AI term
Processing-in-memory
What is Processing-in-memory?
A design approach that places computing units directly within memory to reduce data movement.
In other languages
- 한국어메모리 내 연산
- 데이터를 프로세서로 옮기는 대신 메모리 안에서 연산을 수행하는 방식입니다.
- 日本語メモリー内演算
- メモリー内に演算機能を組み込み、データ処理を行う方式です。
Related Terms
- Compute-in-memoryComputing architecture that performs data processing directly within the memory unit to reduce power consumption and latency
- Near-memoryA hardware design approach that minimizes the distance between the processor and memory to reduce data transfer delays and energy use.
- Von Neumann architectureComputer design model separating memory and processing units, requiring data transfer between the two during computation
- Unified Memory ArchitectureComputer design where processors share one memory pool instead of separate memory areas
- Shared MemoryA method of inter-process communication that allows multiple running programs to access the same region of RAM to exchange data efficiently without complex message passing.
- Unified memoryComputer architecture design where CPU and GPU share the same memory pool, allowing faster data exchange and higher capacity for AI workloads
- Zero-copyA computational optimization technique that allows data to be accessed or transferred without copying it between different memory locations, reducing latency and CPU overhead.
- Host-to-device transferMovement of data from a computer’s CPU memory to memory accessible by an accelerator.
- Sparse computingComputing methods that skip operations involving zero or inactive data to reduce processing work.
The fastest, most reliable AI news, straight to your inbox