Back to Table of contents

Primeur weekly 2012-03-05

The Cloud

Australian National University selects SGI to help harness explosion in data ...

Big science teams up with big business to kick‐start European Cloud computing ...

AMD to acquire SeaMicro: accelerates disruptive server strategy ...

HP launches Enterprise Security Cloud Connections partner programme ...

Dell advances public Cloud security with Trend Micro SecureCloud solutions ...

EuroFlash

Stories from the Grid launched at International Symposium on Grids and Clouds ...

IBM selected to build new supercomputer of the Max Planck Society ...

A supercomputer for research: University of Oldenburg to upgrade technical infrastructure ...

Workshop Call for Abstracts: Collaborative research using eScience infrastructure and high speed networks ...

USFlash

Cray's YarcData Division launches new Big Data graph appliance ...

Cray reports 2011 full year and fourth quarter financial results ...

National Institute of Genetics adopts SGI HPC solution for new supercomputer system ...

NASA scales SGI Pleiades InfiniBand Cluster to 25,000 Intel Xeon processor cores ...

SGI announces new President and CEO Jorge Luis Titinger ...

IBM Research advances device performance for quantum computing ...

ORNL completes first phase of Titan supercomputer transition ...

Researchers harness Kraken supercomputer to model explosions via transport ...

Climate scientists compute in concert ...

High-Performance Computing customers gain greater capabilities to address Big Data requirements through NetApp's collaboration with OpenSFS and EOFS communities ...

New supercomputer to accelerate Taiwan's Render Farm ...

On the path to 1 terabit-per-second networks ...

XSEDE's Extended Collaborative Support programme shares insights via symposium series ...

IBM captures leadership position in worldwide server market in fourth quarter of 2011 ...

Illinois supercomputers' expertise to help determine winner of genomics prize ...

SDSC and UC Santa Cruz to host summer school on astroinformatics ...

New supercomputer to accelerate Taiwan's Render Farm

29 Feb 2012 Taipei - Taiwan's National Center for High-Performance Computing (NCHC) independently researched and developed the Formosa series supercomputer. The latest self-built largest GPU cluster Cloud-computing system in Taiwan, Formosa 4, was ranked the 234th on the 38th TOP500 list released in the SC11 Conference in November, 2011. This cluster system, including hardware, operating system, network infrastructure, and software installation, was accomplished by NCHC's professionals. This ability not only fully demonstrates NCHC's solid strength in cluster and parallel HPC researches, but also proves that related technologies in Taiwan are in the leading position of the cluster industry already.

Formosa 4 has 88 servers, each server equipped with two six-core Xeon 5670 2.93GHz processors, InfiniBand 40Gb/s interconnection, three Nvidia Tesla M2070 GPUs, 48GB RAM, and storage nodes for a 42TB HDD capacity. To further reduce the construction cost and increase the energy efficiency, Formosa 4 adopted a CPU/GPU hybrid computing framework. Formosa 4 costs NT $37 million dollars, with an optimal performance of 70 Tflops. If it were changed to a CPU-only framework with the same computing performance, it would cost over NT$120 million.

Implementing the testing and adjustment of the overall computing efficiency of Formosa 4 was a tough task, which involved modifying cores of the operating system, altering and adjusting the network card driver, setting the network switch and network card parameters, trying numerous sets of advanced compilers, and testing several sets of computational libraries, message passing libraries, and algorithm optimisation. Since 2000, NCHC had begun developing cluster techniques, so that it could accomplish this difficult task in a short period.

The difference between the Formosa series and the "Windrider" supercomputer, ranked the 42nd on the TOP500 list released in June, 2011 is that the Formosa series is the reflection of NCHC's best parallel cluster techniques in Taiwan. It signifies NCHC's capability to handle and build Taiwan's own brand of supercomputers. Due to Formosa 4's GPU framework allowing general-purpose computation in addition to graphics processing, it provides a complementary effect on the R&D application services of the "Windrider".

Formosa series also establishes the "Render Farm" platform to provide Cloud-based animation and special effect rendering services for the local film and animation industry. The "Render Farm", officially launched on November 11, 2011, has subsequently integrated a dynamic simulation system and animation software to cultivate rendering techniques for special effects.

Formosa 4 is undergoing its final test, said Alpha Wang, manager of the planning and promotion division at the NCHC. The system, which encompasses a graphic processing unit (GPU) cluster and is located at the Tainan branch of the NCHC, is expected to come online in mid-June, she said. "Usually it takes about one year to finish the computation of some complicated animations and special effects", she stated. "Formosa 4 will be able to shorten the time and help companies introduce their works earlier."

"Some companies are using the previous generation cluster computer, Formosa 3, to compute their animations", Alpha Wang stated, without divulging the names of the companies due to confidentiality agreements.

Formosa 4 has an optimal performance of 70 teraflops (trillion floating point operations per second), nearly nine times faster than the 8 teraflops of Formosa 3, which was launched in August 2011.
Source: National Center for High-Performance Computing

Back to Table of contents

Primeur weekly 2012-03-05

The Cloud

Australian National University selects SGI to help harness explosion in data ...

Big science teams up with big business to kick‐start European Cloud computing ...

AMD to acquire SeaMicro: accelerates disruptive server strategy ...

HP launches Enterprise Security Cloud Connections partner programme ...

Dell advances public Cloud security with Trend Micro SecureCloud solutions ...

EuroFlash

Stories from the Grid launched at International Symposium on Grids and Clouds ...

IBM selected to build new supercomputer of the Max Planck Society ...

A supercomputer for research: University of Oldenburg to upgrade technical infrastructure ...

Workshop Call for Abstracts: Collaborative research using eScience infrastructure and high speed networks ...

USFlash

Cray's YarcData Division launches new Big Data graph appliance ...

Cray reports 2011 full year and fourth quarter financial results ...

National Institute of Genetics adopts SGI HPC solution for new supercomputer system ...

NASA scales SGI Pleiades InfiniBand Cluster to 25,000 Intel Xeon processor cores ...

SGI announces new President and CEO Jorge Luis Titinger ...

IBM Research advances device performance for quantum computing ...

ORNL completes first phase of Titan supercomputer transition ...

Researchers harness Kraken supercomputer to model explosions via transport ...

Climate scientists compute in concert ...

High-Performance Computing customers gain greater capabilities to address Big Data requirements through NetApp's collaboration with OpenSFS and EOFS communities ...

New supercomputer to accelerate Taiwan's Render Farm ...

On the path to 1 terabit-per-second networks ...

XSEDE's Extended Collaborative Support programme shares insights via symposium series ...

IBM captures leadership position in worldwide server market in fourth quarter of 2011 ...

Illinois supercomputers' expertise to help determine winner of genomics prize ...

SDSC and UC Santa Cruz to host summer school on astroinformatics ...