Slashdot Mirror
A Plasmonic Revolution for Computer Chips?
Roland Piquepaille writes "Today, we're using basically two ways to move data in our computers: transistors carry small amounts of data and are extremely small, while fiber optic cables can carry huge amounts of data, but are much bigger in size. Now, imagine a single technology combining the advantages of photonics and electronics. This Stanford University report says a new technology can do it: plasmonics. (For more about plasmons, read this Wikipedia article.) Theoretically, it is possible to design plasmonic components with the same materials used today by chipmakers, but with frequencies 100,000 times greater than the ones of current microprocessors. There is still a challenge to solve before getting plasmonic chips. Today, plasmons can only travel a few millimeters before dying, while today's chips are typically about a centimeter across. Read this overview for more details and references about plasmonics, and to discover why it's one possible future for chips' circuitry."
Lets keep it simple, put all of the Roland Piquepaille conspiracy posts here. :)
Editors: GIVE HIM HIS OWN DAMN SECTION SO CAN HIDE HIS POSTS
Your hair look like poop, Bob! - Wanker.
Let us think of a computer processing unit as a juggler, and bytes as mangoes. Older CPUs would juggle one mango at a time, and frequently require modifications to the stage to boot. Around the 1980s, they could juggle two mangoes. Then four around 1990, and today as many as eight at a time! Now you would be expected to be quite impressed with each leap, notwithstanding the fact that you really wanted a juggler that could handle melons, grapefruit, or watermelon slices instead of (or in addition to) mangoes. In addition, the fact that you are juggling in a zoo where a primate is free to grab your fruit and substitute twigs (or worse!) mid-juggle owing to something called "stack smashing" in computer terminology is not supposed to discourage you.
There is a movement towards something called mutable paragraphs, where as in English "words" (groups of bytes) can be of different lengths depending on need. This may mean the ability to exactly fill out a data page for better efficiency, or to allow the CPU to work with communication protocols in their element (if a common network packet is 68 bytes long, a word should be ½NP or 34 bytes in the I/O buffer.) It also means that you use no more CPU space than you absolutely need to for a computational step, decreasing wear and tear on your components.
I guess what I'm getting at is that science fiction has nothing on practical interative design for real world technological improvement. Sure, we might get to the same place we read about 50 years ago, but not all in one step.
Try not. Do or do not, there is no try.
-- Dr. Spock, stardate 2822-3.
Another use for this technology maybe rendering objects invisible to the observer. Using plasmons to stop light from scattering back to the observer's eyes. Unfortunately it can only be used to hide very small objects since the wave lengths of the light need to be near the size of the object that reflecting them.
If anyone wants anymore info on this check out this link.
I know next to nothing about this field but wonder if it would be possible to bridge the gap in distance using using metal or ceramic nano-wires embedded on-chip for the plasmons to travel across?
Wikipedia, because the tyranny of the persistant must be right.
Claiming "100,000 times the frequency" is a little misleading. You're not talking about the processor running at terahertz speeds - simply put, you can't make things small enough to do that. Plasmonic signals, photonic signals, electronic signals - they all travel on the order of light speed. There really wouldn't be much point raising the clock frequency beyond the characteristic length of the processing unit (Pentium 4 designers understand this now - they had to put "drive" stages into the pipeline just to allow signals to propagate, and that deep pipeline lead to a very low IPC).
This would be useful for things like memory and processor interconnects, because you could shove gigantic amounts of data. Hence the reason that the article stresses their use as high-traffic freeways. I'm not sure I see the point in an all-plasmonic chip (unless they've got power advantages) because of size concerns.
If the frequency is 100,000 times that of current CPUs, the chip could be more simplistic because it could process more instructions per second.
.3 or .4 centimeters, so 30 to 40% the size of today's current CPUs, operating at 100,000 times the frequency. I suppose you could even do the dual core thing and have 60-80% of the size of a normal CPU. OR....
I don't know how much fat you'd have to trim to get it down to a "few millimeters." So a few could be 3 or 4 milimeters which would be
While an all-plasmonic chip might be feasible someday, Brongersma expects that in the near term, plasmonic wires will act as high-traffic freeways on chips with otherwise conventional electronics.
That sounds more feasible than building an entire chip in this fashion. Just plug it in where you need it.
My money is still on carbon nanotubes, though.
Plasmons can be easily created in metallic nanotubes. Furthermore, it is possible to create them into an entangled state. This _in principle_ might be exploited on the quantum computation scale.
If you executed 2 = 2.5 then the statement "two plus two does not equal five, even for large values of 2" would be proven false in any following code.
Got time? Spend some of it coding or testing