Could this be the start of the next revolution in computers?
An intergrated silicon optical transmitter based computer along with fast SSD
http://phys.org/news/2012-08-silicon-optical-transmitter-large-volumes.html
I'm envisioning a desktop version with a fiberoptic bus between ram CPU GPU south bridge north bridge etc etc
Silicon dreaming or cold hard reality?
>>52128612
bump
>>52128612
>I'm envisioning
No you're not. All you're thinking of is, "What if things were magically fast?" You don't know enough about anything to actually envision it.
>>52128612
PCI-E 4.0 uses an optical bus so you're not that far off. Optical interconnects are being used for mass storage, and a few companies are looking to explore them to reduce latency and increase bandwidth with MCMs.
>>52129156
your imagination operates under some pretty strict restraints.
>>52129348
It's obvious you don't know anything about how a computer works.
>>52128612
>Next the revolution in computing?
>>52128612
There are 2 relatively interesting notions to using light for basic logic units.
>1. One of the major problems with continuing to use emf transistors is the clock synchronization issue for sufficiently large ICs. Which is why we're in this quagmire of die-shrink hell. With light, we could conceivably create much larger ICs and only die-shrink when form factor became an issue.
>2. Computing would no longer be bound to a 1 wire 1 bit. With sufficiently fast switching we could transfer large amounts of data from element to element over a single connection.
>it will be possible to create optical transmitters small enough to be embedded into CPU modules capable of transmitting large volumes of data at rates of several terabits per second.
Yeah I'm not sure this is going to be on your desk, OP. But it sounds like it could be incredible for things like Watson.
>>52128612
The article clearly says this is for interconnect networks for multiple cpu chip.
>>52129521
Yes it's interesting especially as they appear to use existing fabrication techniques
>>52129935
Yes it does but it isn't a stretch to see this being adapted to other areas of system design
The next step is probably quantum computing - optics will get us only so far.
>>52131180
Another path which admittedly is likely a decade or two away 'could' be processors that use photons rather than electrons
They've been working on this for decades and, despite recent media hype, there's been no significant breakthroughs. I'm not a physicist so I'm not going to pretend to know exactly what the issues are, but I'm not holding my breath for optical processors any time soon.
>>52128612
The basic problem with these prototype optical gates chips is that the optical part is merely a bus for the bit to transfer, they then are detected by photodetectors (they take time to get activated too, decreasing the overall time advantage) which translates them to electrons again.
Optical chips are very very hard to make and offer a very meager performance gain over their pure silicon counterparts.
>>52132051
>I have literally no idea what I'm talking about
Hahaha. You autistic contrarians are so easy to spot.
http://spectrum.ieee.org/tech-talk/semiconductors/optoelectronics/processor-with-photonic-interconnects-built
http://spectrum.ieee.org/computing/hardware/can-hpes-the-machine-deliver
https://www.youtube.com/watch?v=PQXl-atcBsg
>>52131059
It's when the design of an interconnect network on a chip is vastly different than that of a motherboard's network.
> I'm envisioning a desktop version with a fiberoptic bus between ram CPU GPU south bridge north bridge etc etc
DEC thought about this in the '90s.
>>52128612
distance between components is too small for fiberoptic to make a difference
unless we come up with logic gates and components that actually works with photons , what you just posted is just a dreamy future
>>52133485
it would be very nice to reduce power consumption. I think the busses in a chip total would at least eat up 2-10W of power.
>>52133533
i dunno about that, they will waste alot more on [ electric <--> optical <--> electric ] conversions
>>52133626
Thats cute.
Go figure out how a transistor works.
>>52128612
I think OP fucked up when posting the link, he should have linked to this article:
http://phys.org/news/2015-12-demo-processor-ultrafast.html , video: https://youtu.be/JAe_xQyFI4k
this looks really awesome. note that they are using the RISC-V ISA! (I wonder how they got money to build their CPU, though... and who, and how, will pay for it. hopefully not the US military/NSA...)
also, from the comments: http://www.kurzweilai.net/skyscraper-style-carbon-nanotube-chip-design-boosts-electronic-performance-by-factor-of-a-thousand
there seems to be a lot going on in computing these days...
>>52134747
btw, I'll emphasize: WATCH THE VIDEO, they are connecting a CPU and a memory module through an optical link...
>I wonder how they got money to build their CPU, though...
nevermind, >>52132124 has the response:
>http://spectrum.ieee.org/tech-talk/semiconductors/optoelectronics/processor-with-photonic-interconnects-built
>Engineers at MIT, the University of California, Berkeley, and the University of Colorado, Boulder founded a start-up to commercialize what they call their “zero-change” approach to chip-making.
>... the federal Defense Advanced Projects Research Agency, which funded the work...
so, there you go. of course, the US military is funding this.
paper (someone should upload this to libgen...):
http://www.nature.com/nature/journal/v528/n7583/full/nature16454.html
and, on related news...
http://phys.org/news/2015-12-team-two-phonon-quantum-world.html
