IBM SoftLayer adds Nvidia Tesla K80 as an option for GPU acceleration in the cloud

Big Blue adds more oomph to its cloud servers for HPC applications

IBM's SoftLayer cloud division has extended its bare metal server offerings for high-performance computing (HPC) by adding Nvidia's most powerful GPU accelerator card as an option for customers looking to build supercomputing applications.

SoftLayer has enabled customers to choose bare metal server servers equipped with GPU acceleration for several years now, allowing them to build HPC clusters to meet specific application requirements, and pay for them only as long as they need to.

The firm has now added the Nvidia Tesla K80 to the mix to provide higher levels of performance. Announced late last year, the Tesla K80 is Nvidia's top-end card and features twin GPU chips, each with 12GB of memory.

This option joins SoftLayer's other GPU accelerated cloud servers, which are currently powered by Nvidia Grid K2 and Tesla K10 GPUs.

SoftLayer chief technology officer Marc Jones said that the new capabilities allow the firm's bare metal servers to meet the requirements of most supercomputing workloads.

"By introducing the K80 accelerator (pictured) on IBM Cloud, we're giving our customers an even more powerful tool to run demanding applications," he said.

In fact, SoftLayer claims to be the only infrastructure-as-a-service provider to offer GPU-accelerated computing on bare metal servers.

In making the compute power of the Tesla K80 available as a cloud service, IBM is looking to give companies of all sizes easier and more affordable access to supercomputing resources.

This is especially important for organisations such as startups and research facilities, which typically need to start at a small scale, using only a few servers and GPU accelerators for testing and development workloads, SoftLayer said.

The announcement has only just been made public, but IBM said that it has several customers that have already been using bare metal servers with the Tesla K80, including New York University and startup firm MapD, which uses the capabilities to analyse large data sets.

Meanwhile, AMD unveiled its own latest offering for HPC in the data centre this week in the shape of a new GPU accelerator card for servers. The FirePro S9170 offers higher performance than the Tesla K80, according to AMD.