[Syonyk]

> So how long does the microchip shortage continue?

Warning: Paid Pessimist point of view (I've done computer security for far, far too long).

Indefinitely. It doesn't recover. Because (modern, leading process) microchips are the sort of globe spanning supply chain problem that works great, right up until it doesn't. Coolant for some laser is made in this country that's at war. Some critical surface processing chemical is made by that company that had a major power outage and their facility caked up with goo that has to be scraped out. Some semiconductor fab has a worker shortage. Or a water shortage. Or got blown to pieces when some other country tried to invade (TSMC would be nuts if they weren't wired with demolition charges and quietly sure China knew it).

There will always be "something" that interferes with the smooth generation of chips. And there will be some supply, but not as much as people hope for, and they'll be rough around the edges, and, besides, you can't get those surface mount resistors in any quantity right now for your design. And eventually, this starts impacting the funding of the companies working on the newest, latest and greatest process tech, and they don't have the money to continue pushing forward.

We're long past the point when anyone but the most well funded multinationals can even consider a leading process tech foundry - and there are only a few companies on the planet who can make the hardware for it. The supply chain for foundries is just as bad as anything else, if you have a couple billion to drop on one.

And at some point, enough people will find workarounds that don't involve modern silicon that (at least in my arc of the future) demand will drop, so you won't be able to justify the investment in a leading edge fab much beyond the current stuff. Throw a solid recession/depression in, and consumer electronics spending is likely to drop substantially - so there goes a lot of the leading edge demand for some while.

The question, "How much damage can an efficient, just in time optimized global economy take before it fails?" has been discussed for many, many years, and the pessimistic point of view as of now would be, "Less than it's taken and is going to take in the next few years." And at some point, things like "food" and "basic energy" are more important to keep working than the latest consumer toys.

I'm entirely aware this doesn't paint a pretty picture of the future, and it certainly doesn't involve us going to the stars in any quantities, outside perhaps some token boots on Mars. But it's the sort of ragged decline we've seen throughout history, and, I'd argue, that we're firmly in the middle of right now. "Rattling down the backside of the arc of empire, undergoing catabolic collapse" seems to better predict things than a lot of other mental models lately.

[cowtools]

I mean it's not like most consumer hardware has to be on the latest nodes. The current hardware is more than enough for any consumer applications I can think of.

Maybe just need to reuse more hardware. Standardize and commodify replacement parts for mobile devices like framework/fairphone. Reduce the amount of IoT crap in toasters/etc. Use multiseat instead of thin clients. Upcycle old computers.

[Syonyk]

> I mean it's not like most consumer hardware has to be on the latest nodes. The current hardware is more than enough for any consumer applications I can think of.

Electron says, "Hold my beer!"

I would love to see lower software requirements, more long lived systems, etc.

Meanwhile, Microsoft is obsoleting... basically the last decade+ of hardware for Windows 11. It's an uphill fight.

[zdragnar]

Meh, I keep multiple electron apps, node processes, and chrome windows and tabs open at the same time, and they're really not at all CPU intensive (except for node running tsc builds or tests). It's the RAM usage that is annoying, and even that isn't too terrible.

Edit: of course, I'm running swaywm on linux, so my environment is generally pretty light as it is compared to Windows.

[Syonyk]

I generally try to use ARM small board computers as much as possible. They really struggle with running a bunch of Electron stuff at the same time, and while you can reasonably use some modern websites on them, it's amazing just how much absolute crap you can get rid of with NoScript/Ghostery/etc, and still have a perfectly functional website.

My massive, powerful, "Do absolutely everything!" computer in college was a dual Pentium III 866, with 768MB RAM. And a hex core, 2GHz ARM box with 4GB RAM can't do half as much, because of just how heavily bloated software is. I still write code, browse the web, chat with people... and I need radically more resources because that's what software development has decided is easy. Meanwhile, stuff like Hexchat for IRC just... uses no resources and works as wonderfully as ever.

[zdragnar]

I'm guessing RAM is your limiting factor. I've got about 20 tabs open across two chrome windows (separate profiles) and I'm using less than 3gb for my entire system. CPU usage is basically nil since I'm not doing much at the moment but typing.

Throwing slack / spotify / vscode / steam / whatever else on top of that could easily bump it up several more gb. VScode is pretty good, up until you start adding in language servers. The quality of the rest is really hit or miss.

Edit: Out of curiosity, I opened steam, and it immediately took up a little bit of ram, then by the time it finished "loading" it ballooned up to a gig- presumably because of all the fancy animated images and video players and what not the storefront needs. I could definitely do without those things, though I imagine they do correlate to more purchases (there's a name for it, I forget- something about moving images triggering a capture of attention going back to hunting / fight or flight response).

[ticviking]

I strongly suspect that we have stopped trying to reduce bloat. The cost in engineering time cannot be sustained. As a dev I can't even afford the time to optimize my local development machine to reduce bloat.

[riffraff]

Multiple electron apps run just fine on my 2015 MacBook pro too. Yes the ram usage is annoying but it's not unusable.

OTOH my battery is dying, and may not be worth replacing.

[Starlevel001]

> I mean it's not like most consumer hardware has to be on the latest nodes.

Maybe developers should take some responsibility and stop fucking writing code that's slow as shit and unusable on anything past their company-paid for macbook pros.

In an ideal society programmers would be sat down and given a seven year old mid-range laptop and threatened that if their program lags in the slightest they would be fired and blacklisted from the industry.

[Syonyk]

Nah. "Raspberry Pi Day at Google." One day a week, your dual Xeon workstation is replaced with a Raspberry Pi 4. 8GB. Overclocked if you want.

It is absolutely adequate to do a lot of things one wants to do. But it will show when you've done something stupid with CPU.

I'm still bitter that Google ruined the Blogger editor interface. Fancy, shiny new interface... that lagged horribly if you had a low power CPU and a bunch of photos in a post. The old interface handled it perfectly, because I wrote an awful lot of blog posts on an old Atom netbook with a nice keyboard.

But, yes, any new hardware performance is more than chewed up by new software abstractions.

[eastbound]

Well they’ll just limit table rows to 6 instead of 10.

That’s what React does: Instead of showing a table of 200 rows, the performance of React is so poor that it’s a design trend to make you paginate through 10-line pages. So, soon 6, “for better performance”.

[krisoft]

Pay for it my dude. Software developers are perfectly able and willing to develop efficient and snappy software if given enough time to do it.

The reason it is not happening is because market forces don’t favour those solutions. But since you seem to care enough to threaten, fire, and backlist them surely you care enough to pay for the craftmanship required to get what you desire.

[ticviking]

This is the problem, and we can't actually do this because we have a shortage of even basic "blub programmers" who can get business logic correct. Developers who are able to produce basically correct simple code are expensive, how much more expensive are people who are able to build low-level secure and optimized code?

[VLM]

Roll back the tech level and go vertical silo.

The auto industry buys bolts and makes pistons because the marketplace can't make pistons, roughly. They have the know how if the bolt marketplace collapsed to turn steel rod into bolts, maybe not as easily and cheaply as it can be done now, but if the bolt market died they'd make their own bolts.

The commodity microcontroller is dead. My guess is crappy "homemade" older-gen FPGAs made by GM for GM products will be the future of automotive ECUs and other automotive apps.

In 1980, $100M built you an entire fab. It costs GM about $300M to remodel an old assembly plant. They can either go out of business because 2022 chips are unavailable or build their own fab. Hmm I wonder what they'll do?

The nice part about building an older gen FPGA is it quite accurately emulates an older gen chip, usually using a lot more power and requiring a lot more silicon, but at least it works better than "next estimated shipping date 2024"

[bluGill]

A fab is $5billion. I know of a large company (I'm not allowed to say who, but you might be able to guess) that when faced with some 16bit CPU going out of production considered building a fab - manufacturing is the core competency, so why shouldn't they make our own chips, and thus keep those old designs based in obsolete CPUs that still work just find running for longer. The cost of a fab was high enough they decided not to. Of course on hindsight if they had built a fab it would have opened just as the supply chain problems started, and the company would have made a mint.

I don't think GM will open a fab on their own. Too expensive for what they need. However a joint GM/Ford/Toyota venture seems possible (if it can get past anti-trust laws!), and that would find customers in the likes of Honda.

[mlyle]

> Indefinitely. It doesn't recover. Because (modern, leading process) microchips

Most of our current supply chain problem is on trailing nodes.

Sure, there's not quite enough capacity on the leading nodes: there never is. It's a bit worse now than usual, but...

The big thing that's new here is increased demand plus some disruption on production has shown how little excess production exists for microcontrollers and various low-end ICs in consumer and industrial goods. It's difficult to justify building additional production on mid-end nodes, as margins are likely to fall back down in the future... but one can imagine various ways this nets out OK (e.g. countries deciding domestic production of these items is critical for national security reasons and supplying subsidy).

[fswd]

So basically... Unrestricted Warfare