HP and Nvidia have officially opened a joint GPU Technical Centre of Excellence in Grenoble, France, as part of a joint initiative aimed at further boosting the acceptance of GPU-based high-performance computing (HPC) systems.
For many workloads, graphics processing units (GPUs) offer considerable advantages to traditional central processing units: where a CPU may have as many as 16 processing cores, a single GPU can have hundreds - and while each individual core isn't the most powerful at general-purpose computing, together they can make light work of highly-parallel tasks.
This is the reason why GPU-accelerated systems have been taking over the TOP500 list of supercomputers in recent years, but it's something HP and Nvidia believe could go still further. As a result, the two companies have pooled their resources into the opening of, to give the facility its full title, the HP Nvidia GPU Technical Centre of Excellence.
'HPC systems require huge amounts of compute resources to achieve their expected performance, as well as the expertise to integrate them,' explained HP's Philippe Trautmann, of the partnership. 'The goal of this collaborative GPU Technical Centre of Excellence is to enable perennial solution improvements and facilitate the adoption of HP systems based on Nvidia Tesla GPUs for HPC.'
The facility provides researchers with remote access to an HP cluster platform comprised of 10 ProLiant SL250s, SL270s and ML350p Gen 8 servers, each of which includes integrated Nvidia Tesla GPU accelerator boards along with a smaller number of Nvidia Grid GPU boards for GPU virtualisation experiments. Local users can take advantage of a Z820 workstation to access the cluster, while remote users are offered ProLiant WS460c Gen 8 blade systems.
'GPU-accelerated computing is enabling new breakthroughs across an ever increasing range of scientific and engineering fields,' claimed Nvidia's Walter Mundt-Blum at the opening. 'By providing access to the latest HPC technologies and expert guidance, we are arming European researchers and engineers with everything they need to drive new waves of discovery and innovation.'
The centre isn't to be open-access, unfortunately: the compute facilities, along with the advice and assistance of HP and Nvidia staff at the centre, is made available exclusively to HP's customers and partners - meaning researchers who have opted for a different hardware vendor, or who are rolling their own GPU-accelerated HPC platforms, are unable to take advantage of the new facility.