11/28/2012 3:42:49 PM
O profesionálních akcelátorech na architektuře NVIDIA GK110 Kepler jsme již psali v předchozím příspěvku. Tento týden společnost NVIDIA oficiálně uvedla novinky na trh, a tak je v tuto chvíli již možné Tesla K20X a Tesla K20 zakoupit u autorizovaných prodejců. O výkonu těchto čipů svědčí fakt, že převážně díky nim se superpočítač Cray Titan umístil na vrcholu žebříčku nejvýkonnějších HPC systému světa.
Pokročilejší z obou modelů, tedy K20X, obsahuje těžko uvěřitelných 2.688 jader CUDA na jednom čipu, dále 16 streaming procesorů a 6 GB paměti RAM se zvýšenou paměťovou propustností. S výkonem 1,31 teraflopů v operacích s plovoucí desetinnou čárkou s dvojitou přesností je nový Kepler dvakrát tak rychlý v surových číslech než jeho protějšek Fermi a ještě více v reálných aplikacích. Navzdory tomu spotřebuje méně energie.
Na rozdíl od předchozí generace K10 má řada K20 kompletní vnitřní a vnější paměťovou pipeline standardu ECC, nechybí podpora pro dynamickou paralelnost a funkce Hyper-Q. Použití ECC pamětí si vyžádalo pokles paměťové kapacity o 12,5 procenta a úbytek malé části výkonu paměti.
Podrobnou technickou specifikaci najdete zde.
11/15/2012 3:52:11 PM
Tesla K20 a K20X s jádrem GK110 jsou nové profesionální grafické čipy od společnosti NVIDIA, které najdou uplatnění především v systémech typu HPC, kde kromě vysokého výkonu je vyžadována rovněž podpora technologií umožňující specifické typy výpočtů. Současně s uvedením novinek na trh vydala NVIDIA seznam více než stovky akcelerovaných aplikací, včetně informací o nárůstu výkonu v porovnání s předchozí generací těchto profesionálních čipů. Kompletní přehled naleznete zde
6/6/2012 2:02:44 PM
5/30/2012 5:39:36 PM
Intel Corporation announced today that the Massachusetts Institute of Technology (MIT) has won a national search to become the newest Intel Science and Technology Center
(ISTC) for Big Data and will be hosted by MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL).
4/26/2012 5:40:42 PM
Want to be the first one on your block to own a Kepler Tesla GPU? NVIDIA will be giving three of them away as prizes for a petaflop application contest they are hosting. According to a blog post by Nvidian Sumit Gupta, they are looking for "proposals from researchers around the world on what you would do if you had access to a dedicated petaflop supercomputer." More >>
3/22/2012 6:13:35 PM
NVIDIA debuted its much-talked-about Kepler GPU this week, promising much better performance and energy efficiency than its previous generation Fermi-based products. The first offerings are mid-range graphics cards targeted at the heart of the desktop and notebook market, but the more powerful second-generation Kepler GPU for high performance computing is already in the pipeline. More >>
3/7/2012 6:19:28 PM
Intel officially launched its new Xeon E5-2600 processor family on Tuesday, months after the chips had been deployed in supercomputers at several major HPC sites around the world. The new CPU represents the company's latest Xeon offering for dual-socket servers, and boasts a number of new features including better performance, a new floating point instruction set in AVX, and integrated I/O. The processor will be a formidable competitor in the server chip battle with AMD. More >>
1/23/2012 2:29:45 PM
Chip giant Intel has penned an agreement to acquire QLogic's assets related to its InfiniBand product line. The move is in line with Intel's strategy to have a broader, more diverse set of offerings for the datacenter, especially in the lucrative network and interconnect segment. The chipmaker also sees the acquisition as key to fulfilling its promise to deliver exascale technology by 2018.
12/29/2011 3:03:23 PM
The supercomputing biz seems to have shaken off most of the after-effects of the global recession, with scads of new deployments large and small around the world. China, in particular, continued its big push into HPC, notching its first home-grown super. And Japan ushered in the era of 10-petaflop supercomputing this year with its world-beating K Computer. But, as always, not all the HPC news was rosy. Here are the top hits and misses for the year.
12/1/2011 10:41:23 AM
The leap forward in genomics technology promises to change health care as we know it. Sequencing a human genome, which costs millions of dollars just a few years ago, now costs thousands. And the prospect of mapping a genome for under a thousand dollars is on the horizon.
11/23/2011 10:38:21 AM
As has become apparent to nearly everyone in the HPC community, life beyond petascale supercomputing will be power limited. Many efforts around the world are now underway to address this problem, both by commercial interests and researchers. One such effort that brings both into play is the Mont-Blanc research project at the Barcelona Supercomputing Center (BSC), which is looking to exploit ARM processors, GPUs, and other off-the-shelf technologies to produce radically energy-efficient supercomputers.
11/21/2011 10:36:01 AM
CUDA introduces developers to a number of new concepts (such as kernels, streams, warps and explicitly multi-level memory) that are not encountered in serial or other parallel programming paradigms. In addition, CUDA is frequently used alongside MPI parallelism and host-side multi-core and multi-thread parallelism. The TotalView parallel debugger provides developers with methods to handle these CUDA-specific constructs, as well as an integrated view of all three levels of parallelism within a single debugging session.
11/11/2011 10:32:43 AM
The CUDA Toolkit v4.1 release candidate (RC1) is now available to CUDA Registered Developers. New features include an open source LLVM-based compiler, 1000+ new image processing functions and a redesigned Visual Profiler with automated performance analysis.
10/11/2011 5:00:34 PM
Oak Ridge National Laboratory has awarded a contract to Cray Inc. to increase the Jaguar supercomputer's science impact and energy efficiency. The upgrade, which will provide advanced capabilities in modeling and simulation, will transform the DOE Office of Science-supported Cray XT5 system, currently capable of 2.3 million billion calculations per second (petaflops), into a Cray XK6 system with a peak speed between 10 and 20 petaflops. More here
9/27/2011 4:54:05 PM
Financial services firms are examining every aspect of their infrastructures to squeeze any delays out of their end-to-end computational workflows. This quest to lower latencies in each step of processing trades, and to perform other chores, was a common theme at the High Performance Computing Financial Markets Conference held in New York, last week. More here
9/1/2011 4:10:53 PM
NVIDIA started shipping Tesla M2075 GPU computing modules. It is optimized for power and performance as a successor to Tesla M2070. With Tesla M2075, customers can take advantage of 10% lower power for the same performance as its predecessor. In addition, M2075 improves ECC reporting. Applications that use a lot of cuFFT will see a performance improvement. Solutions based on Tesla M2075 are available from all leading OEMs.
8/1/2011 4:21:09 PM
A researcher at the New Mexico State University is modeling new ways to address common challenges with data-intensive, graph-based problems.
7/28/2011 4:23:55 PM
A new language could improve the quality of parallel code and automate some of the trickiest elements of multicore programming.
6/1/2011 4:29:26 PM
Two Important products have are now in production release: CUDA 4.0- Features include Unified Virtual Addressing (UVA), Thrust C++ Template Performance Primitives Libraries and GPUDirect 2.0 GPU peer-to-peer communication technology. Download at www.nvidia.com/getcuda
. Parallel Nsight 2.0- Features include full support for Microsoft Visual Studio 2010, CUDA 4.0, PTX/SASS Assembly Debugging, CUDA Debugger Attach to Process, CUDA Derived Metrics and Experiments, as well as graphics performance and stability enhancements.
5/23/2011 12:00:00 PM
Computational scientists at the Pittsburgh Supercomputing Center (PSC) and HP Labs have achieved speedups of nearly 10 times with GPUs (graphic processing units) versus CPU-only code (and more than 1000 times versus an implementation in a high-level language) in k-means clustering, a critical operation for data analysis and machine learning.
3/8/2011 12:00:00 PM
AccelerEyes today released version 1.7 of Jacket for GPU computing with MATLAB. Version 1.7 delivers even more speed to MATLAB with a new Sparse Linear Algebra Library, a new Signal Processing Library, a big boost to convolution functions, and more.
2/10/2011 12:00:00 PM
NVIDIA today announced that the third annual GPU Technology Conference (GTC) will return to San Jose's McEnery Convention Center from Oct. 11-14, 2011. In addition, Los Alamos National Laboratory will co-locate its Accelerated High Performance Computing (HPC) Symposium at the conference.
2/1/2011 12:00:00 PM
Revolution Analytics, the leading commercial provider of software and support for the popular open source R statistics language, today released Revolution R Enterprise 4.2, which, for the first time, adds the ability to natively import data from SAS files into the R environment without data size limitations. Now, companies looking to apply the advanced statistical capabilities of R to data locked in the SAS environment can easily import the data into Revolution R Enterprise.
12/22/2010 12:00:00 AM
11/17/2010 12:00:00 PM
11/17/2010 12:00:00 PM
9/2/2010 3:20:00 PM
1/14/2010 12:00:00 PM
11/11/2009 12:00:00 PM
8/14/2009 4:00:00 PM
7/27/2009 8:00:00 AM
6/1/2009 12:00:00 PM