AMD Leads The Way with HPC at International Supercomputing Conference 2014

Blog Post created by brystal.boyd Employee on May 12, 2015

Several advances in the software ecosystem – particularly those associated with OpenCL™ -- inspired AMD to showcase some great innovations at International Supercomputing Conference (Leipzig, Germany, June 22-26).


The conference attracts industry leaders, a variety of exhibitors, and members of academia from around the world and this year it explored the future direction of HPC technologies, life science applications, big data, quantum computing, and the real-world value of HPC.


As you might know, OpenCL™ has made GPUs increasingly attractive for running high-performance computing (HPC) workloads in a scalable and efficient manner. At the show, I had the opportunity to share details on AMD’s new workstation and server GPU solution.

photo 3.JPG.jpg


One of the highlights was the announcement of AMD’s new AMD FirePro™ W8100 workstation graphics cards powered by OpenCL™ for personal supercomputing.  Each card is capable of delivering more than 4.0 TFLOPS of peak GPU compute performance for future-ready 4K multi-display workflows, and deliver up to nearly 38X more floating point compute performance than the most comparable competing product(1).


Another highlight was a sneak peek of the new AMD FirePro™ S9150 – our newest server GPU for HPC. This forthcoming product offers massive parallel processing power that combines with OpenCL™ to accelerate applications beyond just graphics, and compute intensive workloads are expected to benefit from technologies and features such as:
- More than 2.0 TFLOPS peak double precision performance
- Half rate double precision floating point
- High performance per watt
- 16GB ultrafast GDDR5 memory
- AMD STREAM technology(2)
- Mechanical and thermal design optimized for standard servers


We treated attendees to a demonstration of the AMD FirePro™ S9150 running a prototype version of the open source numerical computation package Scilab. This side-by-side dataset comparison took a look at Scilab running on a CPU only, alongside Scilab with the sciGPGPU module providing support for OpenCL™. The demonstration – powered by four AMD FirePro™ S9150 server GPUs in a Supermicro SuperServer chassis – showed how for the first time, users will have access to GPU acceleration for mathematical modeling enabled by the industry standard OpenCL™ API. The Scilab application provides a simple interface to access the power of GPU compute, even for users who aren’t familiar with the intricacies of programming for GPU acceleration.


photo 1b.JPG.jpg


AMD FirePro™ S9150 GPUs support AMD STREAM technology(2), which was first announced at Computex 2014.  AMD STREAM technology powers the ecosystem that enables AMD FirePro™ graphics cards – including the new AMD FirePro™ S9150 – to be used for compute-intensive workflows leveraging the massively parallel processing power of AMD GPUs.


This technology can be used to accelerate many applications beyond just graphics, including critical research and operations for a variety of scientific and mathematic purposes across many industries and disciplines. For example, AMD has supported the University of Frankfurt SANAM Supercomputer with AMD FirePro™ S10000 server GPUs, with the same underlying technology as the AMD FirePro S9150 to enable groundbreaking power and efficiency for specific research projects.


AMD STREAM technology(2) includes:
- ECC Memory Support
- Fast Single & Double Precision Performance
- Peer-to-Peer Multi-GPU Support
- Bi-directional PCIe® Memory Transfers
- GPU Optimized OpenCL™ Libraries


The goal for AMD STREAM technology enabled on AMD FirePro S-Series server GPUs is to deliver immense compute capability, performance and flexibility to handle the dense, multithreaded workloads and extremely large, complex datasets associated with supercomputing and HPC.


If you aren’t doing so already, try working with OpenCL and assess the performance of AMD FirePro™ graphics for your particular compute workloads.


Learn about AMD FirePro graphics and University of Frankfurt SANAM Supercomputer

Niles Burbank is a senior manager, server GPU products, at AMD. His postings are his own opinions and may not represent AMD’s positions, strategies or opinions. Links to third party sites are provided for convenience and unless explicitly stated, AMD is not responsible for the contents of such linked sites and no endorsement is implied.

OpenCL and the OpenCL logo are trademarks of Apple Inc. used by permission by Khronos. Apple and Mac Pro are trademarks of Apple Inc., registered in the U.S. and other countries.


1 SiSoftware Sandra test details:
System Description: AMD FirePro W8100 vs. Nvidia Quadro K5000 - Dell T3610, Intel Xeon E5-1620 v2 @ 3.60 GHz, 8GB DDR3, Seagate HDD 7200RPM, Win7 64-bit SP1, 1920x1080 resolution
AMD Driver 13.352.1009 | Nvidia Driver 333.11


2 AMD STREAM technology is a set of features offered with select AMD FirePro graphics cards for the acceleration of compute-intensive workflows.  Not all products have all features and full enablement of some capabilities may require complementary software.  Check with your system manufacturer for specific capabilities and supported technologies.

*Originally Posted by System Admin in AMD Business on Jul 9, 2014 11:02:59 AM