Ridiculous trivia.........

[plexi12000]

So, i'm reading an article about how transistors work, etc., etc.  It goes on to say, the modern computer chip, the size of a fingernail, may contain between 500 million and 2 BILLION transistors.

That is just insane...at least to me!!  How in the H*LL do you make a machine that can make that miniscule stuff?!

technology...mass production....etc., simply amazing and fantastic.

[karbomusic]

Oldie but a goodie...

[mth5044]

Moore's Law!

[MrStab]

My now-dated Intel i5 2500K has 1.16 billion transistors on it, so go figure. That doesn't mean all those transistors work, though. I always went with Intel for my past overclocks because they had the stronger products at the time, but apparently what AMD used to do (or still do) was lock off faulty cores on chips and market them as lesser models. Some folk would unlock those cores.

If you think about how sensitive a MOSFET is in pedal-land, it's just unlikely that that high a number would ALL be functional. I don't know how it works and how they avoid constant crashes etc., but surely it's accounted for. With multithreading, transistor count is probably a bit less relevant than it used to be. My point is, i guess, that it's a tad less-perfect a process than you might think. just a tad.

It's you guys' fault i'm no longer a computer hobbyist/gamer. Seriously. Luring me in with all those circuits. tsk.

[PRR]

> How in the H*LL do you make a machine that can make that miniscule stuff?!

Find a film camera. Take a picture of your family. Your 6-foot brother comes out on the negative just 1/4" high. With a different lens and film, he might be 1/100th of an inch tall. With good lens and film, many of his details will be there in the image.

Historically, we laid-out the chip table-size, and used a series of camera reductions to get an image on the Silicon.

And historically the process worked fine unless there was a "defect" in the Silicon. We want a Single Crystal with all the atoms lined-up in perfect rows. 40 years of intense research has made that almost happen.

Some years back, a new CPU might have a "yield" of 10% the first week but >50% after some months adjusting the process. Because the quantity of Silicon is very tiny, you can make big money if half your CPUs come out right.

In the last decade the size of the transistor-bits has become smaller than a ray of lamp-light. They shifted to blue and then UV to get the small detail; they may be further-out by now.

The image on the Silicon is in a "mask". The light parts wash-out. Then you wave it in hot steam (Oxygen), Arsenic gas, Aluminum vapor, layer by layer, until you build-up the desired transistors and wires. A little like 3-color printing, but much more fussy and nasty.

> That doesn't mean all those transistors work

In most CPUs, in most engineering generally, ALL the parts work. If you can sell it with a bad part, then you didn't need that part, so why waste time making it?

The mass-production leverage of Silicon processing invites tricks. I hear that some chips have enough extra parts that many defects can be worked-around by re-wiring at the test level.

That's different from AMD's habit of working on the BIGGEST (many-core) CPUs, and re-badging the rejects as less-core CPUs. On one hand, this smacks of Plymouth making a V-8, but one bad piston, and selling it as a "V-7". (We had one, though it said V-8 on the fender.) When Chevy wanted a V-6 as good as their V-8, they just made 3/4 of the V-8, didn't bother making extra parts. 150 pounds of iron is not to be sneezed at. A quarter of a pepper-size chip patched-out is not a big problem.

[plexi12000]

PRR......  you and some of these other fellas are some smart Mo-Fo'S!!  hahaha  Obviously, you have a naturally talent for it all.  Do you work in that field?

[Rob Strand]

Transistor have been smaller than the wavelength of light for some time.

[samhay]

> Transistor have been smaller than the wavelength of light for some time.

Indeed. Single-molecule transistors aside, we are now at the 10's nm scale (visible light is about 400-700 nm). As I understand it, things start to get interesting now as tunnelling and then eventually the Compton wavelength should put a stop to Moore's law unless we can find something other than electrons to push around in these things.

[amptramp]

I used to work in cockpit display technology where we used cadmium selenide thin-film transistors in LCD displays.  There were lots of deposition steps but it was all done with optical masks.  Cadmium selenide was championed by Dr. T. Peter Brody, the man who invented the phrase "active matrix".  Since CdSe transistors can withstand hundreds of volts, the first thing he made was an active matrix electroluminescent display running at two hundred volts.  He eventually added a television front end and created a black and greenish yellow television as his first project.  Due to the lack of cleanliness in his prototype process, there were numerous lines out, but he still had the unit at his house in Pittsburgh where I saw it in 2001 and it was fully operational.

[mth5044]

I agree with Hermandusa.