AMD has recently announced its Radeon Pro Duo desktop dual-GPU graphics card, featuring two Fiji GPUs and providing 16 TFLOPS of computing performance and, today, it has unveiled a similar graphics card for High Performance Computing (HPC) market as well, the AMD FirePro S9300 X2.

Based on two fully-enabled Fiji GPUs, clocked at 850MHz, and each packing 4GB of High Bandwidth Memory, the FirePro S9300 X2 provides a total of 13.9 TFLOPS of FP32 compute performance while keeping a 300W TDP.

Targeting HPC applications that need maximum single-precision FP32 compute performance, like deep neural networks, machine learning, geo-science, molecular dynamics, data processing and analysis and development platforms for Exascale computing, the FirePro S9300 X2 obviously has its own market.

Since its double-precision FP64 compute performance is locked at 1/16 of FP32, and since it only comes with 4GB of HBM of non-ECC memory, the new FirePro S9300 X2 is still nowhere near the Hawaii-based FirePro S9170, which offers significantly higher FP64 performance and focuses on other HPC markets.

According to AMD, the new FirePro S9300 X2 is up to two times faster than Nvidia Tesla K80 in wave equation modelling performance, making it the current king in FP32 compute performance.

AMD's new FirePro S9300 X2 will be available in the second quarter of this year with a price set at US $5999. It will be available alongside the rest of the FirePro S9100 series lineup based on Hawaii GPU and providing better double-precision compute performance.









Source: AMD.com.