TO BEGIN, let’s reflect on what Moore’s Law
actually is. It comes down to the disarmingly
simple observation that transistor densities in
integrated circuits double every two years. In
other words, Moore’s Law says that twice as
many components are squeezed into a given area of computer
chip every couple of years. This is the pace at which the chip
industry has progressed for decades and the result, in hardware
terms, has been spectacular, exponential growth in computing
complexity and capability.
If that seems to also imply a halving in cost over the same
time frame, the reality isn’t quite so simple. Certainly, Moore’s
Law proved accurate from 1975, when the co-founder of Intel,
Gordon Moore (the ‘Moore’ in Moore’s Law), adjusted his earlier
observational time frame from doubling every year to every two
years, until around 2010, when there were the first signs that the
wheels might be coming off one of the most remarkable runs in
engineering and economic history.
To illustrate that with some numbers, way back in 1970, Intel’s
first microprocessor, the 4004, packed 2,250 transistors, which
was impressive for the day. Intel’s latest desktop chip, Alder
Lake, contains over 20 billion transistors. Of course, you’d expect
today’s CPUs to be dramatically more complex than that 50-year-
old processor, so try this for a more recent and arguably more
brain-bending comparison. The Intel 486 CPU of 1990 boasted 1.2
million transistors. Roughly speaking, that means 20,000 486s
would fit inside the die space of a current Intel Core i9-12900K,
were the 486 built on the same 10nm production node as the new
chip. That’s a staggering notion for anyone who can remember
a 486 when it was the latest and greatest desktop powerhouse.
Or how about the first “Willamette” Pentium 4, from 10 years
after the 486? That was a 42 million transistor chip. Using the
same rationale, you’re looking at around 500 of those to rack up
the same transistor count as a 12900K. Even a fairly modern Core
i7 processor from 2010 would fit inside an Alder Lake die 30 times
over, again if it were produced on Intel’s latest node.Moreover, do the math on those 10-year transitions and you’ll
see that they mostly line up with Moore’s Law. From 1.2 million in
1990 to 42 million in 2000 is nearly bang-on doubling every two
years. From 2000 to 2010 and 42 million to 774 million, there’s a
little slippage evident. As for the 10 years from 2010 to 2020 (or
2021, at least, with the arrival of Alder Lake), the step up from
774 million to 22 billion is, again, pretty much in line with the
expectations set by Moore’s Law.
Of course, you could cherry-pick examples that wouldn’t fit
quite so nicely. Likewise, Intel is just one company, not the entire
semiconductor industry. The new Apple M1 Max chip, as found inEUV lithography should keep Moore’s Law going for a few more years.MooreNs law
30 MAXIMU MPCMAR 2022
© WIKIMEDIA, ASML, APPLEApple’s M1 Max chip packs around 57 billion transistors.