It may be possible to develop faster, smaller and greener computers, capable of processing information at a speed up to a thousand times faster than the existing speed today, by converting the silicon material on which they are based with innovative materials capable of quickly switching between different electronic states.
[Translation by Dr. Moshe Nachmani]
It may be possible to develop faster, smaller and greener computers, capable of processing information at a speed up to a thousand times faster than the existing speed today, by converting the silicon material on which they are based with innovative materials capable of quickly switching between different electronic states.
It is possible to overcome the current size and speed limitations of computer processors and memory volume by converting the silicon material into innovative materials called 'phase-change materials' (PCMs, Wikipedia), materials capable of reversibly switching between two structural instances with different electronic states - the first crystalline and electrically conductive, and the second glassy and electrically insulating - within a billionth of a second.
Running models and conducting tests in devices based on such materials showed that logic processing operations can be performed in non-volatile memory cells using defined combinations of extremely short voltage pulses, which are not possible for silicon-based devices.
In these innovative devices, the logical operations and memory are in a common location, and not in separate locations, as is the case in silicon-based computers. These materials could, in the end, enable processing speeds that are 1000-500 times higher than the speed rates that exist today in portable computers, and this even while using less energy. The research findings were published in the scientific journal Proceedings of the National Academy of Sciences.
The new processors, developed by researchers from the University of Cambridge and the University of Singapore, make use of a type of show-changing material based on chalcogenous glass that can be melted and recrystallized in half a nanosecond by transferring an appropriate voltage. The calculations performed by most computers, mobile phones and tablet computers are made by silicon-based logic devices. The solid-state memory used to store the results of these calculations is also based on silicon. "However, as the demand for faster and faster computers increases, we are rapidly approaching the threshold of silicon's capabilities," said Professor Stephen Elliott from the Department of Chemistry at the University of Cambridge, who led the research.
The main method for increasing the power of computers has so far been to increase the number of logical devices included in them by increasingly miniaturizing their size, but physical limitations of the architectures of these devices mean that we are rapidly approaching the threshold of possible size. Today, the smallest silicon-based logic and memory devices are about 20 nanometers in size, and are arranged in layers. As the devices become more and more tiny, in order to increase their number on the chip, the gaps between the different layers will eventually be so small that the electrons may "leak" from the device (a phenomenon known as 'tunneling') and cause information loss. Devices based on secondary-instance materials can overcome this size threshold since they work properly when they are 2 nanometers in size.
Another alternative for increasing the processing speed without increasing the number of logical devices is to increase the number of calculations performed in each of these devices, which is not possible with the help of silicon, but it is possible with the help of second-instance materials.
First developed in the 60s, secondary-instance materials were originally used in optical memory devices, such as recordable DVD discs. Today, they are beginning to be used in electronic memory applications and to replace silicon-based flash memories in some smartphones. Despite the advantages inherent in secondary-instance materials, they also have several disadvantages: currently, they are still unable to perform calculations at speeds approaching silicon, and they exhibit instability in the initial amorphous state. At the same time, the researchers found that if the process is carried out in reverse order - starting with the crystalline state and then moving to the molten state, then these materials are much more stable and fast.
The switching speed built into these materials, i.e. the speed of formation, is at the rate of 10 nanoseconds, a rate suitable for the replacement of flash memory. By increasing the speed even more, to rates of one nanosecond (as demonstrated by the researchers back in 2012), these materials will one day be able to replace the DRAM type memory ('Dynamic Random Access Memory').
In silicon-based systems, information moves from one place to another, which leads to an increase in energy and time. "Ideally, we would like the information to be created and stored in the same location," said one of the researchers. "Silicon is a temporary phase: the information is created, moves to another location and is stored in a different location. However, if we use logical devices based on secondary materials, the information instance remains in the location where it was created."
4 תגובות
Please note: for twenty years there has been talk of replacing the existing MOS array with self-assembled monolayers with "islands" of longer molecules or spaces on the surface.
Another note: the 20 nanometer technology, produced - in part by Intel, rotates the old MOS arrangement by 90 degrees.
Such computers and even the smaller ones must consider quantum mechanics, and indeed in the seventies of the last century the great Isaac Asimov used the concept of "quantum computing" to create a "positron brain" for his robots.
Jesus
Right. I think the illustration implies the intention. In the conventional model of the computer, called the von Neumann model, there is a separate processor and a separate memory. Even in parallel computers that exist today, this structure is still preserved - each processor has a large amount of memory.
The brain is built as described in the article - a collection of small units where each unit is both a processor and a memory. The importance is twofold: such a computer will be able to solve the same types of problems that the brain is good at. We are not good at calculations but great at identification. The second point is more theoretical: there are problems that the normal model of the computer is not able to solve (it doesn't matter how fast or big the computer is) and the new model is able to.
By and large, he says that the ram will also be the cpu. And both will be smaller
Sounds amazing, even though I didn't really understand everything at first reading...