| 规格 | NVIDIA Quadro RTX 8000 | NVIDIA T4 |
|---|
| 架构 | Turing (TU102, TSMC 12nm FFN, 18.6B 晶体管) | Turing (TU104-895) |
| 显存 | 48GB GDDR6 with ECC | 16GB GDDR6 with ECC (默认开启) |
| 显存带宽 | 672GB/s (GDDR6 14Gbps) | 320GB/s (早期一页版datasheet口径300GB/s) |
| 显存位宽 | 384bit | 256bit |
| 算力 | FP32 16.3 TFLOPS (Boost) / Tensor 130.5 TFLOPS / FP16 32.6 TFLOPS / INT8 206.1 TOPS / RTX-OPS 84T / FP64=FP32的1/32≈509GFLOPS | FP32 8.1 TFLOPS / FP16·FP32混合精度 65 TFLOPS / INT8 130 TOPS / INT4 260 TOPS |
| CUDA核心 | 4608 (72 SM) / Tensor Cores 576 (图灵二代) / RT Cores 72 (一代) | 2,560 (40 SM) / Tensor Core 320 / RT Core 40 (白皮书口径) |
| 功耗 | 总板功耗 295W / 图形功耗 260W (含 USB-C/VirtualLink 为 295W, PNY 口径) | 70W (Total board power, 默认) |
| 互联 | NVLink 100GB/s 双向 (TU102 双 x8 链路, 双卡拼 96GB, NVLink 售卖) / PCIe 3.0 x16 | PCIe Gen3 x16 (支持x8运行, Interconnect BW 32GB/s) |
| 发布年 | 2018 (2018-08-13 SIGGRAPH 温哥华发布[媒体多记08-14], Q4 2018 上市, 预计市场价 $10,000) | 2018 (Product Brief v01 2018-10-24初版, Dec18版datasheet © 2018 NVIDIA) |
| 形态 | Dual-slot 全高, 4.4"H x 10.5"L, Active 主动散热, 1x 8-pin 外接供电 | 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 1.4 + VirtualLink (USB Type-C) | — |
| 编解码 | 1 Encode / 1 Decode | — |
| VR | Yes (VR Ready, 业界首个 VirtualLink 接口) | — |
| API | DirectX 12.0 (FL 12_1) / OpenGL 4.6 / Vulkan 1.1 / CUDA / OpenCL / DirectCompute | — |
| 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 |