SoylentNews is people
SoylentNews
https://www.fau.eu/2017/09/25/news/research/the-fastest-light-driven-current-source/
Controlling electronic current is essential to modern electronics, as data and signals are transferred by streams of electrons which are controlled at high speed. Demands on transmission speeds are also increasing as technology develops. Scientists from the Chair of Laser Physics and the Chair of Applied Physics at Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU) have succeeded in switching on a current with a desired direction in graphene using a single laser pulse within a femtosecond – a femtosecond corresponds to the millionth part of a billionth of a second. This is more than a thousand times faster compared to the most efficient transistors today.
[...] For their experiments, the scientists fired extremely short laser pulses with specially engineered waveforms onto graphene. When these light waves hit the graphene, the electrons inside were hurled in one direction, like a whiplash. 'Under intense optical fields, a current was generated within a fraction of an optical cycle – a half femtosecond. It was surprising that despite these enormous forces, quantum mechanics still plays a key role,' explains Dr. Takuya Higuchi from the Chair of Laser Physics, the first author of the publication.
Light-field-driven currents in graphene (DOI: 10.1038/nature23900) (DX)
(Score: 2) by ledow on Wednesday September 27 2017, @05:08PM (7 children)
The problem is not speed, but distance.
These things are constrained by the speed of light. The reason chips can't go faster at the moment is because - at the speed of light - it starts to take longer than a clock cycle for a signal to traverse the chip. Modern chips rely on a common clock across the chip.
And when you make the chips small enough to traverse in the time you need, the distance again comes into place, as the current leaks through things you don't want it to on that fine a scale, and generates a lot of heat in a tiny area.
The problem of THz computing isn't being able to trigger a fast light impulse. It's being able to do anything useful with that light impulse if you want it to pass through billions of "gates" etc. along the way and get an answer out the other end that's correct, consistent, and timed properly.
Though this is useful in all kinds of context, THz computing is deca-years ago. We've stagnated at about 3-4GHz in home computers for a reason. And even the fastest chips are only 8-9GHz.
(Score: 1, Interesting) by Anonymous Coward on Wednesday September 27 2017, @05:15PM (1 child)
Removing the need for a monotonic clock pulse could very well be one of the benefits of this technology. While clockless traditional CPU designs do exist, they haven't gotten transistor switching time down to the half femtosecond like is being described here. Combine this with RISC architectures that reduce the number of transistors needed to being with and we really start to see some real potential with this kind of technology.
(Score: 2) by ledow on Thursday September 28 2017, @07:38AM
But removing the need for a monotonic clock pulse could also be done WITHOUT this technology.
Even at traditional transistor speed, they can't manage to match clocked chips.
It may also mean a complete change of architecture and many even programming paradigm (i.e. everything suddenly becomes asynchronous).
But it doesn't need THz transistors to get there.
(Score: 2) by c0lo on Wednesday September 27 2017, @05:22PM (2 children)
You may want to recheck your calculations. 1THz => 1 femto-light-second = 3 m
It's not the travel space that causes the processing delay, it's the maximum switching speed of the transistor gates forming/maintaining/modifying the signal.
https://www.youtube.com/watch?v=aoFiw2jMy-0 https://soylentnews.org/~MichaelDavidCrawford
(Score: 1, Informative) by Anonymous Coward on Wednesday September 27 2017, @05:44PM (1 child)
"tera" is 1012. "femto" is 10−15. You want to be using "pico", 10−12, to keep your bases consistent.
A 1 THz wavelength is 1 pico light-second which is 0.003 m which is 3 mm. It's tiny.
(Score: 2) by c0lo on Wednesday September 27 2017, @06:10PM
Yeap, my mistake.
https://www.youtube.com/watch?v=aoFiw2jMy-0 https://soylentnews.org/~MichaelDavidCrawford
(Score: 2) by takyon on Wednesday September 27 2017, @05:26PM
Just increase the core count to hundreds or thousands.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 2) by Grishnakh on Wednesday September 27 2017, @10:29PM
The reason chips can't go faster at the moment is because - at the speed of light - it starts to take longer than a clock cycle for a signal to traverse the chip. Modern chips rely on a common clock across the chip.
Right. The way to deal with this is to put the CPU in a small subspace field so that the speed of light is faster.