According to the researchers, a machine with this technology is able to make calculations and save data in the same place. This saves energy and achieves up to 200 times faster speed compared to cutting-edge classic computers. Understand in the following lines what in-memmory computing is and how the IBM project works.
IBM Scientists Can Store 330TB of Data on Cassette Tape
IBM’s non-processor computer is based on the concept of in-memory computing , a type of computing that happens inside the memory chip. The method differs from conventional computers and smartphones, which rely on two independent physical elements to store and process information: the processor and memory. Machines that use in-memmory computing can dispense with the CPU in operation.
The novelty keeps data from being transported all the time between the processor and memory – which makes computing more efficient. On personal computers, the difference in efficiency may not be so great. However, according to IBM researcher Abu Sebastian, the alternative is likely to have a significant impact on artificial intelligence systems .
The mechanism works with specific memory chips that IBM calls computational memory. The components are made of an alloy composed of germanium, tellurium and antimony, materials that respond efficiently to electrical stimuli. According to the researchers, the physical behavior of the material is essential to endow the memory with computational capacity.
Memory chips are not meant to perform the typical calculations of processing units, something that forced IBM to look for alternatives to unify the functions. To do so, scientists have explored the physical attributes and state dynamics of memory devices. Upon receiving an electric current, the material undergoes changes that cause the memory to perform calculations based on the modifications of the structure.