Intel on Monday disclosed a version of its Xeon processor line optimized for supercomputers and announced a partnership with NEC to develop future supercomputers.
At Supercomputing 2009 in Portland, Ore., Intel unveiled a future version of its "Nehalem-EX" processor optimized for supercomputers. The six-core chip will run at higher speeds than eight-core versions of the Nehalem-EX processors and will offer advantages for supercomputer specific tasks, Intel said in a statement. Intel also refers to supercomputing as high-performance computing, or HPC.
The chip architecture will offer greater memory speeds and capacity and will allow customers to build single computers or "nodes" with up to 256 such processors, according to Intel. This will be available next year, Intel said.
Intel said Monday that four out of every five supercomputers on the Top500 list published Monday are powered by Intel processors.
Intel also announced that it is partnering with Japan's NEC--that country's largest supercomputer vendor--to jointly develop technologies "that will push the boundaries of supercomputing performance," according to a joint statement.
NEC will use the technologies in future supercomputers based on the Intel Xeon processor and other technologies such as AVX (Advanced Vector Extensions), an extension to Intel's x86 instruction set architecture.
AVX will be used with Intel's upcoming Sandy Bridge microarchitecture due in 2011, according to Intel.
"With NEC further innovating on Intel Xeon processor-based systems, Intel is poised to bring Intel Xeon processor performance to an even wider supercomputing audience, " said Richard Dracott, general manager of Intel's High Performance Computing Group, in a statement.
Fumihiko Hisamitsu, general manager of HPC Division at NEC, said: "NEC's substantial experience in the development of vector processing systems...is a natural fit for taking Intel architecture further into new markets."
A vector processor design can perform operations on multiple data elements simultaneously. Intel Xeon chips are good at scalar processing, which handles one data item at a time.
The initial focus of the collaboration will be the development of technology to boost the memory speed and scalability--the latter refers to expanding a system to increase performance or capacity. "Such enhancements are intended to benefit systems targeting not only the very high end of the scientific computing market segment, but also to benefit smaller HPC installations," the two companies said.
NEC will also continue to sell its existing SX vector processor-based products. NEC, for example, currently markets the SX-9 supercomputer.
As Intel prepares to invade Nvidia turf, large companies at the Intel server chip rollout Monday stated--in some cases quite objectively--what graphics chip suppliers need to do to make this technology more palatable for high-performance computing.
Lincoln Wallen, head of research and development at DreamWorks Animation
(Credit: Screen capture by Brooke Crothers)Besides competing in the gaming graphics market, Intel is eying large high-performance computing customers such as Dreamworks Animation (whose "Monsters vs. Aliens" opened last weekend to large box office numbers) for its future Larrabee graphics chip.
Nvidia is already a player in the so-called General Purpose GPU space, which applies graphics processing units (GPUs) to high-performance computing. As described by Nvidia, high-performance computing on the GPU uses a CPU and GPU together in a heterogeneous computing model, with the "sequential" part of the application running on the CPU and the computationally-intensive part running on the hundreds of processing cores built into the GPU.
Application developers have to modify their application to take the compute-intensive kernels (core components of an operating system) and map them to the GPU. The rest of the application remains on the CPU.
At the Intel "Nehalem" server chip event on Monday, a panel of representatives from large companies addressed the issue of CPU versus GPU. Currently, these customers are using CPUs to do their data crunching.
Keith Gray, manager, high performance and technical computing at oil giant BP, spelled out why he has hesitated to use GPUs to date while expressing interest in adopting them in the future. "Our business is about accelerating our development of new seismic imaging research algorithms. At this point we actually believe the level of programming difficulty (and) lack of standardization of application development tools make the move to accelerated computing a bit risky," he said.
CPU (left) versus GPU
(Credit: Nvidia)Gray continued. "We are watching the evolution of the programming interfaces. Once those are better standardized, once the issues of moving data back and forth from the general purpose system to an accelerator is addressed, we'll be very interested in taking advantage of it," he said.
Lincoln Wallen, head of research and development at DreamWorks Animation, is also looking into exploiting power of the GPU for tasks such as rendering. "We're looking forward to exploit more flexible compute models, perhaps involve more of the graphics processing functionality but tightly coupled with very powerful CPUs to address the particular way in which we generate images, very soft body, lots of geometry generation," he said.
Wallen continued that, as he sees it, Larrabee offers an advantage because of its tight coupling between the CPU and GPU. "The promise of Larrabee with that tight coupling and the programming model offers a great opportunity to start to explore that type of architecture for our particular workloads," he said.
A Dell slide shown Tuesday was a reminder that a future 80-core processor is still in sight.
Flash back two years to the Intel Developer Forum when CEO Paul Otellini pledged to deliver an 80-core processor in five years.
Otellini said at the time that the chips will be capable of exchanging data at a terabyte a second and that the company hopes to have these chips ready for commercial production within a five-year window.
Michael Dell referred to a slide showing an 80-core chip Tuesday at SC08, a conference in Austin, Texas, focused on high-performance computing.
The trend of packing more compute power into small supercomputing enclosures "is really driven by what's going on in microprocessors. The x86 revolution continues. You see more and more cores. Increased performance. But also without more power required," he said, speaking during the keynote.
Dell slide shown Tuesday at SC08
(Credit: Dell Computer)In various venues, Intel has spelled out its intention to bring out many-core processors including its upcoming Larrabee graphics chip and future server processors that may reach 32 cores. Currently, Intel's Dunnington processor gets the prize (at Intel) for the most cores: six. Sun Microsystem's "Rock" processor will have 16 cores.
The 'Stallion' Visualization Cluster.
(Credit: Dell)Democratize IT. A banal catch phrase until you see off-the-shelf gaming boxes from PC maker Dell being used for visual supercomputing.
CEO Michael Dell showed the "Stallion" Visualization Cluster at the Texas Advanced Computing Center (TACC) running on standard Dell XPS gaming machines during his keynote Tuesday at SC08, a conference in Austin, Texas, focused on high-performance computing. (The keynote was streamed over the Web.)
The Stallion "visualization wall" uses XPS boxes to power 30-inch Dell displays. "The largest display of its kind in the world, at 307 million pixels," Michael Dell said.
"Literally these are gaming systems. We just leverage what was going on the commodity technology market," said Kelly Gaither, associate director at TACC, speaking as part of Dell's keynote address.
Dell is also looking to Nvidia to democratize supercomputing and bring it down to the desktop. "Advances in graphics technology are actually creating some new opportunities in supercomputing," Dell said. "We announced today that we're extending our partnership with Nvidia to advance their CUDA architecture in Dell's precision workstations," he said.
"So this really is the supercomputer on your desk. Adding one (Nvidia) Tesla card to Dell Precision workstations delivers a theoretical performance of 1 teraflop," he said. "That's seven times higher than (a high-end) Thinking Machines (supercomputer) back in 1993." (A teraflop is one trillion floating point operations per second.)
Dell also announced Tuesday that it has teamed up with Intel and Lawrence Livermore National Laboratory (LLNL) to build the Hyperion hyperscale computing environment at LLNL. The National Nuclear Security Administration's Advanced Simulation and Computing Program at the facility expects Hyperion to speed the development and reduce the cost of powerful high-performance computing clusters vital to U.S. Department of Energy and National Nuclear Security Administration missions, including climate change, and other global challenges.
"Hyperion is a test bed that will share those breakthroughs with the entire open-source community," Dell said.
Dell Precision workstation becomes supercomputer with Nvidia Telsa technology
(Credit: Dell)But Dell's big message was how mass-market and standard commercial computing technologies have invaded the supercomputing space. "429 of the top 500 supercomputers are based on the x86 architecture," Dell said, referring to the computing architecture being used in laptops and desktops today. "What you see here is some of the things from the commercial world in managing large data centers really penetrating very heavily (into high-performance computing)."
"Three years ago, using our blade chassis, we put 240 cores in a full-size 42U rack with 2.8GHz CPUs, and that was 1.3 teraflops of theoretical peak performance per rack. Today, we get 512 cores in a 42U rack with 3.3GHz CPUs, and that's 6.82 teraflops per rack," Dell said. ("U" is a unit of measure that describes the height of equipment used in a rack computer. Typically, 1U equals 1.75 inches.)
More addressable memory space--critical for high performance computing--will come with Intel's Nehalem processor, he said. Nehalem will support memory spaces of up to 1 terabyte (trillion bytes) of system memory, Dell said. Most PCs today support 4 gigabytes (billions of bytes).
In related news, Nvidia announced that Tokyo Institute of Technology (Tokyo Tech) will use Nvidia Tesla GPUs to boost the computational horsepower of Tokyo Tech's Tsubame supercomputer.
Through the addition of 170 Tesla S1070 1U systems, the supercomputer now delivers nearly 170 teraflops of theoretical peak performance, placing it among the world's Top 500 Supercomputers.
"Tokyo Tech is constantly investigating future computing platforms and it had become clear to us that to make the next major leap in performance, Tsubame had to adopt GPU computing technologies," said Satoshi Matsuoka, division director of the Global Scientific Information and Computing Center at Tokyo Tech in a statement.
- prev
- 1
- next
