Supercomputer software firm to support Nvidia GPUs

Platform Computing has made its supercomputing software products GPU-aware, allowing workloads to be run and analysed on both CPU and GPU hardware

Platform Computing will support graphical processing unit computing hardware from Nvidia across its major software products, the company announced on Tuesday.

Platform Computing, which makes workload management and automation software for high-performance computing and cloud hardware stacks, had previously only supported CPUs for hardware management in grids and clusters.

"We do see a trend in interest in GPU [graphical processing unit] utilisation in the HPC market," Ken Hertzler, vice president of project management at Platform Computing, told ZDNet UK on Monday. "We have added the ability to intelligently schedule HPC workload on resources that have GPU cards only. For example, if a cluster has some servers that have dedicated GPU cards installed in them and others do not have GPU cards, you want to have the ability to schedule [GPU-specific] work to just the ones with GPU cards," he said.

Support is currently limited to the Nvidia GPUs, including Tesla, because of their use of the Cuda software instruction set architecture. Platform Computing uses Nvidia's GPUs because they are Cuda-capable, and no other instruction set has the sufficient sophistication for workload deployment, Platform Computing said.

From Tuesday, GPU support will be incorporated into Platform Computing's core product lines for cluster computing — Platform HPC — and grid computing — Platform LSF — and Platform Symphony.

Support will be free for Platform HPC and Platform LSF users and will carry an additional add-on charge for Platform Symphony users, Platform Computing said on Tuesday. Exact pricing was not available at the time of writing.

Customers will be able to use Platform Computing's products to deploy specific workloads, via Cuda, a software and hardware architecture for Nvidia's graphics cards, to Tesla GPU nodes.

"What we have today is basically only CPU workload schedulers. What we're saying is we can now distribute an application to a resource that has GPU resources on it," Tripp Purvis, a vice president of business management at Platform Computing, told ZDNet UK on Monday.

GPUs can be used in sectors like financial services, said Purvis. "We see algorithms used in the financial service industry, like Monte Carlo algorithms enabled to run on a GPU core," he added.

A Monte Carlo algorithm uses a combination of random numbers and known boundaries to compute rough numerical solutions to problems which are too complex to solve normally.

Platform Computing will expand support to other GPUs when units capable of handling processes such as the Monte Carlo algorithm come to market, Hertzler said. "Whether Intel or AMD come out with something similar in the future... I think we'll have to wait for their announcements," Purvis said.

The world's fastest supercomputer — the Chinese Tianhe-1A — uses 7,168 Tesla GPUs as part of its HPC architecture. HP released a three-GPU blade server in October.

GPUs are also becoming more popular with financial companies, according to financial technology consulting company CS Technology.

"We are hearing the buzz [for GPUs] in the industry at the moment," James Dow, CS Technology's chief technology officer, told ZDNet UK on Tuesday.

Platform Computing competes in the HPC management software market with companies such as IBM and Cray. It has strategic relationships with Cray, Dell, HP, IBM, Intel, Microsoft, Red Hat and SAS.


You have been successfully signed up. To sign up for more newsletters or to manage your account, visit the Newsletter Subscription Center.
See All
See All