AMD today announced the launch of AMD FirePro S9300 x2, the fastest graphics card in the world for servers thanks to its 13.9 TFLOPS of computing power in single precision, something made possible by the use of Dual-Fiji GPU core offering 8192 Stream Processors @ 850 MHz. Meanwhile, on the memory side, AMD leaves the card’s HBM memory untouched, with each GPU getting 512GB/sec of memory bandwidth, for an aggregate 1TB/sec of bandwidth. Memory interface of 8192 bits which translates into a theoretical 13.8 TFLOPS of FP32 compute performance (slightly slower than -16 TFLOPs of Radeon Pro Duo) and 0.8 TFLOPs of FP64 and that’s all with a TDP of 300W that allows it cooled passively.
AMD FirePro S9300 x2, another Dual-Fiji GPU for HPC
The finished result is that the S9300 X2 is an intriguing niche item intended for a specific business sector segment that need solid FP32 execution most importantly else – and everything else held equivalent, don’t use massive data sets. It’s a to some degree restricted corner, thus, yet one AMD trusts they can do exceptionally well in given what sort of FP32 execution S9300 X2 is able to do, particularly as NVIDIA doesn’t have a FP32 HPC-focused dual-GPU Geaphics card of their own. Its suggested retail price is $ 5,999.
AMD FirePro S Series Specification Comparison
|FirePro S9300 X2||FirePro S9170||FirePro S9150||FirePro S9000|
|Stream Processors||2 x 4096||2816||2816||1792|
|Memory Clock||1Gbps HBM||5Gbps GDDR5||5Gbps GDDR5||5.5Gbps GDDR5|
|Memory Bus Width||2 x 4096-bit||512-bit||512-bit||384-bit|
|VRAM||2 x 4GB||32GB||16GB||6GB|
|FP32||13.9 TFLOPs||5.2 TFLOPs||5.1 TFLOPs||3.2 TFLOPs|
|Transistor Count||2 x 8.9B||6.2B||6.2B||4.31B|
|Target Market||HPC||HPC||HPC||HPC + VDI|
|Manufacturing Process||TSMC 28nm||TSMC 28nm||TSMC 28nm||TSMC 28nm|
|Architecture||GCN 1.2||GCN 1.1||GCN 1.1||GCN 1.0|
|Launch Date||Q2 2016||07/2015||08/2014||08/2012|