Slashdot Mirror
Light-Based Memory Chip Is First To Permanently Store Data
sciencehabit writes: Scientists have developed the first ever memory chip that’s entirely light-based and can store data permanently. Sciencemag reports: "Today's electronic computer chips work at blazing speeds. But an alternate version that stores, manipulates, and moves data with photons of light instead of electrons would make today's chips look like proverbial horses and buggies. Now, one team of researchers reports that it has created the first permanent optical memory on a chip, a critical step in that direction. If a more advanced photonic memory can be integrated with photonic logic and interconnections, the resulting chips have the potential to run at 50 to 100 times the speed of today's computer processors."
Light doesn't have to worry about magnetic fields. An electrical current creates a magnetic field. While that field is being created, it takes power from the current flow. This slows does transitions. It's called inductance.
There is also the opposite of inductance, capacitance. You need to charge the gate of a fet before it will turn on. That takes time and slows things down too.
No the real question is it made of crystals. Every future computer is made is crystals
While you're deliberately being a moron, the answer actually happens to be yes, though it kind of depends.
The article just says they successfully fabricated a optical storage cell (3 bits per cell) using standard chipmaking process that's based off a phase change material, like the stuff used for rewritable DVDs, as the storage mechanism. The PCM material is partially crystalline and partially amorphous (non-crystalline) depending on how much energy is dumped into when it's being written to; that is how they get the 8 separate levels that allows it to store 3 bits per cell. So, how crystalline it is depends on what you have in memory at the time.
No. And this matters - when your dealing with 3GHz+ clocks, it actually becomes a problem getting a signal from one side of the chip to the other and back again within a single clock cycle.
Could this not be done the same way CRTs scan a grid of pixels, just on a micro scale with higher resolution?
This reminds me of an early computer memory, the Williams tube, that enjoyed a brief period of popularity in some first generation machines. It worked by storing bits as charged spots on the phosphor face plate of an oscilloscope tube. Although access was random and fast (12 microsecond read/write cycle as implemented by the IBM 701), its refresh requirements effectively halved its performance, and it was notoriously unreliable. Positioning the electron beam was by electrostatic deflection, requiring accurate sub-microsecond switching of high voltages. IBM's implementation used precision counter-wound resistors to achieve the required control, the counter-winding preventing the resistors from also behaving like inductors. Unfortunately, the counter-winding also led to occasional electrical arcing inside the resistors, mispositioning the beam and causing the "Navajo Blanket" effect where the resulting data corruption had a visual appearance like its namesake woven blanket. Error-free operation seldom exceeded a handful of hours, and the Williams tube was quickly supplanted by magnetic core memory.
Absolutely. Another huge problem is skew, where dissimilar wire lengths result in signals (for example, the bits making up a word) arriving at their destination at different times. This is not a problem exclusive to integrated circuits: Seymour Cray addressed this problem in the CDC 6600 (circa 1964, discrete Si transistors) by using wires of identical lengths for interconnections. If you look for a photo of the CDC 6600 back plane, you'll readily see what I mean.