IBM Fiber Optic Breakthrough
#1
Posted 27 March 2007 - 09:26 PM
They have figured out how to implement fiber optic connectors in place of copper in chips and mobo's.
They anticipate the connectors will be 8 times faster than current copper based setups with speeds hitting "160 billion bits per second".
and these links on the subject:
http://www.verivox.d.......=47529&pm=1
and:
http://domino.research.ibm.com/comm/pr.nsf...ransceiver.html
I'm sure it'll cost a mint when it hits the streets ... but imagine the possibilities
#2
Posted 27 March 2007 - 10:41 PM
#3
Posted 27 March 2007 - 11:22 PM
"Supercomputer" will lose it's status... and nobody with a really decent pc can say they don't know if light is fast!
#4
Posted 28 March 2007 - 03:49 AM
#5
Posted 30 March 2007 - 03:42 AM
another step closer towards building my holodeck! nice
#6
Posted 30 March 2007 - 03:55 AM
Edited by stormrosson, 30 March 2007 - 03:57 AM.
#7
Posted 30 March 2007 - 04:52 AM
#9
Posted 30 March 2007 - 10:49 PM
stormrosson, on Mar 30 2007, 05:55 AM, said:
No, the speed gain will be from the amount of data send through one line. As opposed to a copper wire, in one optic fibre line, there can be almost uncountable (okay, not entirely uncountable) amounts of data streams flow at the same time (different angles of entry, different wavelength are already two methods to send in the case of the entry angle 8 different streams)
Thus, a higher dataflow despite the fact that the speed of which the data travels is pretty close to the same (light actually is, in theory, slightly faster, see Einsteins Relativity Theory)
----
For those who don't completely understand the physics of fibre optics, here's a short and as simple as I can explanation.
A fibreoptic wire looks a bit like concentric circles when you look into it (ok, it's way thinner offcourse!)
The optical data can enter this wire at 8 points, namely the top, and then every 45 degrees around the circle another one.
A lightwave entering the fibreoptic cable, kinda follows the following pattern, every layer breaks the ray a bit towards the center.
This makes the light follow a sort of sinus path, until it's bounced at the edge of a wire. (I won't go into detail about how this all works.
---
Then, at every entry point, there offcourse needs to be an exit point. So this must be very carefully calibrated, or else the wire is useless for the 8point entry system. I'm guessing, for inside a computer, they'll be able to calculate this to maximize the dataflow.
---
Next, there's the matter of wavelength. Maybe you know, maybe you don't, but lightwaves can be send through fringes, making them brake and create light and dark spots on a wall (please consult Wikipedia for details of that).
Since every wavelenght has a different spacing between the points on a wall that are lit, and which are dark, usually the fourth top from the center highest is used as the dataspot.
---
So, with the knowledge of all wavelengths having different points and multiple waves can be send through one wire, you can hopefully now a bit imagine that on all 8 points you can at least send 200 (yes, 200) different datastreams at the same time. thus 200*8 makes 1600 bits per transfer. at the speed of light.
This is not entirely true, because sending only one bit is pretty useless, the data must carry an identifier, so the computer knows which data is belonging to what process, but you get the idea I hope.
If you have questions, you're free to ask offcourse!
#10
Posted 30 March 2007 - 11:48 PM
... so was calling it a reduced lag time a mis-nomer on me?
Rather I shold have beat down the skinny copper for not hauling enough juice?
#11
Posted 01 April 2007 - 04:20 AM
Edited by stormrosson, 01 April 2007 - 04:23 AM.
#12
Posted 01 April 2007 - 05:14 AM
0 user(s) are reading this topic
0 members, 0 guests, 0 anonymous users