02 Mar 21
@sonhouse saidWhile this is really impressive, don't hold your breath to see see it on the market… As I understand the paper the setup is very special to find out the switching time for the single device.
https://phys.org/news/2021-01-optical-sub-picosecond.html
So say it is 1/10th of a picosecond, that would be 10 TRILLION operations per second.
So challenges are:
* reproduce the result
* find out how to make an actual working device from it.
* characterize that one.
* find out how to make a complex device without losing the advantage
* characterize that one
* find a way to mass produce devices.
* show that to investors
* go for mass production.
In my opinion we suffer in science from having to beat our drum, and to project far into a future to find the "breakthrough".
The research is great in finding an ultrafast switch and being able to characterize it. An application in computing is very far away.
02 Mar 21
@ponderable saidIf it's anywhere comparable to cancer drug R & D it'll be at least 10 years.
While this is really impressive, don't hold your breath to see see it on the market… As I understand the paper the setup is very special to find out the switching time for the single device.
So challenges are:
* reproduce the result
* find out how to make an actual working device from it.
* characterize that one.
* find out how to make a complex device without losin ...[text shortened]... an ultrafast switch and being able to characterize it. An application in computing is very far away.
03 Mar 21
3 edits
@wildgrass
When the engineering is done and they have a viable product, it sounds to me like it would make for a supercomputer in the size of a sound bar.
Another path to small supercomputers: "Wafer scale processors"
1.2 TRILLION transistors on one wafer, a bunch of parallel processors all on one 12 inch wafer.
This technique has been tried for 40 years but just recently succeeded because of the problem of defects spoiling the mix.
https://cerebras.net/wafer-scale-processors-the-time-has-come/