Energy efficiency of L-CSC supercomputer powered by AMD FirePro GPUs on way to Exascale goal

Posted by Tony DeYoung on November 25, 2014
image

Interesting comment and quote from an article on Scientific Computing regarding the AMD FirePro-powered L-CSC supercomputer that won the Green500.

Assuming that L-CSC’s energy efficiency could be scaled linearly to an exaflop supercomputing system, one that can perform one trillion floating-point operations per second, such a system would consume on the order of 190 megawatts (MW). “Although this 190-megawatt power envelope is still far from DARPA’s optimistic target of a 67-megawatt power envelope, it is approximately 16 times better than the initial projection of a nearly 3000-megawatt power envelope from 2007 when the first official Green500 list was launched,” says Wu Feng of the Green500.

Niles Burbank gave a talk at SC14 in which he describe the DARPA goal of energy efficiency.

Tags: Events, HPC

The Bourbon St Afterparty for the AMD FirePro Green500 #1 Ranking

Posted by Tony DeYoung on November 21, 2014

AMD FirePro S9150 GPUs power the L-CSC cluster at the GSI research facility that yesterday was awarded the #1 position for supercomputers on the latest Green500 List ranking of the most energy-efficient supercomputers in the world. 

It was a big achievement with the new AMD FirePro powered supercomputer 20% more energy efficient than the top ranked cluster from June. The afterparty was worth attending!

It is New Orleans and it is Bourbon St, so beads, lights, alcohol and celebration were the rule for the night.

image

image

image

Tags: Events, HPC

Announcement of GSI supercomputer w/ AMD FirePro S9150 GPU Ranking #1 in Green500

Posted by Tony DeYoung on November 20, 2014

Announcement of Top Ranked Green500 energy efficient supercomputer at SC14:  TheGSI L-CSC supercomputer based on AMD FirePro S9150 GPUs.
This supercomputer breaks the 5000 MegaFlops / Watt barrier!

Tags: Events, HPC

Top Ranked Green500 energy efficient supercomputer is L-CSC based on AMD FirePro S9150 GPUs + OpenCL

Posted by Tony DeYoung on November 20, 2014
image

Today at Supercomputing 2014, the announcement that everyone was waiting for has finally come:
The Top Ranked Green500 energy efficient supercomputer in the world is:  The L-CSC cluster, powered by AMD FirePro S9150 Server GPUs.

Each of the 160 ASUS ESC4000 G2S nodes of the new L-CSC cluster include four AMD FirePro S9150 server GPUs, for a total GPU peak of 3.25 petaFLOPS single precision and 1.62 petaFLOPS double precision performance. (Read more on the L-CSC cluster).

With 5:27 GFLOPS per watt using 600 AMD FirePro S9150s, the GSI L-CSC cluster delivers 20% better efficiency than the top ranked cluster on the June 2014 Green500 List.  The cluster is enabling Lattice Quantum Chromo Dynamics (Lattice QCD) computational research using one of the fastest OpenCL implementations for research applications in the world. Lattice QCD calculations are applicable to a range of studies in high energy and heavy ion physics.

Quote from the chairman at Frankfurt Institute for Advanced Studies: “Supercomputers are inevitable in today’s research. The scientific challenges require computers as fast as possible, but we have to keep power consumption and costs in mind, in order to use our available resources as well as possible,”  ...The ASUS ESC4000 G2S servers and the AMD FirePro S9150 GPUs are an extremely powerful basis for the L-CSC cluster and they provide the compute capabilities we need for our research. The large 16GB of memory of the AMD FirePro S9150 server GPUs allows us to run most LQCD computations on one GPU without inter-GPU and inter-node communication resulting in very efficient LQCD application operation. We have chosen a multi-GPU approach with four GPU boards per server to enable an extremely cost-efficient and energy-efficient configuration.”

image

image

image

image

Tags: GPGPU, HPC

Power efficient HPC clusters with AMD FirePro S-series GPUs

Posted by Tony DeYoung on November 19, 2014

This presentation by Niles Burbank at the AMD FirePro SC14 theater presents an overview of the FirePro S-series products for compute as well as practical recommendations on deploying GPUs in HPC environments,
Niles Burbank is Sr. Product Manager, AMD FirePro S-series Server GPUs.

Tags: Events, HPC

What going on with AMD FirePro at Supercomputing 14? OpenCL, OpenAAC, Partners, Theater

Posted by Tony DeYoung on November 18, 2014

AMD FirePro is at Supercomputing 2014, showing supercomputing performance using SPEC ACCEL  (OpenCL and OpenACC), GROMACS (molecular dynamics) , SHOC (Scalable HeterOgeneous Computing / OpenCL) benchmarks.

New HPC server configurations include:
- Supermicro 1047GR-TRF server, 1U 2P with 3 S9150s
- Gigabyte G210-H4G blade, 2U enclosure, 4 blades, 1P with 4 S9150s
- Exxact FXR410-512R workstation (secured) 2P with 4 W9100s
- ASUS ESC8000 G3 3U rack mount with 8 S9150s

Check out the video overview and pics below.

Theater
image

Booth
image

AMD FirePro S9150
image

Supermicro 1047GR-TRF server with 3 S9150s, ASUS ESC8000 G3 3U rack mount with 8 S9150s
image

Exxact FXR410-512R workstation (secured) 2P with 4 W9100s
image

Gigabyte G210-H4G blade, 2U enclosure, 4 blades, 1P with 4 S9150s
image

Apache Hadoop on AMD ARM processors
image

Pre-release of SYCL for OpenCL
image

Also check out the theater schedule:

TUESDAY, NOVEMBER 18TH

11:15 AM OpenACC on AMD GPUs and APUs - Learn how to use OpenACC on your AMD APUs and Radeon GPUs using the PGI compilers
Michael Wolfe, Compiler Engineer - PGI

1:00p ASUS Server Launch with AMD FirePro S-series GPUs

1:15p The L-CSC cluster - An AMD-GPU-based cost- and power-efficient multi-GPU system for Lattice-QCD calculations at GSI
Dr. David Rohr, Frankfurt Institute for Advanced Studies

2:15 PM Distributed Symmetric Multiprocessing - Shattering Affordability and Scalability Barriers to Large Shared Memory, Many Core HPC
Richard Anderson, CEO - Symmetric Computing

3:15 PM Power efficient HPC clusters with AMD FirePro S-series GPUs - An overview of AMD FirePro S-series products for compute and practical recommendations on deploying GPUs in HPC environments
Niles Burbank, Sr. Product Manager, AMD FirePro S-series Server GPUs

4:15 PM Raising the Bar on HPC Innovation with PBS Works
Scott Suchyta, Director of Partner Solutions and Integrations, Altair

5:15 PM Sandia National Laboratories’ Experiences with AMD APUs and HSA Technology
Si Hammond, Scalable Computer Architectures, Sandia National Laboratories


WEDNESDAY, NOVEMBER 19TH

11:15 AM HSA for Application Developers - Learn the intricacies of using HSA and the benefits seen in some HPC workloads
Jonathan Gallmeier, AMD Research Software Engineer

2:15 PM Using the SYCL for OpenCL open standard to accelerate C++ code - SYCL for OpenCL is an upcoming standard C++ programming model for OpenCL from the Khronos group. Ideally suited to C++ metaprogramming, it allows a single C++ source file to be compiled for both host and OpenCL devices. This talk will introduce the provisional specification and enable developers to provide feedback for the final specification.
Andrew Richards, CEO Codeplay

3:15 PM Porting Gromacs from CUDA to OpenCL. Lessons learned
Vincent Hindriksen, CEO, Stream Computing

4:15 PM Kernel Compilation and execution for HSA
Greg Rodgers, Design Engineer, AMD Research

5:15 PM The Value of Supercomputing Field Data - Dr. Nathan DeBardeleben of LANL will discuss supercomputer reliability studies, in collaboration with AMD RAS and Research teams, and the impact they will have on next generation system design and ExaScale planning for the Department of Energy.
Nathan DeBardeleben, Ph.D., Head of Ultrascale Systems Research Center (USRC), HPC Resilience Lead, Los Alamos National Laboratory (LANL)

THURSDAY, NOVEMBER 20TH

11:15 PM Distributed Symmetric Multiprocessing - Shattering Affordability and Scalability Barriers to Large Shared Memory, Many Core HPC
Richard Anderson, CEO - Symmetric Computing

1:15 PM Penguin Computing’s APU / HSA Development Cluster
Phil Pokorny, CTO - Penguin Computing

Tags: Events, HPC

Size Matters: Times Square digital billboard powered by 3 AMD FirePro GPUs spans a city block

Posted by Tony DeYoung on November 17, 2014

The new Times Square digital billboard stands 8 stories tall and spans a full city block (100 meters),  wrapping around the front of the Marriott Marquis Hotel on Broadway between 45th and 46th streets. The seamless digital canvas is visible from virtually every vantage point in the Square’s ‘bowtie’ where more than 300,000 pedestrians pass by daily.  Imagine it on New Years Eve!

Great quote : “People go to the Grand Canyon to see the most visually stunning natural canyon in the world. They come to Times Square to see the most digitally striking canyon in the world” said Tim Topkins, president of the Times Square Alliance.

The 24 million pixel display is driven by 3 AMD FIrePro W9000 GPUs and will light up on Tuesday night (Nov 18).

image

How Sugar TV made hi-end TV commercial using Mac Pro w/ dual AMD FirePro cards + NUKE + Resolve

Posted by Tony DeYoung on November 17, 2014

Digital Arts has a very interesting and comprehensive case study about how Chris Smith of Sugar TV used a single Mac Pro, with dual AMD Fire-Pro GPU, NUKE STUDIO, DaVinic Resolve for colour correction and editing, Final Cut Pro X for offline editing, Logic Pro X for sound editing/mixing and MODO for 3D content creation to creae a commercial for HEB Grocery Stores “Signs” campaign.

What makes the article especially compelling is the pagination which ties a specific image to components of the narrative.  This is a case study worth reading for both content and presentation!

On great quote: “The thought of doing it the old way of jumping out to a separate 3D app and render passes back in suddenly started to seem cave-man like!”

Video of the final commercial below.

ASUS ESC8000 G3 server can hold up to 8 AMD FirePro 9150 S-Series GPUs

Posted by Tony DeYoung on November 15, 2014

The new ASUS ESC8000 G3 is a 3U rack mount server solution with up to 8 AMD FirePro S9150 server GPUs.
Download the AMD FirePro and ASUS ESC8000 G3 data sheet for full specs.
Learn more about HPC in the Data Center accelerated by OpenCL on AMD FirePro S-Series Server GPUs.

image

image

The bank branch of the future - NY’s SEFCU uses Ventuz + AMD FirePro W9100 to make it real

Posted by Tony DeYoung on November 13, 2014

New York’s SEFCU wanted to put reality into the bank of the future. They commissioned stereolize to design a 4k multi-touch video wall and 4k multi-touch top to let customers explore and interact with the banks products and services

The interactive wall and table were driven by Ventuz software and a single AMD Firepro W9100 GPU.

image

image

image

image

More info about the project at Ventuz Used for Bank Showroom in New York.

Page 1 of 93 pages  1 2 3 >  Last »

Close