Previously William J. Dally, Chief scientist at Nvidia Corp, predicted Nvidia GPUs in 2015 will be implemented on 11 nm process technology that feature roughly 5,000 cores and 20 teraflops of performance. Current Nvidia GPUs have 500 gigaflops of performance in single precision. 20 teraflops would be 40 times faster. 570 times faster in 2016 would be 285 teraflops. However, if Huang was referring to double precision then the increase would be from the current 100 gigaflops going up to 57 teraflops of double precision performance. This seems to make more sense and is more consistent with the 20 teraflop in 2015 statement. 57 teraflops of double precision performance in 2016.
3 teraflop GPGPU chips should be available from Nvidia in late 2009 or early 2010.
The current Tesla GPUs are running at 1.3-1.4 GHz and deliver about 1 teraflop, single precision, and less than 100 gigaflops, double precision. Valich speculates that a 2 GHz clock could up that to 3 teraflops of single precision performance, and, because of other architectural changes, double precision performance would get an even larger boost.
This statement was made as part of "GPU Computing Revolution" keynote speech by Jen-Hsun Huang at the Hot Chips 21 conference.
Huang - who made his comments at the Hot Chips symposium in Stanford University - explained that such advances could enable the development of realtime universal language translation devices and advanced forms of augmented reality. Huang also discussed a number of "real-world" GPU applications, including energy exploration, interactive ray tracing and CGI simulations.
Nvidia website for its Tesla GPGPU computing solutions is here
There is a GPU technology conference Sept 30-Oct 2, 2009 in San Jose
The transcript (from Seekingalpha) of Nvidia's quarterly earnings conference call was on August 6, 2009
After three years of evangelizing, GPU computing has surely reached the tipping point. CUDA has been adopted in a wide range of applications. In consumer applications, nearly every major consumer video application has been or will be accelerated by CUDA. We estimate there are over 1200 research papers based on CUDA. We’ve highlighted 500 of them on CUDAZone.com.
CUDA now accelerates Amber, an important molecular dynamic simulation program used by more than 60,000 researchers in academia and pharmaceutical companies worldwide to accelerate new drug discovery. CUDA sped up Amber 50 times.
For the financial market, numerics and compatible announced CUDA support for their new counter party risk application and achieved an 18 times speed-up. Numerics is used by approximately 375 financial institutions.
There are broad ranging uses for CUDA including astro physics, computational biology and chemistry, fluid dynamic simulation, electromagnetic interference, CT [image reconstruction], seismic analysis, raytracing and more.
Another indicator of CUDA adoption is the ramp of our new TESLA GPU for computing business. There are now more than 700 GPU clusters installed around the world with new Fortune 500 customers ranging from Schlumberger and Chevron in the energy sector to BNP Paribas in banking.
And starting this fall with the launch of Microsoft’s Windows 7 and Apple’s Snow Leopard, GPU computing will go mainstream. In these new operating systems, the GPU will not only be the graphics processor but also a general purpose parallel processor accessible to any application.
Recently John Petty, a leading industry analyst, forecast the global graphics market to grow nearly 22% in 2010, based in part to the rise of the GPU as a co-processor. The report states the continued expansion and development of heterogeneous computing and GPU compute will stimulate growth in 2010, enabled by Apple’s and Microsoft’s new operating system, new programming capabilities using OpenCL, Direct Compute, and NVIDIA's CUDA architecture will remove barriers to the exploitation of the GPU as a serious economical and powerful co-processor in all levels of PCs.
TESLA is available as a module, a desk side personal super-computer or server for high performance computing clusters. TESLA achieved its first significant quarter of revenue with approximately $10 million in sales. Virtually every major OEM, including [Cray], Dell, HP, IBM, Lenovo, Silicon Graphics, or excuse me, SGI, Sun, and Super Micro now offers TESLA based solutions
We have over 50 HPC specialized VARS currently selling TESLA today. We estimate there are approximately 1,000 VARS actively involved in the HPC market which we have yet to engage.
We estimate TESLA to address a $5 billion market opportunity for us over the next three years.
We also know that high resolution displays and projectors are becoming more affordable than ever. The Sony 4K projectors, digital projectors are very affordable and people need scalable resolution, scalable visualization solutions to be able to address that and so we created a new product called Quadro SVS. And the Quadro SVS virtualizes both the application as well as the display, so you could literally run one application across up to four GPUs completely virtualized and invisibly, and then you can take the output of that and literally drive it up to 32 million pixels without the application ever knowing anything about it. And so this virtualization technology both at the GPU level as well as the display level is a groundbreaking idea and it’s something that we are really excited about.
TESLA servers consumers nearly 20 times less power than a conventional CPU server and the reason for that is because of the amount of performance that you get out of it.
Nextbigfuture had an interview with Nvidia's Sumit Gupta.
Forbes had an interview in June 2009 with Huang