I do think that linking the Quality feature tighter with Quality Control / Quality Assurance used in manufactoring processes is the way to go.
It will probably lead to more "Oh, this makes sense" moments, rather than "I don't like nondeterministic games".
ie:

• Manufactoring processes are imprecise, but you might get lucky

• Higher tech level -> More precise tools -> Higher quality products

• Higher tech level -> More precise measuring tools -> Tighter specifications

• Naming scheme could than be how far the item deviates from (or conforms to) the specification

It might be fun to have the Quality Control as an explicit step for discovering the fitness of an item, with the factory providing some sort of baseline. This would make it more of an opt-in for when I actually care about a higher quality product / have a tighter specification. Depending on how it's implemented, it could make filling trains full easier, by not forcing you to filter out higher quality items.

[Yoshifizzle]

You know, with semiconductor/microchip manufacturing being pulled out so much, does anyone have some source numbers showing high failure rates? Because a limited, cursory Google search is coming up with results that suggest the failure rates are often measured in (single digit) parts per million.

You’re Googling the wrong key words. That’s the defect rate of products delivered to customers, where the product doesn’t conform to its listed specs, and it indicates a failure in their QA process to detect defects. The rate of defects in manufacturing is much higher, but they usually catch it rather than shipping defective chips. A cursory search shows an article about Samsung reducing the defect rate on their 4nm process to below 25%. 25% is a far cry from 90%, much less 99%, but conceptually it’s reminiscent of what they seem to be going for with this mechanic. They will often stamp out a whole bunch of chipsets on the same wafer, and statistically some of them just won’t work, it’s just a question of how many, detecting which ones, and if they are salvageable or have to be trashed.

The term to search for is manufacturing yield and it can be significantly less than 100%, especially for the newest technology nodes. Think ~30% for a leading edge node (e.g. 3 nm) and ~80-90% for a mature technology (e.g. 20 nm).

[FuryoftheStars]

Yeah, sorry, it's still going no where. I have no idea what you were trying to say in that wall of text. Aside from what I've already answered, you talk about specialized application testing processes for concrete (which is well above and beyond the Quality system the devs are introducing), different game mechanics other than what we're talking about, and something about forum post sample sizes that isn't relevant to the point I was trying to make to Qon. I pretty much stopped reading at that point, sorry.

TL DR: you keep arguing about how it's done real life which doesn't matter since it's a game, it's symbolic, abstract, it's not made to model the reality perfectly and maybe its just a lack of knowledge that prevent you from seeing the parralels because you don't read the articles others or even yourself posted while still being the one complaining that the discussion is called "of poor level".

I keep arguing how it's done irl because others keep trying to say this is how it is irl. Which it's not. Semiconductors are the closest to that, but ultimately aren't really and are represented in the game by 3 things out of dozens, all of which are very different in how they are manufactured compared to semiconductors.

It's much more rationnal to trust people that worked for years on video game than someone who post his own idea he got while reading the FFF. Because they had time to try ideas, and discard those that leaded to no-fun gameplay. ( haha high quality is random in video game too, but you can influence it by repeating a lot the random process of "getting an idea",testing, and repeating which none of the readers of the FFF did).

To go with the generic, I've tried out plenty of games that had over a decade of development time that turned out to be pretty bad for one reason or another.
To go with the specific, these devs have made a number of decisions people have disagreed with over the years. Some the devs relented on. That's part of how we got such a good game for v1, is it not?

( Khagan spotted a thing, but if you read the whole discussion you'd realize that intel's rate of success for 10nm are mentionned as "probably abysmal", and why do they even call this a chiplet ? the very concept of which seem to me to be a modular engraving of 8 core so that it can still be sold under a different name if only 6 or 4 are functionnal which cannot be more explicit in showing the whole process is difficult to predict... Random)

Yes, I saw this. This is because of Intel's process being less modular than AMD's, so they are running a higher risk of not being able to get their final product to specs. The rest, I already covered above.

[blazespinnaker]

Technically speaking, anything that does not meet the required specs and either goes back to be recycled or is rebranded as a lower tier product is a "failure" for the specs that it was being manufactured for.

Heh, well, even quality has varying levels of quality.

[mmmPI]

I keep arguing how it's done irl because others keep trying to say this is how it is irl. Which it's not. Semiconductors are the closest to that, but ultimately aren't really and are represented in the game by 3 things out of dozens, all of which are very different in how they are manufactured compared to semiconductors.

In the game circuits are made of copper, iron and plastic, when in real life it's a bit more complex no ? That's not a problem as much as the "quality" feature ? the devs took some decisions to make a game, like when they simplified the early oil processing and it sparked controversy on the forum or as someone posted in the previous 35 pages " a storm in a bottle " , or " the ultimate bikeshed ".

I am going to say for the 3rd time that maybe if you were to read the articles the "others" posted, you may have changed your mind. For concrete you need to produce and wait for long time to improve the mix. That would be ridiculous and no-fun in a video game like factorio, therefore some simplification are made, because you don't realisticly expect that every single of the 50+ item there is in factorio has a way to be recycled EXACTLY like in real life do you ?

I personnaly like playing with mods that adds bakelite or epoxy or gold nickel tin manganese neodymium lead silicon hydroclhoric acid and others, that made me wanting to learn about those, it's more "realistic", it doesn't that mean it would be a nice addition for every player, it's a niche gameplay.

Plus i'm sorry to insist but in real life the process you say it's not like in the game , but i think it is, that's just perception it's not EXACTLY like in the game. It's endless.

( Khagan spotted a thing, but if you read the whole discussion you'd realize that intel's rate of success for 10nm are mentionned as "probably abysmal", and why do they even call this a chiplet ? the very concept of which seem to me to be a modular engraving of 8 core so that it can still be sold under a different name if only 6 or 4 are functionnal which cannot be more explicit in showing the whole process is difficult to predict... Random)

Yes, I saw this. This is because of Intel's process being less modular than AMD's, so they are running a higher risk of not being able to get their final product to specs. The rest, I already covered above.

Well yeah, it seem to be the new norm since 2017 and when AMD was able to catch up on intel that was using "monolithic model" were the rate of defect were so high, that they couldn't really progress in their CPU power. So really the reason they couldn't improve in quality was the random dust or little electromagnetic variation from things like solar waves and other RANDOM process that created defect.

Which was solved by creating "modular component", the "chiplets" instead of "monolithic model". It helped massively AMD in at least 2 different ways, 1) reducing the cost by being able to use a higher % of the input, ( by selling CPU with diffferent amount of cache for different price when there was not the full cache available due to defect ). And selecting the "good pieces" that had a lower percentage of chance of being made, randomly to sell them at a higher price than the others. ( I see many similarities )

2) making faster CPU, not just cheaper, due to the ability of combining chiplets to create a single processor, making it possible overall to assemble something of higher quality than what was possible with another way of doing. ( they added quality modules in the process step ? ).

Also, that's not just CPU, we mentionned rocket engines earlier, and i remember learning that : "Why do they use 30 + engines in their spaceX rocket ? "
Because it's statistical that the rocket will launch even if 1 or 2 or 3 engine have defect and can't turn on, in which case they also turn off the twin engine, located in the opposite diameter to balance out the thrust.

That's the idea that you remove randomness and drive quality up by multiplying the attempt to the point where things are statistics congrats, that's a very subtle introduction to megabasing !

[AvengerStar]

In the game circuits are made of copper, iron and plastic, when in real life it's a bit more complex no ? That's not a problem as much as the "quality" feature ? the devs took some decisions to make a game, like when they simplified the early oil processing and it sparked controversy on the forum or as someone posted in the previous 35 pages " a storm in a bottle " , or " the ultimate bikeshed ".

The "ultimate bikeshed" comment was from this thread, and the context is as follows:

In my opinion discussing the names of the qualities is the ultimate bikeshed. The names can be very easily changed; what about the mechanics?

Read as is, it was referring to how discussion on the topic of quality, for the most part, hyperfocused on the names on the quality tiers, something extremely trivial to change and therefore easier to conceptualize alternatives for than adjusting more complicated matters, in this case the mechanics. Hence why such was called a bikeshed.

The "storm in a bottle" comment was from the previous thread, and it implied that the massive backlash was without merit because a lot of people were quick to complain about quality. Which wasn't a very a useful argument, it was rather reductive and addressed absolutely nothing about why any of it was supposedly mere hysteria as opposed to legitimate concern.

As for this current discussion, I think there's some sort of miscommunication going on. It appears that everyone participating agrees that this game, much like a lot of games, abstracts some concepts for the sake of the player and/or the developer. The idea that I believe Fury is attempting to express is that just because there are realistic grounds for a particular direction in design, doesn't necessarily mean it has to be replicated in some or even any way in a game.

Rather, that defense of quality often defaults to it being somewhat a realistic implementation of real-life manufacturing processes, but that on its own doesn't address whether or not it would be good for the gameplay loop of the mid-to-late-game.

[FuryoftheStars]

I keep arguing how it's done irl because others keep trying to say this is how it is irl. Which it's not. Semiconductors are the closest to that, but ultimately aren't really and are represented in the game by 3 things out of dozens, all of which are very different in how they are manufactured compared to semiconductors.

In the game circuits are made of copper, iron and plastic, when in real life it's a bit more complex no ? That's not a problem as much as the "quality" feature ?

???????

You've completely missed what I said. And the rest of your post is about specialized/niche applications.
(Edit: you're going on about how some things irl are like this, but then turn around and pull out examples about how the game isn't like rl in order to counter my arguments about how rl isn't actually like that?)

As for links, as far as I can recall, aside from the two I provided, there's been one other.


Edit: Let me see if I can clarify this some by generalizing the discussions that sparked this:

• Devs: release FFF with Quality mechanic using randomization
• Some players: don't like the randomization for various reasons
• Some other players: semiconductors are like this and so are some of these other things
• Me: while I believe semiconductors are not to this extreme, they are also a small fraction of everything represented in the game, and the other irl examples supporting randomization are specialized, niche applications or have nothing to do with the mechanic the devs have introduced

This "irl" discussion is merely counter-arguing another "irl" argument. It's not a core reason I do not support the randomization method.


Edit2: Your concrete example, for example, is not an example of quality as it appears to be implemented. It's an example of generalized product improvement. This would be best represented in game if there were researches that could improve the bonuses given by quality.

[adam_bise]

Lol you guys sure have a lot to say about quality when it's an optional mechanic. Do you think it will not be fun to use? Comparing to advanced real-world manufacturing statistics seems extreme. There is no real-world comparison for transmitting "speed" to machines using radio waves, but everyone is fine with that.

[FuryoftheStars]

Lol you guys sure have a lot to say about quality when it's an optional mechanic. Do you think it will not be fun to use? Comparing to advanced real-world manufacturing statistics seems extreme. There is no real-world comparison for transmitting "speed" to machines using radio waves, but everyone is fine with that.

It's "optional" to an extent. As per the dev's own statement in the FFF, if you're just looking to "finish" the game, then you never need to touch it. If you want to keep going and get bigger, then it's not so optional anymore.

But regardless of how optional it is, this is still something that seems like a major feature of the expansion. I don't want to have to ignore a whole major feature of something - something that I'd have to pay good money for - just because I don't like it. It's like being offered to buy a bag of fruit or something that has fruit in it that you don't like.

But I'm not arguing to convince the devs. I've already said my piece there. I'm arguing about how "optional" or "semiconductors are like this" are being thrown around as if these invalidate my (and others) dislike of the new system.