NVIDIA® DGX-2™

erek

[H]F Junkie
Joined
Dec 19, 2005
Messages
10,894
The World’s Most Powerful AI System for the Most Complex AI Challenges
Tackle the world’s most difficult AI problems with the world’s most powerful AI system. The NVIDIA® DGX-2™ delivers 2 petaFLOPS of deep learning performance in a single system—combining the power of 16 fully interconnected NVIDIA® Tesla® V100 GPUs with the outstanding ease of use of the NVIDIA DGX platform and services.


  • 16 NVIDIA Tesla V100 “Volta” GPUs
  • 125 TFLOPS double-precision, 250 TFLOPS single-precision, 2 TensorPFLOPS of performance
  • 81,920 NVIDIA CUDA cores
  • Total of 512GB of high-bandwidth GPU memory, interconnected with NVSwitch tecnology
  • 12 NVSwitches with up to 2.4TB/sec of bisectional bandwidth interconnecting the Tesla GPUs
  • Two Intel Xeon Platinum Scalable CPUs
  • 1.5TB DDR4 2666Mhz System Memory
  • Eight, Dual Port 100Gbps EDR InfiniBand Adapters
  • 30TB of NVMe storage
  • 10U Rackmount Form Factor
  • Redundant, Hot-Swap power supplies
  • Power Consumption: 10kW at full load
  • Ubuntu Server Linux operating system

EJ2LnKsXYAEXTAI?format=jpg&name=900x900.jpg




@NVIDIADC
DGX-2 is here on display at #SC19 Booth 519.
 
"125 TFLOPS double-precision, 250 TFLOPS single-precision, 2 TensorPFLOPS of performance"

this hypothetically and or theoretically is capable of far surpassing the threshold of performance required to achieve real-time photo-realistic 3D rendering according to coretek
 
Back
Top