Toward more efficient computing, with magnetic waves(news.mit.edu)
news.mit.edu
Toward more efficient computing, with magnetic waves
http://news.mit.edu/2019/computing-magnetic-waves-efficient-1128
3 comments
This work is part of the spintronics field. There are some good youtube videos about it:
TEDxCaltech - David Awschalom - Spintronics: Abandoning Perfection for the Quantum Age https://www.youtube.com/watch?v=g9NAUPbqAvE
TEDxCaltech - David Awschalom - Spintronics: Abandoning Perfection for the Quantum Age https://www.youtube.com/watch?v=g9NAUPbqAvE
If you can get an event that needs a wave above a certain height to trigger, you can make a gate by having input of waves a fraction of that height that, when they constructively interfere, exceed the threshold.
If your logic is based on wave interference, you may be able to get a lot of interesting behaviors even more useful than primitive logic gates.
If your logic is based on wave interference, you may be able to get a lot of interesting behaviors even more useful than primitive logic gates.
https://arxiv.org/ftp/arxiv/papers/1901/1901.08934.pdf - "Fundamentals of magnon-based computing" (PDF)
This seems to me like very important work.
Besides the obvious potential for memory devices, logic and even analog computation seem possible.
An inverter is the first bit of logic to perfect. Usually an AND or OR is not hard to get once you have an inverter, and once you have either, you have Turing completeness.
That is aside from any possible qubits.
Besides the obvious potential for memory devices, logic and even analog computation seem possible.
An inverter is the first bit of logic to perfect. Usually an AND or OR is not hard to get once you have an inverter, and once you have either, you have Turing completeness.
That is aside from any possible qubits.
Only peripherally related: As we keep increasing the percentage of the chip that is memory, I can't help but recalling all that talk years ago about storing multiple bits per structure.
This has shown up in SSDs, but not in cache memory. Are we going to skip right over that plan and go to spintronics?
This has shown up in SSDs, but not in cache memory. Are we going to skip right over that plan and go to spintronics?
3 bits per cell, a year ago: https://ieeexplore.ieee.org/document/8544012
Time to market is ¯\_(ツ)_/¯
Time to market is ¯\_(ツ)_/¯
Maybe with chiplets things like this and async circuits will have more of a chance to shine.
Quite a few such things popping up recently, looks like alternative/natural computing is trending.
I wonder what the reasons for that are - Moore’s law drying up?
You can read more here https://news.ycombinator.com/item?id=21665931 and (OC) https://juretriglav.si/computing-with-nature/.
Any other interesting sources?