AI term
Multi-Instance GPU
What is Multi-Instance GPU?
A GPU technology that divides one physical accelerator into multiple hardware-isolated instances, each with dedicated computing and memory resources.
In other languages
- 한국어멀티 인스턴스 GPU
- 하나의 물리 GPU를 여러 독립 실행 파티션으로 나누는 기술입니다.
- 日本語マルチインスタンスGPU
- 1基のGPUを複数の独立した計算領域に分割する技術です。
Related Terms
- GPU virtualizationSoftware that divides a physical GPU into virtual resources for separate workloads.
- Streaming multiprocessorA parallel processing unit within an NVIDIA GPU that executes groups of computing tasks.
- Heterogeneous GPU memory allocationAssigning memory across different types or configurations of graphics processors to support a computing workload.
- Unified memoryComputer architecture design where CPU and GPU share the same memory pool, allowing faster data exchange and higher capacity for AI workloads
- CUDA IPCNVIDIA CUDA mechanism that lets separate processes share GPU memory handles for interprocess communication
- GPU ManagementInfrastructure practice for scheduling and allocating GPU workloads across tasks, priorities, and hardware constraints
- Heterogeneous ComputingThe use of multiple types of processors or specialized cores within a single system to optimize performance and energy efficiency for diverse computational tasks.
The fastest, most reliable AI news, straight to your inbox