AMD FirePro™ Accelerators for High Performance Compute

The World’s First GPU Accelerator with 1TB/s Memory Bandwidth

GPUs are a compelling solution for HPC, boasting impressive densities and exceptional performance per watt with GPU-accelerated supercomputers dominating the Green500 list. Thanks to major advances in GPU technology and the rise of OpenCL™, the open standard for parallel programming, more science and engineering firms are turning to GPU servers and clusters to crunch complex datasets.

Powerful Double Precision performance for HPC - AMD FirePro™ S9150 Server GPU

Whether you’re doing scientific computing, data analytics, or seismic processing, the AMD FirePro S9150 can help accelerate all of your demanding GPU compute needs. Accelerate past our competition, with over 50* more double-precision performance than the comparable Tesla K40. DGEMM, or Double-precision General Matrix-Matrix multiplication, measures floating point execution rate for double precision, real matrix-matrix multiplication. There are many real-world applications that take advantage of double-precision matrix operations. These include computational fluid dynamics, finite element analysis and structural modelling, and molecular dynamics. With our AMD OpenCL BLAS implementation, we are able to achieve 2 TFLOPS of sustained DGEMM performance with the S9150, while the Tesla K40 achieves 1.3 TFLOPS DGEMM.

Hardware configurations: K40 - The OS is SLED 11 SP3, with driver 340.65 and CUDA 6.5 SDK, S9150 - The OS is SLED 11 SP3, AMD 15.20 driver

Recently, PathScale collaborated with Cirrascale to conduct a test on the OpenACC SPEC-ACCEL benchmark using the AMD FirePro™ S9150. The PathScale ENZO compiler generating an AMD FirePro S9150 executable is now the fastest OpenACC solution as of March, 2015, according to PathScale.