Hacker News Reader

Comment by eitally | Hacker News Reader

That's a silly anecdote and I have a hard time believing chip fab line workers are too much different from those in other types of high-tech (where I spent 15 years working on test automation & quality systems). Some smart folks, for sure, but almost zero roles required even actual engineering background. A lot of management had strong technical education, but the vast majority of line workers are just following simple instructions.

I could be wrong about how fabs work, of course, and would love to learn more.

View on HN

Having seen a few 'high tech' places myself: there's high tech, and then there's _high tech_.

Assembling a motherboard is to semiconductor fabs as flying a toy drone is to landing a Boeing 747. One you can learn in an afternoon; the other takes years to learn.

You cannot learn to assemble a motherboard using modern manufacturing equipment in an afternoon. Pick-and-place operation is a week-long course alone (specific to that machine, and assuming you are already familiar with the process), and there are at least a screen printer and reflow oven involved as well.

Yea fair lol - def an exaggeration. (And my experience is def not first hand, here)

But still - a lot easier than semiconductors

One of the most critical aspects of engineering is making it so advanced high tech things can be manufactured by the least skilled and least number of skilled people possible.

Don't confuse that with the workers not being skilled - many are, but just like good software, you want your designs to run well even when executed by the worst hardware.

Actually engineering often means need more skilled workers as the unskilled work was replaced. 1 person running the machine who understands more of the engineering as opposed to 100 with a saw.

Making PhD engineers operate/maintain/monitor a wire bonding machine seems expensive.

I worked in an R&D fab (albeit, as an intern, many years ago) with an abnormal number of PhDs around (the fab was essentially a small scale test bed for new products and processes - everything tested there first before scaling up overseas). I think we had a 5:1 ratio of techs to engineers. This was in a fab that basically did nothing but R&D and low volume manufacturing for defense.

It's comparable to any kind of high-tech manufacturing where you have engineers designing/testing stuff (and designing the tools to make and test the stuff in the first place) while there have to be techs/mechanics/machine shop workers to actually do the work of making it. It's inefficient and expensive to put the engineers on that task, while they'll still need to do it from time to time.

> Assembling a motherboard is to semiconductor fabs as flying a toy drone is to landing a Boeing 747.

I disagree. I live next to a fab and have coworkers that have worked the fabs.

Humans in fabs are taken out of the loop as much as possible. That's because when dealing with nanoscale structures, human error is simply too common.

One of my coworkers worked at the fab during a period where they had humans running the forklift that moved the wafers from one stage to the next. That was cut out because the tiny bumps caused by a human operating the controls caused imperfections in the chips that decreased yield (the metric that matters most for a fab). They ultimately removed that work and job and replaced it with robots to carefully move the wafers.

What's complex about a fab ends up being not the frontline work, but rather the layer or 2 in the back (like designing the lithography filter for a given chip). That stuff happen outside the actual plant.

> Humans in fabs are taken out of the loop as much as possible.

This. We are talking about Atom scale here. I have as much admiration for skilled mechanics as the next guy. I heard about a lathe operator at Patek Phillip who could turn an arbor to within 1 micron precision, just by listening to the pitch the cutting tool made when trimming it down.

And when chip features were on the micron scale, humans in the loop made sense. But chip feature are ten thousand times smaller than that now--4 orders of magnitude. Anything that doesn't need a Ph.D. in solid state physics to do is going to be automated.

I'm probably revealing exactly which fab, but one issue the fab had is when it was built it was placed fairly close to the interstate. That did no matter when they were at something like 500nm. However, as they slowly pulled the node size down they started noticing random errors within their chips. It took a while to track it down but it turns out semi trucks driving past the interstate were causing defects in the chips.

They ended up installing shock absorbers everywhere to counteract this problem.

There are still, though, lots of working fabs at much larger process sizes. And some for things that won't ever modernize to a smaller one. Like parts that handle lots of power, or chips for RF purposes, etc.

I think we agree? The point is that a semiconductor fab (like a 747) is so highly automated that you don't need a bunch of low- or medium-skilled folks to drive forklifts; you need a few high-skills folks to design, monitor, debug & optimize the huge system(s).

I think you underestimate the amount of people with high school education doing very precise, skilled work right now. Do you think skilled machines tend to have PHds or even bachelors?

Im not saying its as simple as CNC. I am saying that I doubt you need an IQ of 130 or a PHD. Start investing in the people whose labor were undercut. In fact a skilled machinist or technician seems like a much closer match (in terms of skill and willingness to do the job) than some with an electrical engineering PHD or something.

Skilled machinists?

It takes education and practice to take a CAD model or technical drawings and run the machines to produce it.

Sorry, I was just trying to clarify what "skilled machines" meant. I think they meant "skilled machinists"?

> Assembling a motherboard is to semiconductor fabs as flying a toy drone is to landing a Boeing 747.

A 747 can pretty much land itself on autopilot.

What are the conditions and steps necessary to engage autopilot during landing? What are the failure conditions, and the steps necessary during a malfunction when autopilot is engaged?

On a clear day with perfect conditions ground control can probably get you on the ground and you walk away safely. The plane may not be safe to fly again, but you walk away. When there is bad weather (common) or other mechanical issues you need a lot more training, and of course if you want the plane to survive to fly again you want some more training.

The only hard part is contacting ground control in the first place. Radio frequencies change all the time.

https://www.washingtonpost.com/travel/2023/05/20/passengers-...

WaPo tried it out last year and only the fellow who was used to WW2 plane simulators was successful at a probably didn't die landing.

> Assembling a motherboard is to semiconductor fabs as flying a toy drone is to landing a Boeing 747. One you can learn in an afternoon; the other takes years to learn.

The hard one to learn is flying a drone, right?

I think people are also underestimating the intelligence of people in the once very large manufacturing workforce.

It doesn't help that the union culture is opposed to promoting smart people. If you work an assembly line you are not better than anyone else on the line, so their position makes sense - smart people are not worth more. However it also means smart people that want to move to management (want - if you are content that is your choice) find things worse: you are required to start at the bottom and your previous experience doesn't count for anything (including the pension)

The decline in American manufacturing happened alongside a decline in union membership in manufacturing. If unions were the causal factor, why do these trends coincide rather than oppose?

Alternative theory: Protectionism industrializes and trade liberalization deindustrializes. Alexander Hamilton's protectionist policies weren't just successful at industrializing the USA, he and his successors were correct as to why they were successful. The relatively recent shift to trade liberalization, by contrast, had the generally accepted effect of such policies on the industrial base. Of course, assets got pumped along the way and that's what really mattered to the people in charge.

Trade liberalization industrializes the world.

Unions were gutted because the labor market grew beyond international borders, but the labor movement didn't (thanks to western capitalists making war on socialists.)

the only cause here is higher wages make other countries more compelling.

The not promoting is a problem but it isn't the cause of decline.

> but almost zero roles required even actual engineering background

Small anecdote: pat gelsinger, intel ceo, started as a technician at intel (iirc) at like years old or something like that.

That entry level job paid enough that he could have a roof on his head and i guess covered his university tuition?

I’m not making this up: it’s all in his book he wrote about his life (not sure I recommend the book)

True, but Pat Gelsinger isn't a good example of what public policy we should pursue to help Joe six-packs enjoy a middle-class life. Gelsinger is one of the few people I would follow blindly into the maws of hell itself; his intellect is exceeded only by his ability to lead.

LTI and Santa Clara are not public institutions.

Have you worked with the guy, like, in person or are you just being an enthusiastic fan?

Just trying to understand your point of view.

I am a little bit fascinated by the way that many people seem to think so highly of Gelsinger, and I even think it is possible that he will succeed in leveraging that to make Intel competitive.

However, just from his public persona of wildly optimistic sales/marketing pitches and showmanship, while outsourcing their own manufacturing to TSMC, I have a very bad impression.

Then his claims of Intel zettaflop machines by 2027 honestly make me think he is totally delusional, but maybe that does not matter.

Just in case they are paying attention, I would love to relocate and go to work for Intel. Following simple instructions sounds terrific, to me. Where do I send my resume?

You're missing the broader point here which is that TSMC pays roughly $50k/year for people with these skills. Sure, you may be able to find the talent pool, but at what price?

I think an important question here is: how exactly can TSMC afford to pay $50k for people with these skills? And why is that impossible in the US? I think this comes down to relative cost-of-living differences: the US is just too expensive to live in these days. Taiwan is not a super low-cost-of-living country by any means, but the US has gotten ridiculously expensive compared to most other peer nations, and this means it's pricing itself out of the market with anything where it needs to compete.

Personally, I think America's reliance on cars and its suburban layout is a big part of this. Workers in Taipei don't need to own a car to get to work, and the overall infrastructure cost per capita is much lower.

Reply on news.ycombinator.com