| 规格 | NVIDIA RTX 6000 Ada Generation | NVIDIA T4 |
|---|
| 架构 | NVIDIA Ada Lovelace architecture | Turing (TU104-895) |
| 显存 | 48GB GDDR6 | 16GB GDDR6 with ECC (默认开启) |
| 显存ECC | Yes | — |
| 显存位宽 | 384-bit | 256bit |
| 显存带宽 | 960 GB/s | 320GB/s (早期一页版datasheet口径300GB/s) |
| CUDA核心 | 18,176 | 2,560 (40 SM) / Tensor Core 320 / RT Core 40 |
| Tensor Core | 568 | — |
| RT Core | 142 | — |
| FP32算力 | 91.1 TFLOPS | — |
| RT Core性能 | 210.6 TFLOPS | — |
| Tensor性能 | 1457.0 TFLOPS | — |
| 功耗 | Total board power: 300 W | 70W (Total board power, 默认) |
| 互联 | PCIe 4.0 x16 | PCIe Gen3 x16 (支持x8运行, Interconnect BW 32GB/s) |
| 散热 | Active | — |
| 形态 | dual slot, full height (4.4in H x 10.5in L) | 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 | — |
| 供电接口 | 1x PCIe CEM5 16-pin | — |
| 编解码引擎 | 3x encode, 3x decode (+AV1 encode and decode) | — |
| NVLink | No | — |
| 算力 | — | FP32 8.1 TFLOPS / FP16·FP32混合精度 65 TFLOPS / INT8 130 TOPS / INT4 260 TOPS |
| 发布年 | — | 2018 (Product Brief v01 2018-10-24初版, Dec18版datasheet © 2018 NVIDIA) |
| 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 |
| SR-IOV | — | Supported; 16 VF（支持） |
| 工艺 | — | 12nm FFN, 136亿晶体管, 545mm2 |
| Compute APIs | — | CUDA, NVIDIA TensorRT, ONNX |