Hacker News new | past | comments | ask | show | jobs | submit login
Inside the Intel 1405: die photos of a shift register memory from 1970 (righto.com)
62 points by jekub on Dec 21, 2014 | hide | past | web | favorite | 16 comments

Lovely stuff. I wonder how close you could get to making this stuff in your garage now, given a silicon wafer and maybe a vacuum chamber. (I imagine the etching chemicals would probably make you unpopular with your neighbors...)

She didn't do anything remotely this complicated, but a few years ago Jeri Ellsworth did do a couple videos of "homemade" ICs.


Really interesting how chips haven't changed much for almost 50 years - except for getting (much) smaller. This could have been a modern die photo except for the scale of things...

Uhm no, absolutely not. The basic operational principle remained the same, but absolutely everything has been changed many times over. Trillions of dollars were invested to make it happen.

The only constant is probably that it is still made on a silicon wafer. But even this wafer changed a lot, and I am not talking about its size.

Would you also compare a horse carriage to a Tesla?

Would you also compare a horse carriage to a Tesla?

I think it would be closer to a comparison between a car engine in the 1900s and one today. Lots of differences, but the basic principles remain the same and if you were to give someone from one time period a part from the other, they'd easily be able to see what it was.

No. Chemicals are different, structures are different, design is different (cmos), etching equipment is different. I am at a loss to name a recognizable thing.

The wafer is a slab of polysilicon, and has been that way for a long time. Most of those trillions you mention were invested to make the lithography smaller, which is how you get more transistors into the same die.

Edit: silicon, not poly silicon. Not sure why I said poly there...

No, simply no.

ICs are made of monocrystalline silicon, not polysilicon. There have been massive efforts to understand how to make it cleaner (internal gettering) and how to get wafer mechanics under control (Nitrogen doping). That does not even touch the changes in production technology to go from 10 mm Wafers to 300 mm Wafers. (You just make them larger, right...)

A lot of new concepts and materials have been introduced to ICs during the last decades. Especially since the 130 nm ground rule, lithography has not been the main limiter. Examples: Copper, Tungsten, Diffusion barriers, Silicide, Strained Silicon, High-k, metal gates, finfet, 3d integration etc...

10 mm wafers? Were they ever that small? I've never heard of less than 1 inch...

The transistors used today are very different from the transistors used back then, not just in size but also in structure and materials. The silicon die is pretty much the only part of the system that has stayed the same -- everything else has been changed, one part at a time.

You probably said "polysilicon" because this chip has some polysilicon deposited on top.

Getting a modern die photo requires using an electron microscope or a scanning-probe microscope, because feature sizes are down to about a tenth the wavelength of visible light. That alone makes the photos look pretty different.

Modern chips also have quite a few more layers (process steps); e.g. IBM's old 0.13μm process supports 8 metal layers, not counting other steps like epitaxial growth and oxides: https://www.mosis.com/vendors/view/ibm/8rf-dm

is there any good book to study about this ?

I don't know much about it myself, so I'm not the right person to ask. Much of the little I do know came from reading http://designinganalogchips.com/, a book by Hans Camenzind, one of the all-time great sandbenders.

I give a thumbs up for Designing Analog Chips - I found it really helpful and interesting, and it's free online at the above link.

And if you're interested in 1970s-era digital chips, I recommend Mead and Conway's "Introduction to VLSI systems" at http://ai.eecs.umich.edu/people/conway/VLSI/VLSIText/VLSITex...

Not sure what you mean by not changing much. Those are probably NMOS circuits. We use CMOS today.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact