Technology

The Durham College GPU Accelerate AI Systems Lab

Durham College is well-positioned to grow as a leader in helping companies to develop and adopt AI solutions. A GPU accelerated AI computing lab was established in 2017, including a large capacity GPU-based computer system, for teaching and research applications in machine and deep learning.  Students trained in this lab will bring advanced practice to projects with local businesses.  In addition to its on-campus resources, the AI Hub offers expertise and access to industry-leading, domain-specific AI platforms that address a variety of business use cases that are undertaken in industry projects.

GPU Accelerated Platform

The Durham College GPU Accelerated compute system is an IBM Power8 server with state-of-the-art NVIDIA Tesla GPUs.  The IBM Power8 system uses the NVLINK technology to provide high-bandwidth connection between the core compute tasks in the CPU, system memory, and the GPU compute tasks.  This system is used primarily for research, project work and teaching in the fields of machine learning and deep learning where the multiple-GPU cores can be used in parallel to achieve high throughput model development.

Technical Specifications

Model IBM Power System S822LC for HPC and AI
Number of nodes 1
CPUs Dual POWER8 10Core Processor
GPUs Two Nvidia Tesla P100 GPUs
System Memory 512GB
Storage Dual 480GB Read Intensive SATA 5xx SFF-4 SSD (mirror)
Operating System Ubuntu (Linux)
ML Software IBM Power AI (includes Caffe, Tensorflow, Theano, Torch, Chainer, DIGITS, OpenBLAS, Bazel, NCGL, NVIDIA CUDA)
System Software Python, Jupyter Notebook, Keras