The main question many will ask is, “what’s the big deal?” or more specifically “What is this change, in technology and programming focus, and how is it going to affect me?”
Hopefully by the end of this article you’ll have an understanding of things to come and how it’s going to affect the world.
During this last week, I recently was lucky enough to attend a Nvidia’s GPU Technology Conference, which I’ll hereafter refer to as GTC. GTC was focused primarily on Programming Nvidia’s GPU processors that are installed in several lines of their video cards. The most common & recognized by many is the GeForce line of graphic cards, made more for the average consumers & game enthusiasts. The next line of cards discussed at GTC was the Quadro cards, with more of a tool for visualization industry, including design and animation. The final card discussed at their show was the Tesla cards, made specifically for parallel computing and to take advantage of the many cores prevalent in the GPU processor architecture.
GTC was really a three tier conference, combining development, research and an intro into new companies that are making great strides In the fields of GPU programming technology. The first tier was an emerging company’s summit where CEO’s and their corresponding corporations got the chance to speak in a lively and fast-paced exchange, interacting with a panel of industry analysts, investors and technology leaders in an 16 minute timeslot, with 8 minutes being a presentation of technology and their company and 8 minutes for discussion with analysts.
The second tier of GTC was Nvidia’s Researchers summit, more catered to researchers interested in using GPUs in science and engineering, and focused on learning how GPU computing can drastically increase computational power and dramatically reduce time-to-discovery. In this choice of sessions, researchers were shown how other colleagues are achieving tenfold to hundredfold speedups using GPUs in fields ranging from molecular dynamics to astrophysics, from genomics to fluid dynamics. Some of the highlighted research posters can be found on Nvidia’s Website here: http://www.nvidia.com/object/research_summit_posters_2010.html
The third and final tier of the summit was the Nvidia’s Developers summit, focused towards programmers and Developers, this summit had several sessions introducing key tools and technologies used in coding to the CUDA architecture. For those not familiar with CUDA, it’s the computing engine in NVIDIA graphics processing units (GPUs) that is accessible to software developers through variants of industry standard programming languages. Part of the highlights of these sessions included several tools and options for programming. NVidia’s own Parallel Nsight, was a major tool demonstrated for programming/Debugging CUDA apps in Microsoft’s Visual Studio suite (C,C++, Visual Basic), but also included several tools for those that aren’t familiar with that type of coding, for instance, Matlab computational tools, or even just programming in a spreadsheet such as MS Excel.
All three of these tiers were combined with a networking and exhibitor showcase that was open during lunch and around the evening, where attendees could see some of the technology in action. There were quite a few real highlights shown during the exhibit, some of the most memorable for me were demo’s on facial recognition, body movement & object tracking, racing simulations, touch screen and scalable displays, and not to mention many examples of stereoscopic 3D in use, including even portable phone LCD’s being shown without the need of any glasses. I was however a bit dismayed as most of the exhibits seemed to be focused on making arrangements to buy some of the new supercomputer server builds combined with Multi GPU’s solutions, which could really only be used by high level researchers or those with plenty of money to spare. The prices of some of these servers were upwards of $50,000 with even renting time on the servers costing $1200 a day for computational use. Some of the more recognizable names selling Supercomputer servers I saw were by such companies as Dell, GE intelligent platforms, Appro, SuperMicro, and Mazda Technologies.