Infrastructure

GCOE's infrastructure is composed of a GPU Cluster and GPUs and Embedded Supercomputing kits which we loan out to those who to pursue research and development using GPUs.

GPU Cluster

Currently we are running a GPU Cluster facility of 4 GPUs with the following configuration:

login node:

Intel(R) Core(TM)2 Duo CPU E7500 @ 2.93GHz with 2 cores 4 GB RAM
4 GB SWAP

CUDA toolkit and driver version 7.0
CentOS Linux release 7.1.1503 (Core) with Linux kernel 3.10.0-229.14.1.el7.x86_64 Slurm Control Daemon 15.08.1

GPU node 1:

Intel(R) Xeon(R) CPU E5-2620 0 @ 2.00GHz with 24 cores
32 GB RAM
30 GB SWAP

2 x Tesla K40c GPUs with 2880 CUDA cores, 12GB VRAM, CUDA Compute Capability 3.5

1 x Tesla K20c GPU with 2496 CUDA cores, 5GB VRAM, CUDA Compute Capability 3.5

GPU node 2:

Intel(R) Core(TM) i7-3770 CPU @ 3.40GHz with 8 cores

16 GB RAM
8 GB SWAP

1 x Titan X GeForce GTX GPU with 3072 CUDA cores, 12GB VRAM, CUDA Compute Capability 5.2

GPUs Loaned

We have also loaned out Tesla K40, Tesla C2070, GTX 650Ti to various researchers for use on their machines, in the IIT Bombay Campus

Embedded Supercomputing:

We have provided 1 Jetson TK1 kit to the IIT Bombay's Autonomous Car Team, 1 to the Electrical Department and 1 to the Centre for Studies in Resources Engineering for use in the IIT Bombay Campus

Expansion plans:

We have plans to further expand the infrastructure by adding two more GPU Compute Nodes and separating our infrastructure between those who wish to develop and optimize their own code and those who wish to run different software to get a taste of GPUs.

We are also exploring providing Cloud-based GPU instances to researchers who want to utilize GPUs.