- 主な特長
- AI およびグラフィックス アプリケーション向けの NVIDIA Tensor および Ray Tracing Core
- 6144 CUDA コアと 32 GB GDDR6 メモリ
- PCIe Gen3 と NVIDIA NVLink の高速インターコネクト
- NVIDIA、CUDA、OpenCL ソフトウェアのサポート
- MOTS+ 極限環境保護 (オプション)
- ANSI/VITA 65.0 準拠
- GPU
- 2x embedded NVIDIA Quadro TU104 GPU with the Turing architecture
- 6144 total processing cores (3072 lanes per GPU), 384 Tensor cores, 48
Ray Tracing cores
- 21.7 peak theoretical single-precision (FP32) TFLOPS
- 32 total PCIe 3.0 lanes (16 lanes per processor)
- 32 GB total GDDR6 Memory (16 GB per processor)
- 512-bit memory interface (256-bit per processor)
- 896 GB/s memory bandwidth (448 GB/s per processor)
- 4x display port display outputs (2 per GPU) to front panel (air-cooled
only) and OpenVPX backplane
- 2x analog VGA display outputs (1 per processor) to front panel (air-cooled
only) and OpenVPX backplane
- Dual 52-Lane Configurable PCIe Switches
- Configurable switches enable multiple system-level configurations:
- Non-transparent bridging and enumeration
- x16 PCIe 3.0 connections to GPU (32 lanes total)
- x32 PCIe 3.0 total connections to backplane
- x16 PCIe 3.0 OpenVPX P2 expansion plane
- x16 PCIe 3.0 OpenVPX P5 expansion plane
|
- IPMI (System Management)
- On-board IPMI controller
- Voltage and temperature monitor
- Geographical address monitor
- Power/reset control
- Onboard CPLD, FRU EEPROM interfaces
- OpenVPX Multi-Plane Architecture
- System management via IPMB-A and IPMB-B link on P0 management plane
- Dual full x16 or dual x8 PCIe on P2 and P5 expansion plane
- 4x display port display outputs on P6 mezzanine I/O plane
- 2x analog VGA outputs on P3 mezzanine I/O plane
- Mechanical and Open Systems Architecture
- OpenVPX (ANSI/VITA 65.0) and VPX-REDI
- VITA 46.0, 46.3, 46.4, 46.6, 46.11, and VITA 48.1, 48.2, 48.4, 48.7 (REDI)
- ANSI/VITA 65.0 module profile MOD6-PAY-4F1Q2U2T-12.2.1-n
- 1.0 インチ ピッチ, シングルスロット
- Cooling options: air-cooled, air flow-by, conduction-cooled, or liquid
flow-through
- オプション MOTS+ 極限環境保護 パッケージ
- 消費電力
- typ. 170W アクセラレータ毎, 電力は構成変更可能です
|