Tiny Chiplets: A New Level of Micro Manufacturing | Hacker News Reader

Tiny Chiplets: A New Level of Micro Manufacturing

nytimes.com

36 pointsrmah12 comments

I was going to submit this. Then I read it and realized it said nothing. And in a confusing way.

This is brilliant.

One problem with today's chip manufacturing tech is if you want to build system with multiple manufacturing technologies(digital/RAM/Flash/Analog) it increases your price by much because you have to package each part(which also limits your designs), test it and assemble it on a relatively large pcb. So in order to save money, you manufacture everything using the same un-optimal process.

So this is why microcontrollers like the arduino have relatively low memory.

Now if you could easily and cheaply assemble miniature electronic components - suddenly you get the optimal composition of your circuit, by picking an matching low cost components.

So you want to make an arduino cpu with megabytes of ram and flash ? now that's possible and very cheap(probably in the cents). Even something like microcontroller that could run linux and python would seem possible.

Can't wait to see it deployed.

One solution to this using conventional CMOS technology is to use 3D stacking, where each layer uses a different technology.

You're right , but that's a pretty difficult and expensive solution.

The more practical solution for now is 2.5D stacking - stacking all the un-packaged chips on silicon/glass die instead of a pcb. That's already being used in some fpga and communications equipment.

But 2.5d is still pretty expensive and complex. Not yet ready for low cost stuff.

This technology looks very intriguing to me but I'm confused about potential applications for it. That is, what's it good for?

If I read it correctly, essentially, tiny circuit boards full of components that can be printed. It seems like it would make the manufacture of electrical devices easier, and perhaps even more accessible to hobbyists, if the production machines and "chiplets" are affordable.

Lots of NYTimes articles on HN lately. While I think most of the topics are worthy of the front page, it really sucks when you hit their "10 articles a month" pay wall. Wish posters would look for alternative, free versions of some of these stories (I'm aware that might not always be possible).

The Times's paywall is so porous that this really isn't much of a problem. The content still loads behind the JavaScript overlay, so you can use Safari's Reader or a similar browser plugin to read the articles. You can also get around the paywall by reloading the page in a private/incognito window, etc.

I subscribe to the Times but have never bothered even to log in to the website on my iPhone, since I can always just click the Reader button—which I would do anyway—to read the article when I get paywalled.

If you are hitting their pay wall every month, may I suggest you subscribe to the Times? You obviously want to read their content at least that much. To get that kind of quality writing isn't free, the writers really have spent years on their craft (a close family member of mine is a journalist). I don't subscribe, but I typically am not interested in their articles more than a couple times each month.

>The emerging printing technology poses a heretical idea: Rather than squeezing more transistors into the same small space, why not smear the transistors across a much larger surface?

Did they forget about the speed of light?

This isn't for performance, it's for the capability at all.

Would you rather have:

1. No ability to create your own silicon logic outside of FPGA/CPLD chips or having an ASIC fabbed, all of which tend to be very high cost per unit and have large board outlines, and are limited in i/o by the predefined capabilities of the chip?

2. The ability to whip out cheap prototypes whose performance is nowhere near state of the art in a few hours with a laser printer style device onto substrates on an arbitrary surface.

New tech is frequently disruptive in ways different from how things were envisioned - for example, I doubt people reading Stephenson's "Diamond Age" would have envisioned the plastic PLA/ABS fiber printers as the first major incarnation of the 3D printing revolution.

I don't have a good mental model for this. What sorts of problems are more suited to a network of tiny ICs rather than a FPGA or a microprocessor?

Reply on news.ycombinator.com