The OpenGL facilities were not activated on the GPUs. NVIDIA has been collaborating with Cray and CSCS to activate the OpenGL features on the chips so that the scientists not only can use them for the computation but also for the visualisation. This opens a whole new avenue for research and interactive use of the machine which the researchers could not conceive of before.
On the GPUs the scientists have experimented with a number of applications that include neuron visualisations and chemical simulations.
John Biddiscombe went over to the NVIDIA booth at the ISC'14 Exhibition to showPrimeur Magazineon one of the screens a live simulation running on Piz Daint coupled to the OpenGL visualisation that we could see happening as it took place.
On the screen John Biddiscombe showed a simulation of molecular dynamics running on 64 nodes of Piz Daint using the NAMDI solver. The visualisation is using one of the GPUs that has been enabled for graphics on the same machine. The simulation and the visualisation are coupled together so this was happening live, taking place in Switzerland as we spoke.
The work is done by collaborators at NVIDIA and the University of Illinois at Urbana-Champaign. It really shows the advantage of having the GPU enabled for graphics so no data is saved to disk, nothing is being copied. It is rendered as it happens and you get these sorts of images, John Biddiscombe showed.
In addition to this, the scientists have also enabled ParaView, Viz-it, and custom OpenGL applications which can run on the same nodes. The advantage is that, even if the scientists don't perform any visualisation, they can render the data on the same machine as it was generated on, on the same scratch file system so they don't have to copy data and they don't have to mount the file systems on other machines. This is a huge game for everybody involved.