Pilot GPU nodes on Biowulf cluster (Biowulf)

Graphics processing units (GPUs) are specialized microprocessors originally
designed for video and rendering. More recently, compute-intensive programs
in the life sciences have been ported to GPUs to explore potential performance
benefits of their massive compute power.

A pilot GPU cluster has been installed to investigate the performance of GPUs
for the types of computations that are performed here at NIH. The Biowulf GPU
nodes contain:

	* 2 x Nvidia M2050 (Tesla Fermi), each: 2.8 GB memory, 448 cores
	* 2 x Intel Xeon X5650 (2.67 GHz), each: 12 cores
	* 48 GB DDR3 memory
	* 450 GB disk (7200 rpm SATA)
	* Each node is connected to two networks: QDR Infiniband (32 Gb/s) and
		1 Gb/s ethernet.

The GPU nodes are integrated with the Biowulf cluster and can be accessed via
'qsub' as usual. We have installed and tested the CUDA versions of NAMD, Amber
and several bioinformatics programs on these nodes. Details about submitting
jobs to these nodes and the results of our benchmarks are available at

http://biowulf.nih.gov/gpu.html


Users are invited to investigate these GPU nodes for their jobs. Please let us
know of any benchmarks you run, and of any other GPU-enabled programs that 
you would like installed.

Biowulf/Helix staff.