AMD on Wednesday let loose its FirePro S9150 server card, supposedly the most powerful server GPU ever built for High Performance Computing (HPC) and the first to support double precision and break the 2.0 TFLOPS double precision barrier. Based on AMD's Graphics Core Next (GCN) architecture, the FirePro S9150 is specifically designed for compute workloads and is aided by 16GB of GDDR5 memory on a 512-bit memory interface for up to 320GB/s of memory bandwidth.
It has a maximum power consumption of 235 watts, and at full blast, the card is capable of 5.07 TFLOPS of peak single-precision floating point performance, which is up to 18 percent more than the competition, AMD says. Double-precision floating point performance peaks at 2.53 TFLOPS. It's made up of 2,816 stream processors (44 GCN compute units) and is ready to support OpenCL 2.0.
"Today’s supercomputers feature an increasing mix of GPUs, CPUs and co-processors to achieve great performance, and many of them are being implemented in an environmentally responsible manner to help reduce power and water consumption," said David Cummings, senior director and general manager, professional graphics, AMD. "Designed for large scale multi-GPU support and unmatched compute performance, AMD FirePro S9150 ushers in a new era of supercomputing. Its memory configuration, compute capabilities and performance per watt are unmatched in its class, and can help take supercomputers to the next level of performance and energy efficiency."
AMD also rolled out its FirePro S9050 GPU with 12GB of GDDR5 memory on a 384-bit bus for up to 264GB/s of memory bandwidth, 1,792 stream processors (28 GCN compute units), and 225W maximum power consumption.