When it comes to AMD FirePro™ S-Series GPUs, there is never “too much of a good thing.” During my conversations at trade shows and conventions, customers have been asking for a server GPU solution that not only provides the speed to accomplish computational tasks, but they also want accuracy for precise results. Earlier this year, we announced the top-of-line AMD FirePro™ S9150 – the most powerful server GPU ever built for High Performance Computing1 measured by performance/watt. But customers also told us they want a product for mainstream, budget conscious projects that still delivers equally exceptional performance for HPC.
To meet that demand, we’ve expanded our server GPU portfolio. By infusing our most current iteration of the AMD GCN architecture2 into AMD FirePro™ server GPUs, we have been able to offer two products that deliver leading edge double-precision floating-point rates, and operate with tremendous efficiency.
The new AMD FirePro™ S9100 server GPU is supreme in its class, with high peak single- and double-precision floating point performance, including 2.11 TFLOPS peak double-precision and up to 9.4 GFLOPS per watt TDP double-precision performance3. Compute-intensive workloads are where this card excels. With 2,560 stream processors and 12GB of GDDR5 memory, the AMD FirePro S9100 is equipped to handle supercomputing tasks. For those working in scientific compute, research, or structural analysis, it would be fair to say this GPU is a compute powerhouse. On top of the performance, the card’s maximum TDP comes in at 225W, and its passive cooling solution makes it ideal for deployment in server environments.
We’re seeing a big part of the demand for GPUs driven by widespread adoption of OpenCL™ support. Professionals have told me they want to use open standards. They want the freedom and accessibility that comes with OpenCL™, and software developers are heeding this ask. Across our product offerings, AMD has long been an advocate for open standards and this is no different with our server GPUs. Leveraging the latest version of OpenCL™, you’ll experience great compute performance, accelerating complex calculations at lightning speed.
Another complementary method for handling workloads is OpenMP -- an API supporting high-level parallelism in the C, C++, and Fortran languages. In segments such as Oil and Gas, Computer Aided Engineering and Computational Science, many organizations have made substantial investments in OpenMP to create scalable workloads. AMD’s collaboration with Pathscale to provide support for OpenMP 4.0 will allow customers in these HPC fields to harness the compute power of AMD FirePro S9100 server GPU.
AMD is also working together with an ecosystem of hardware providers to help ensure customers have access to great technology for their projects. To foster that ecosystem, AMD FirePro S9100 server GPUs will be available from ASUS, Gigabyte, Supermicro and TYAN.
Designed to accelerate applications beyond just graphics, AMD FirePro S9100 joins the lineup of AMD FirePro S-series cards for servers that are powered by AMD STREAM4 technology to support compute intensive workloads by offering technologies such as:
ECC memory support (on GDDR5 memory only)
Fast single- and double-precision floating-point performance
Peer-to-peer multi-GPU support
Bi-directional PCIe® memory data transfers
GPU-optimized OpenCL™ libraries
We’ve listened to our customers and we understand their needs by delivering two products: one with the best possible performance at the high end, and another option for more budget conscious deployments that need great performance in the data center.
Learn more about AMD FirePro S-series server graphics cards
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.PCIe is a registered trademark of PCI-SIG Corporation.
1AMD FirePro™ S9150 max power is 235W and delivers up to 2.53 TFLOPS peak double and up to 5.07 peak single precision floating point performance. Nvidia’s highest performing server cards in the market as of June 2014 are the Tesla K40, max power of 235W, with up to 1.43 TFLOPS peak double and up to 4.29 peak single, and the K10, max power 225W, with up to 4.58 TFLOPS peak single and 190 GFLOPS peak double precision. Visit http://www.nvidia.com/object/tesla-servers.html for Nvidia product specs. FP-97
2 Select AMD FirePro graphic cards are based on the GCN Architecture and include its associated features (AMD PowerTune technology, AMD ZeroCore Power technology, PCI Express 3.0, etc.). Not all features are supported by all products—check with your system manufacturer for specific model capabilities
3 AMD FirePro™ S9100 max power is 225W and delivers up to 2.11 TFLOPS peak double and up to 4.22 TFLOPS peak single precision floating point performance. Nvidia Tesla K20X max power is 235W and delivers up to 1.31 TFLOPS peak double and up to 3.95 TFLOPS peak single. Visit _http://www.nvidia.com/object/tesla-servers.html for Nvidia product specs. FP-109