The processes to manufacture CPU’s are getting smaller and smaller. There will be the maximum hit with silicon in the near future. How can we build more efficient CPU’s?
To answer this question we have to go back to the year 2008. In this year HP did research on project “Corona” (named after the sun’s corona), this was an optical CPU.
It would have 256 cores, each with 4 threads and has 10 trillion floating points operations a second. They claimed that just 5 of this CPU’s could have the speed of a super computer!
And they would only need a fraction of the power, compared to a normal CPU. This CPU was simulated but never made. Because one important piece was missing: the on-chip laser.
It could not made small enough.
Optalysys has a FHE encryption optical CPU which can work on encrypted data without the need to decrypt it. So the data stays safe. They also did use the CPU’s for fast DNA sequencing.
Lightmatter has an optical CPU for AI machine learning called “Envise”. In “Passage” there are optical interconnected CPU’s. They are with the functionality of a rack on a single chip. The optical CPU can have multiple cores and multiple light wavelengths combined. This makes a huge improvement in performance. And there is “Idiom”: it interfaces with standard deep learning frameworks and model exchange formats, while providing the transformations and tools required by deep learning model authors and deployers.
As you can see there is much going on in this area. So the future of this technology will be bright.