HN user

nsrango

29 karma
Posts0
Comments10
View on HN
No posts found.

General purpose digital computation is often power limited (either from the supply energy or thermals of the device) and there is an (academic) resurgence of analog computation on data prior to digital blocks to reduce the power footprint of a part. Most applications of this are RF. Conditioning signals as you say these days and in the future isn't just bandwidth limiting and shifting the baseband.

It depends on what you mean by fully formed. For example the idea of vectors and quaternions, which seem so natural, is quite new (~250 years old) [1]

When maxwell derived and unified electromagnetic theory, he didn't use constructs like the gradient and divergence of vector fields (those concepts didn't exist), instead performing those operations 'just' partial derivatives [2]. Sure, the math is explicitly identical, but the modern concepts of operators on vector fields that is so powerful just didn't exist which, to me, is rather telling about the evolution mathematical thinking: we are all doing the same thing (and have been for a long time) but way we think about it evolves with our notation. And notation that we are used to is actually quite new

[1] https://www.math.ucdavis.edu/~temple/MAT21D/SUPPLEMENTARY-AR...

[2] http://rstl.royalsocietypublishing.org/content/155/459.full....

With respect to CPUs this is not entirely true.

Processors with different clock frequencies roll off of the same assembly line but small variations in manufacturing change the quality of the yield quite a bit. This includes the side of the wafer a chip is on or how far the chip was from the center of the wafer. As a result, processor manufacturers test their ICs and sell them based on how each one performs