| 规格 | AMD Instinct MI100 | AMD Instinct MI300X |
|---|
| 架构 | CDNA / gfx908 (CDNA1 世代开山, TSMC 7nm FinFET) | CDNA3 |
| 显存 | 32GB HBM2 (Full-Chip ECC, 4 stacks) | 192 GB |
| 显存带宽 | 1.2 TB/s (官方页口径; ROCm 架构文档 1.228 TB/s @1.2GHz) | 5.3 TB/s |
| 显存位宽 | 4096bit (4 stacks × 1024bit) | 8192-bit |
| 算力 | FP64 11.5 / FP32 23.1 / FP32 Matrix 46.1 / FP16 184.6 / BF16 92.3 / INT8·INT4 92.3 TOPS (TFLOPS, Peak Engine Clock 1502MHz) | — |
| 计算单元 | 120 CU / 7,680 流处理器 (8 Shader Engine × 15 CU, Matrix Core) | — |
| 功耗 | 300W Peak (TDP, 被动散热) | — |
| 互联 | Infinity Fabric 3链 (峰值92GB/s/链, 2-4卡hive) / PCIe 4.0 x16 (兼容 3.0 x16) | PCIe® 5.0 x16 |
| 发布年 | 2020 (官方 Launch Date 11/16/2020) | — |
| 形态 | PCIe Add-in Card (全高全长双槽 10.5"/267mm, 被动散热) | OAM Module |
| 显存类型 | — | HBM3 |
| FP8算力 | — | 2.61 PFLOPs |
| FP16算力 | — | 1.3 PFLOPs |
| FP32算力 | — | 163.4 TFLOPs |
| FP64算力 | — | 81.7 TFLOPs |
| INT8算力 | — | 2.6 POPs |
| 功耗TBP | — | 750W Peak |
| Infinity Fabric Links | — | 8 |
| Infinity Fabric带宽 | — | 128 GB/s |
| 流处理器 | — | 19,456 |
| 末级缓存LLC | — | 256 MB |
| 制程 | — | TSMC 5nm | 6nm FinFET |