| 规格 | NVIDIA RTX A5000 | NVIDIA T4 |
|---|
| 架构 | Ampere (GA102, GA102-850-A1) | Turing (TU104-895) |
| 显存 | 24GB GDDR6 with ECC | 16GB GDDR6 with ECC (默认开启) |
| 显存带宽 | 768GB/s | 320GB/s (早期一页版datasheet口径300GB/s) |
| 显存位宽 | 384bit | 256bit |
| 算力 | FP32 27.8 TFLOPS / RT Core 54.2 TFLOPS / Tensor 222.2 TFLOPS (含稀疏加速) | FP32 8.1 TFLOPS / FP16·FP32混合精度 65 TFLOPS / INT8 130 TOPS / INT4 260 TOPS |
| CUDA核心 | 8192 / Tensor Core 256 (三代) / RT Core 64 (二代) | 2,560 (40 SM) / Tensor Core 320 / RT Core 40 (白皮书口径) |
| 功耗 | 230W (主动散热, 1x 8-pin PCIe供电) | 70W (Total board power, 默认) |
| 互联 | NVLink 112.5GB/s 双向 (低矮桥接器2槽/3槽, 双卡拼48GB, 桥接器另购) / PCIe Gen4 x16 | PCIe Gen3 x16 (支持x8运行, Interconnect BW 32GB/s) |
| 发布年 | 2021 (2021-04-12, 与RTX A4000同日发布) | 2018 (Product Brief v01 2018-10-24初版, Dec18版datasheet © 2018 NVIDIA) |
| 形态 | Dual-slot 双槽全高, 4.4"H x 10.5"L, Active主动散热 | 单槽半高半长 Low-Profile 6.6英寸(长169.53mm), 被动散热, 板重301g |
| 显示输出 | 4x DisplayPort 1.4a | — |
| 编解码 | 1 Encode / 2 Decode (+AV1解码) | — |
| VR | Yes (VR Ready, vGPU 软件支持 vPC/vApps/RTX vWS/vCS) | — |
| 时钟 | — | Base 585MHz / Boost 1590MHz (Base为70W功耗墙下调校) |
| 媒体引擎 | — | 1080p30解码 H.264 32路 / HEVC 44路 / VP9 32路 (白皮书Table 2, 官方页称38路full-HD) |
| SR-IOV | — | Supported; 16 VF（支持） |
| 工艺 | — | 12nm FFN, 136亿晶体管, 545mm2 (白皮书TU104口径) |
| Compute APIs | — | CUDA, NVIDIA TensorRT, ONNX |