JANUARY 2017: TURNKEY DEEP LEARNING SOLUTION WITH SUPERMICRO SYS-1028GQ-TXR(T) + NVIDIA Tesla P100 SXM2 WITH NVLINK

The Supermicro SYS-1028GQ-TXR(T) 1U server supports four Tesla P100 SXM2 accelerators and can handle workloads with up to 85 TFLOPS FP16 of peak performance. It supports the dual Intel Xeon processor E5-2600 v4/v3 product families; 512GB DDR4-2400MHz in 16 DIMM; optional dual on-board 10GBase-T ports; and redundant Titanium-level digital power supplies. This result is a dense and compact server capable of generating massively parallel processing power and unrivalled GPU peering via 80GB/s NVLINK for Machine Learning applications.

NovaGlobal provides a turnkey Deep Learning Solution by integrating the server with Open Source and free software including:

• Ubuntu OS, GPU Driver, CUDA Toolkit 8.0 & CUDNN Library
• NVIDIA DIGITS, Caffe, Torch, Theano, TensorFlow & BIDMACH
• Docker engine and containers for Deep Learning Frameworks