Is Moore's Law wrong, or is it wrong-headed (6th time around)

[jasong]

Correct me here, but I believe Moore's law "died" around the time companies stopped spending everything they could on fabs and started cost-cutting more on production.

It makes me wonder if we've been doing the math with the wrong terms. Do you guys think a best fit geometric curve would look accurate on a graph of

(total computing power generated) divided by (cost of fab) <---- x variable
(time) <-- y variable

[Brian-E]

Perhaps the answer is to be found in one of the six other threads which you have started about Moore's Law through the years?
http://www.mersenneforum.org/showthread.php?t=18885
http://www.mersenneforum.org/showthread.php?t=17637
http://www.mersenneforum.org/showthread.php?t=17167
http://www.mersenneforum.org/showthread.php?t=8259
http://www.mersenneforum.org/showthread.php?t=6952
http://www.mersenneforum.org/showthread.php?t=4602

[jasong]

Originally Posted by Brian-E

Perhaps the answer is to be found in one of the six other threads which you have started about Moore's Law through the years?
http://www.mersenneforum.org/showthread.php?t=18885
http://www.mersenneforum.org/showthread.php?t=17637
http://www.mersenneforum.org/showthread.php?t=17167
http://www.mersenneforum.org/showthread.php?t=8259
http://www.mersenneforum.org/showthread.php?t=6952
http://www.mersenneforum.org/showthread.php?t=4602

What can I say, I like the topic.

Btw, Brian, does my avatar make me more, or less, appealing? ;)

[LaurV]

Trying to combine intel's Tick-Tock model with Kondratiev's cycles, we are now at the end of a big "Tick". Things were shrinking and becoming smaller and denser, faster and less power hungry. But that's not the end of the "law", by far. They will become "multi", or "different". Wait for the "Tock".

[bgbeuning]

Skylake chip geometry is 14 nm (10^-9).
One copper atom is 128 pm (10^-12).
So the Skylake CPU use traces (wires) 109 atoms wide.

Will a 10 atom wide trace work?
I doubt a 1 atom wide trace would work but I am no physicist.

They have already started going vertical. Old school chips
might have traces that are 200 atoms wide and an unknown
(to me) number of atoms tall. Maybe a 5 atom wide and 100 atom
tall trace would work.

(This has probably been covered in one of the previous posts
on this topic. I have not gone back and read them all.)

[LaurV]

Or, we can say, a transistor is as big as the hepatitis B virus

[Mark Rose]

Originally Posted by bgbeuning

Skylake chip geometry is 14 nm (10^-9).
One copper atom is 128 pm (10^-12).
So the Skylake CPU use traces (wires) 109 atoms wide.

Will a 10 atom wide trace work?
I doubt a 1 atom wide trace would work but I am no physicist.

They have already started going vertical. Old school chips
might have traces that are 200 atoms wide and an unknown
(to me) number of atoms tall. Maybe a 5 atom wide and 100 atom
tall trace would work.

(This has probably been covered in one of the previous posts
on this topic. I have not gone back and read them all.)

If I'm not mistaken, one of the advantages of the vertical gates is less electrical leakage from quantum tunnelling.

[retina]

Originally Posted by Mark Rose

If I'm not mistaken, one of the advantages of the vertical gates is less electrical leakage from quantum tunnelling.

There is research being done into using the tunnelling, instead of trying to avoid it, as the basis of the transistor operation. Perhaps in a few years we will have QT transistors?

QT research is just one of seemingly thousands of other avenues begin investigated as successors to the current transistor designs. Perhaps the Star Trek isolinear circuits will be the new thing?

[kladner]

Originally Posted by LaurV

Or, we can say, a transistor is as big as the hepatitis B virus

Fantastic graphic! Thanks for posting it!

[Brian-E]

Originally Posted by jasong

Btw, Brian, does my avatar make me more, or less, appealing? ;)

Well, yours originates from Africa and mine from South America, so we are unlikely to have met each other in which case being "appealing" is out of the question. Your original cat avatar (at least, that's the earliest one I can remember that you had) was quite frightening to the guinea pig.

[jasong]

Originally Posted by bgbeuning

Skylake chip geometry is 14 nm (10^-9).
One copper atom is 128 pm (10^-12).
So the Skylake CPU use traces (wires) 109 atoms wide.

Will a 10 atom wide trace work?
I doubt a 1 atom wide trace would work but I am no physicist.

They have already started going vertical. Old school chips
might have traces that are 200 atoms wide and an unknown
(to me) number of atoms tall. Maybe a 5 atom wide and 100 atom
tall trace would work.

(This has probably been covered in one of the previous posts
on this topic. I have not gone back and read them all.)

Ahhhhhhh, but if we're willing to get into some ultra-weird physics, then we begin to realize that atoms have their own components. Put the right spin on things(pun intended) and you could have a one atom bit, as in a byte, on your hands.

Not that I could explain it. I'm wondering how hard a basic physics MOOC would be.(Forgot about Coursera)