Term of the Moment
AI chip
Redirected from: in-memory computation
Definition: compute-in-memory
A combination processor and RAM in the same circuit. Also called "processing in memory," "memory-centric computing" and "in-memory computation," compute-in-memory systems are designed to handle huge amounts of parallel arithmetic calculations for AI and other math-intensive applications. For decades, numerous designs for a chip that enable the contents of memory to be directly processed have been considered.
CPU <--bus--> Memory
The traditional computer architecture is a CPU connected to memory via a system bus. To be calculated, data in memory traverse this pathway back and forth, and those data transfers take a huge amount of processing time. As AI applications grow larger and more complex, the goal of the compute-in-memory design is to handle matrix multiplications (multiply-and-add) in parallel at the highest speed possible. See matrix multiplication.
Microcontrollers - A First Step
Since the early 1970s, microcontrollers (MCUs) have placed the CPU and memory on the same chip, but there is still a pathway between them. See microcontroller.
Different Approaches
CPU <--bus--> Memory
The traditional computer architecture is a CPU connected to memory via a system bus. To be calculated, data in memory traverse this pathway back and forth, and those data transfers take a huge amount of processing time. As AI applications grow larger and more complex, the goal of the compute-in-memory design is to handle matrix multiplications (multiply-and-add) in parallel at the highest speed possible. See matrix multiplication.
Microcontrollers - A First Step
Since the early 1970s, microcontrollers (MCUs) have placed the CPU and memory on the same chip, but there is still a pathway between them. See microcontroller.
The common computer architecture (top) is a system bus between CPU and memory (RAM). In compute-in-memory designs, the memory and processing circuits are either very close or integrated together.