Nvidia Powering World's Most Powerful Supercomputer
Oak Ridge National Laboratory has chosen Nvidia's Tesla GPU to power a new supercomputer for the U.S. Department of Energy.
Tuesday Nvidia said that Oak Ridge National Laboratory has chosen Tesla GPUs to power a new Cray XK6 supercomputer called "Titan" for the U.S. Department of Energy's open science computing facility located in Tennessee.
Titan will integrate 299,088 AMD Opteron processing cores (18,688 16-core CPUs) as well as 7000 to 18,000 Nvidia Kepler GPUs. The system, which is scheduled to completed in late 2012, will also integrate 600 TB of memory. ORNL said that Titan will be ready for users sometime in 2013.
Nvidia said that the new computing beast has the potential to deliver over 20 petaflops of peak performance, more than two times faster and three times more energy efficient than today's fastest supercomputer, the K computer located in Japan. Steve Scott, the CTO of Nvidia’s Tesla business unit, added that Titan will be ten times more powerful than the current Jaguar machine. 85-percent of the computing power will come from the Tegra chips while the other 15-percent will be handled by conventional AMD CPUs.
According to Scott, GPUs are better at floating point operations than CPUs, and they're more power efficient. "We’ve reached the point where processors have become power constrained," he told AllThingsD. "If you pack all the transistors that you can onto a chip and run it as fast as you can, the chip will melt. We’ve entered a time where performance is constrained by power, and its only going to get worse, so you need processors that are power efficient. It’s a fundamental sea change in the underlying technology of high performance computing."
Jeff Nichols, associate laboratory director for Computing and Computational Sciences at Oak Ridge National Laboratory, said that Titan will be used for research in a broad range of fields, including material science, energy technology, medical research, geoscience, and others. "All areas of science can benefit from this substantial increase in computing power, opening the doors for new discoveries that so far have been out of reach," he said. "Titan will be used for a variety of important research projects, including the development of more commercially viable biofuels, cleaner burning engines, safer nuclear energy, and more efficient solar power."
The first stage of deployment is already underway, as Oak Ridge is currently upgrading its existing Jaguar supercomputer with 960 Tesla M2090 GPUs based on Nvidia's "Fermi" architecture. These GPUs will serve as companion processors to multi-core CPUs in the Cray XK6 supercomputer. In the second phase -- which is expected to begin in 2012 -- Oak Ridge plans to deploy up to 18,000 Tesla GPUs based on the next-generation architecture code-named "Kepler." The peak performance will end up in the range of 10 to 20 PFlops, depending on "architectural options", ORNL said.
The lab hopes that Titan will deliver result in research that targets the simulation of commercial biofuel production, the development of new materials for photocells, and the simulation of extended lifecycles for nuclear power plants.
Stay On the Cutting Edge: Get the Tom's Hardware Newsletter
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.
-
Inferno1217 gmcizzleBut can it run BF3?Reply
BF3 is much better coded than the mess known as Crysis. The highest end cards at the time could not run it at full settings when it was released. BF3 does NOT suffer from this. Crysis was poorly coded.
-
dragonsqrrl gmcizzleBut can it run BF3?Ya, sorry dude, that question just doesn't hold the same weight as it did with Crysis. BF3 maxed at 1920x1200 actually performs quite well on high end single GPU configurations.Reply -
dragonsqrrl It's amazing how quickly the floating point performance of modern supercomputers have increased. It seems like just yesterday the first PFLOP system launched, and next year we could potentially have a 20 PFLOP system. This is no doubt related to the integration of GPU coprocessors in recent supercomputing systems.Reply -
gerchokas With billions (or trillions?) of transistors, how close can it be to emulating a human brain?? I want true AI!!!Reply