Researchers have developed a brand new sort of memory cell that may each retailer data and do high-speed, high-efficiency calculations.
The memory cell permits customers to run high-speed computations contained in the memory array, researchers reported Oct. 23 within the journal Nature Photonics. The quicker processing speeds and low power consumption could assist scale up knowledge facilities for artificial intelligence (AI) techniques.
“There’s a whole lot of energy and a whole lot of power being put into scaling up knowledge facilities or computing farms which have 1000’s of GPUs [graphics processing units] which can be working concurrently,” research co-author Nathan Youngblood, {an electrical} and laptop engineer on the College of Pittsburgh, informed Reside Science. “And the answer hasn’t essentially been to make issues extra environment friendly. It is simply been to purchase extra and extra GPUs and spend extra and extra energy. So if optics can tackle among the similar issues and do it extra effectively and quicker, that might hopefully end in diminished energy consumption and increased throughput machine studying techniques.”
The brand new cell makes use of magnetic fields to direct an incoming light sign both clockwise or counterclockwise by way of a ring-shaped resonator, a part that intensifies light of sure wavelengths, and into one of two output ports. Relying on the depth of light at every of the output ports, the memory cell can encode a quantity between zero and one, or between zero and minus one. Not like conventional memory cells, which solely encode values of zero or one in one bit of knowledge, the brand new cell can encode a number of non-integer values, permitting it to retailer as much as 3.5 bits per cell.
Associated: New ‘petabit-scale’ optical disc can store as much information as 15,000 DVDs
These counterclockwise and clockwise light alerts are akin to ” two runners on a observe which can be working in reverse instructions across the observe, and the wind is at all times within the face of one and to the again of the opposite. One can go quicker than the opposite,” Youngblood stated.. “You are evaluating the pace at which these two runners are working across the observe, and that permits you to principally code each constructive and adverse numbers.”
The numbers that end result from this race across the ring resonator could be used to both strengthen or weaken connections between nodes in synthetic neural networks, that are machine studying algorithms that course of knowledge in methods just like the human mind. That could assist the neural community determine objects in a picture, for instance, Youngblood stated.
Not like conventional computer systems, which make calculations in a central processing unit then ship outcomes to memory, the brand new memory cells carry out high-speed computations contained in the memory array itself. In-memory computing is especially helpful for purposes like synthetic intelligence that must course of a whole lot of knowledge in a short time, Youngblood stated.
The researchers additionally demonstrated the endurance of the magneto-optic cells. They ran greater than 2 billion write and erase cycles on the cells with out observing any degradation in efficiency, which is a 1,000-fold enchancment over previous photonic memory applied sciences, the researchers wrote.Typical flash drives are restricted to between 10,000 and 100,000 write and erase cycles, Youngblood stated.
Sooner or later, Youngblood and his colleagues hope to place a number of cells onto a pc chip and strive extra superior computations.
Ultimately, this know-how could assist mitigate the quantity of energy wanted to run synthetic intelligence techniques, Youngblood stated.