December 13, 2012
In the world of high performance computing, a lot of the important news this year revolved around heterogeneous computing, big data, and HPC interconnects. Two vendors that perhaps embodied those technologies more so than others were Cray and Intel, both of which figured prominently in some of the biggest HPC stories in 2012. Here are HPCwire's highlights and lowlights for the year.
Hit: Accelerators Break Teraflop Barrier
With the introduction of NVIDIA's Kepler K20/K20X GPUs, Intel's newly hatched Xeon Phi line, and AMD's new FirePro GPU offering in November, users can now buy a teraflop computing device for just a few thousand dollars. NVIDIA has established itself as the market leader and its latest K20X GPU looks to keep that role intact. Intel's entrance into the market with its Knights Corner coprocessors, and AMD's re-entrance with its server-grade FirePro S10000, will offer some interesting alternatives though, and should widen the aperture significantly for heterogeneous computing in the year ahead.
Hit: Sandy Bridge CPUs Debut
Before Intel launched the Knights Corner products, the chipmaker was already making its mark on the supercomputing landscape with its mainstream Xeon line. The first Sandy Bridge Xeon parts (the E5-2600 processor family) were officially released in March. At that point though, there were already at least ten TOP500-class supercomputers being powered by the chips using early production parts. Sandy Bridge represents a new microarchitecture, featuring PCIe Gen3 on chip, a faster memory interface, an integrated I/O hub, and AVX (Advanced Vector Extensions). The latter is Intel's new 256-bit vector instruction set that effectively doubles FP throughput.
Miss: AMD Gets Existential
AMD, suffering from a weakened PC market in 2012, announced in October it would lay off 15 percent of its workforce this year. The Opteron chip division, which serves the HPC server space was in somewhat better shape though. The company launched its new "Piledriver"-based Opteron 6300 in November, but there was little hope that the new line would do anything to reverse Intel's dominance of the server chip market. To top it off, in November, rumors surfaced that the chipmaker was looking to sell the company, or at least some its assets, to interested parties. AMD denied the rumors.
Hit: HPC Cozies Up to Big Data
HPC vendors embraced "big data" in a big way in 2012, or at least their marketing departments did. And as some in the industry have argued, HPC and big data are joined at the hip. HPC server makers, interconnect vendors, storage companies, would tend to agree; in 2012, they all jumped on the bandwagon. Supercomputing technology is shaping up to be a key ingredient for big data machinery.
Hit: Cray Yarcs Up Big Data
To make the HPC-big data connection official, supercomputer-maker Cray launched into the business with a brand new division devoted to the technology. Dubbed YarcData (Yarc is Cray spelled backward) the group was announced in February and unveiled its first commercial product in March. The appliance, known as uRiKA, recycles Cray's XMT hardware to deliver a graph analytics platform that is aimed at "real-time knowledge discovery." So far the company has sold a handful of the machines.
Miss: Feds Suck the Energy Out of SC12
Thanks to fallout from GSA financial misconduct at an employee training conference in 2010, US government funding for Department of Energy (DOE) booths and attendance at the Supercomputing Conference (SC12) in November was curtailed significantly this year. The feds subsequently issued an edict to curtail travel expenses, which led to a $100K-per-conference spending limit for the DOE. Effectively that meant no pretty booths for any of the energy agency's labs and a smaller contingent of DOE attendees roaming the conference halls this year.
Hit: Titan Wins Linpack Sweepstakes
The DOE might have been missing in action at SC12, but its supercomputers got plenty of attention. The GPU rework of Cray's Jaguar supercomputer at Oak Ridge National Lab (ORNL) captured the number one spot on the new TOP500 list. Titan, as it is now known, delivered 17.6 petaflops on Linpack, knocking another DOE supercomputer, Lawrence Livermore Lab's Sequoia, into second place. The NVIDIA GPU-powered Titan also bested Sequoia, an IBM Blue Gene/Q machine, in energy efficiency.
Hit: Cray Acquires Appro
The most significant supercomputing acquisition of the year came in November when Cray shelled out $25 million to buy Appro, a privately owned HPC cluster-maker. Appro, which has been in the business for 20 years, makes commodity supercomputer clusters (commodity inasmuch as Appro gear relies on industry-standard interconnects like InfiniBand), mainly for the US market. The move expands Cray's portfolio considerably and is expected to add $60 million worth of revenue to the company's top line in 2013.
Hit: Intel Rolls Up Fabrics
In 2012 Intel bought up two HPC interconnect technologies and started to devise a fabric strategy around them. In January, the chipmaker anted up $125 million to buy QLogic's InfiniBand assets. And in April, Intel added to its collection by acquiring Cray's custom interconnect IP and expertise for $140 million. Intel's strategic interest in the technology is to bring these fabrics on-chip, essentially integrating a NIC onto the processor die. The move dovetails with Intel's plans to provide silicon componentry for exascale machines.
Miss: US Exascale Plans in Limbo
With DARPA's Ubiquitous High Performance Computing (UHPC) program now but a distant memory, it fell to the DOE to pick up the exascale slack for the US. But now we've learned that program is also running into delays. According to a presentation delivered at SC12 by William Harrod, the Research Division Director of the DOE's Office of Science Advanced Scientific Computing Research (ASCR) program, the agency's original plans of fielding its first exaflop machine before the end of the decade is in jeopardy. Harrod said that even if the government approves exascale funding next year, the first such machines are probably not going to arrive before 2020, and more likely, not until 2022.
May 23, 2013 |
The study of climate change is one of those scientific problems where it is almost essential to model the entire Earth to attain accurate results and make worthwhile predictions. In an attempt to make climate science more accessible to smaller research facilities, NASA introduced what they call ‘Climate in a Box,’ a system they note acts as a desktop supercomputer.
Read more...
May 22, 2013 |
At some point in the not-too-distant future, building powerful, miniature computing systems will be considered a hobby for high schoolers, just as robotics or even Lego-building are today. That could be made possible through recent advancements made with the Raspberry Pi computers.
Read more...
May 16, 2013 |
When it comes to cloud, long distances mean unacceptably high latencies. Researchers from the University of Bonn in Germany examined those latency issues of doing CFD modeling in the cloud by utilizing a common CFD and its utilization in HPC instance types including both CPU and GPU cores of Amazon EC2.
Read more...
May 15, 2013 |
Supercomputers at the Department of Energy’s National Energy Research Scientific Computing Center (NERSC) have worked on important computational problems such as collapse of the atomic state, the optimization of chemical catalysts, and now modeling popping bubbles.
Read more...
05/10/2013 | Cleversafe, Cray, DDN, NetApp, & Panasas | From Wall Street to Hollywood, drug discovery to homeland security, companies and organizations of all sizes and stripes are coming face to face with the challenges – and opportunities – afforded by Big Data. Before anyone can utilize these extraordinary data repositories, however, they must first harness and manage their data stores, and do so utilizing technologies that underscore affordability, security, and scalability.
04/15/2013 | Bull | “50% of HPC users say their largest jobs scale to 120 cores or less.” How about yours? Are your codes ready to take advantage of today’s and tomorrow’s ultra-parallel HPC systems? Download this White Paper by Analysts Intersect360 Research to see what Bull and Intel’s Center for Excellence in Parallel Programming can do for your codes.
In this demonstration of SGI DMF ZeroWatt disk solution, Dr. Eng Lim Goh, SGI CTO, discusses a function of SGI DMF software to reduce costs and power consumption in an exascale (Big Data) storage datacenter.
The Cray CS300-AC cluster supercomputer offers energy efficient, air-cooled design based on modular, industry-standard platforms featuring the latest processor and network technologies and a wide range of datacenter cooling requirements.