The HPC User Forum is underway in Houston this week and our volunteer correspondent Sharan Kalwani has been busy tweeting live from the show:
- Dylan Brennan of the Univ of Tulsa, OK talks on the US Burning Plasma organization (USBPO group of scientists) use of HPC – Fusion research
- Lee Berry of ORNL worries about the problems/challenges and successes of using HPC codes for Fusion simulations
- CASL have used the Denovo code to model the recent Japanese nuclear disaster – checking the does rates in that facility
- Tom Evans of the ORNL-CASL (Consortium for Advanced Simulation of Light water Reactors) uses HPC to simulate/design of nuclear reactors
- Cynthia MacIntyre of the Council for Competitiveness talks about their Midwest Project for SME-OEM use of modeling and simulation.
- Suzy Tichenor of ORNL talks about the SciDAC workshop on commercializing the SW developed by the National Labs.
- Melyssa Fratkin of TACC shares their vision of how Industry outreach is being implemented.
- Ashok Krisnamurthy is the first speaker – OSC has done well to come up with some very useful industry portal based applications.
- Merle Giles moderating panel on how to enable/accelerate HPC Industry outreach – OSC, TACC, ORNL, Competitive Council are all represented.
- However using the capabilities for “Big Data” requires NEW ANALYTICS! both for data at rest and data in motion. Scales up and down….
- Dave Turek, VP IBM Deep Computing tells a fascinating story of the IBM cloud for doing the Power7, Sony, MS and Nintendo engineering designs
- Stephen Wheat, Sr. Director of HPC at Intel, branches off into a very different direction – an update on the Democratization of HPC!
- Dr. Newhauser of the Anderson Center @ UT Houston, tells us all about Computational Radiotherapy for Cancer Research (risk/predictions, etc)
- Keith Cheng of Penn State Hershey Medical Center tells us all about the deluge/flood of data from the HPC bio community – post lunch session
- Peter ffoulkes of Adaptive Computing shares a formal product roadmap update on MOAB – Viewpoint 2.0 tuned for HPC as a Service & Cloud
- The uHPC panel agrees on the inescapable need for SW/HW co-design for the eXaFLOP project, the old traditional way is no longer sustainable.
- Arun Rodriguez (Scalable Architecture) of Sandia National Labs is closing out the uHPC session – sharing info on their X-caliber project
- Angstrom is envisioned to be about a 1000 cores, each at 5 TFLOPS/s, 1.5 GHz and 50W (50 mW/tile) power. Each core uses 11nm technology
- Angstrom project == optimize both energy *and* performance