SoftLayer launched a HPC cloud-based solution based on servers using the GPU (Graphics Processing Unit) cards from Nvidia.

The company sees a need for an on-demand, consumptive billing model for High-Performance Computing (HPC) users needing data mining, numerical and seismic analysis which also targets heavy content producers such as advertising agencies and web design shops looking to develop interactive games, applications and 3D content.

“Using GPUs for running applications is an emerging trend in HPC today. Computing requirements continue to skyrocket but users remain as demanding as ever. They want capacity on-demand and prefer to pay only for what they use. It is incumbent on hosters to deliver this capacity in a cloud-like fashion if they want to break into this market,” said Philbert Shih, Managing Director, Structure Research Ltd. “Giving people the ability to scale GPU resources adds incredible flexibility in the way research, engineering and development projects are conducted. By putting such raw computational power into a user’s hands, in an easy pay as you go model, the focus is on research results not acquiring or maintaining the latest technology.” 

SoftLayer’s HPC hosting solution use dual-processor Intel E5-2600 (Sandy Bridge) based servers supporting one or two Nvidia Tesla M2090 GPUs.

The GPU offloads compute intensive portions of the application from the CPU to accelerate overall performance of the machine, dramatically boosting application performance by running in tandem with the CPU.