AI半導体・チップ

01GPUGraphics Processing Unit 02TPUTensor Processing Unit 03NPUNeural Processing Unit 04FPGAField-Programmable Gate Array 05ASICApplication-Specific Integrated Circuit 06CPU（AI文脈）Central Processing Unit for AI 07NVIDIANVIDIA Corporation 08CUDACompute Unified Device Architecture 09cuDNNCUDA Deep Neural Network library 10Tensor CoreTensor Core 11AMD InstinctAMD Instinct 12Intel GaudiIntel Gaudi AI Accelerator 13Google TPUGoogle Tensor Processing Unit 14Apple Neural EngineApple Neural Engine 15Qualcomm AI EngineQualcomm AI Engine 16AI半導体AI Semiconductor 17AIチップAI Chip 18AIアクセラレータAI Accelerator 19エッジAIチップEdge AI Chip 20ニューロモーフィックチップNeuromorphic Chip 21光コンピューティングOptical Computing 22量子コンピュータ（AI）Quantum Computing for AI 23メモリ帯域幅Memory Bandwidth 24HBM（High Bandwidth Memory）High Bandwidth Memory 25VRAMVideo Random Access Memory 26PCIePeripheral Component Interconnect Express 27NVLinkNVLink 28InfiniBandInfiniBand 29半導体微細化Semiconductor Scaling 30プロセスノードProcess Node 31チップレットChiplet 32SoCSystem on Chip 33ウェハーWafer 34ファウンドリFoundry 35TSMCTaiwan Semiconductor Manufacturing Company 36Samsung FoundrySamsung Foundry 37演算精度（FP32/FP16/INT8）Compute Precision (FP32/FP16/INT8) 38混合精度演算Mixed Precision Training 39BFloat16Brain Floating Point 16 40INT4量子化INT4 Quantization 41スパースコンピューティングSparse Computing 42並列計算Parallel Computing 43SIMDSingle Instruction Multiple Data 44GPU仮想化GPU Virtualization 45マルチGPU学習Multi-GPU Training 46データ並列Data Parallelism 47モデル並列Model Parallelism 48パイプライン並列Pipeline Parallelism 49テンソル並列Tensor Parallelism 50推論チップInference Chip 51学習チップTraining Chip 52消費電力効率（TOPS/W）Power Efficiency (TOPS/W) 53ベンチマーク（MLPerf）MLPerf Benchmark