| CPD Seminar 2012-09 HPC and CUDA |
|
| September 2012 | |
|
Compucon CPD Seminar
4:00 – 4:15pm Outside the Square Time Outside the Square Time It is about time that we talk about time. This first word of time refers to calendaring. The second word of time has a much wider scope of meanings. Time has existed longer than human beings and therefore human can consider time as a resource of infinity. This session will avoid stepping into the philosophical nature of time and focus on the efforts human has spent to make time to be more useful to operating our life. The scope contains Greenwich Mean Time, Master Clock, GPS, Network Time Protocol, Olympics time measurement, and a light touch of meta-physics. Parallel & High Performance Computing Most software applications have not been programmed for parallel computing despite the progress made in integrating multiple cores in a CPU (central processing unit). Do we need parallel computing at all then? The short answer is very much so. Do we want to spend 24 hours just to simulate 46ns of bio-molecular activities? Apart from poking into nature like the question above, we would like to produce our movies and vehicle designs in a shorter time frame than our competitors. In an average office, doing more in lesser time contributes to productivity and competitiveness. How fast are the most modern computer systems in the world and on our desktops? This session will provide a high level introduction to the area of parallel and high performance computing. It will address what the global computing industry has been trying to achieve in recent years. It will include a preview of the Pirates of the Caribbean, a glimpse of Andromeda running into Milky Way, and a demonstration of how the concept of Maximus works to speed up a design process. Integrating CUDA Technology Compucon has chosen to ride in the Nvidia CUDA bandwagon as a business decision made on 1 January 2011 after having a couple of direct contacts with Nvidia. The main basis for the decision is the trust in Nvidia being able to see the industry far ahead and to execute its technology visions. CUDA is both a parallel computing framework and a hardware implementation. The framework was first launched in 2007 and has now reached version 4.2 heading for version 5.0 at the end of this year. Hardware cores are in the 2nd generation called Fermi and the 3rd generation called Kepler is being rolled out progressively at present. CUDA is not a cult but is being taught in 500 universities around the world. This session will explain the CUDA framework (software and hardware), real world competitions, and the soon to be released K20 hardware capabilities. We will introduce a low cost CUDA machine for the desktop within the reach of small companies, as well as an entry level CUDA server with as many as 5000 cores for medium size organisations. We expect a sea change of high performance computing landscape in the industry from now on. END |