This professional graphics card designed for servers offers exceptional performance in FP32 and FP16, massive memory bandwidth, and unbeatable energy efficiency. It achieves a computing performance of 13.9 TFLOPS, twice the performance of the most powerful computing cards ever released. It boasts the world's first memory bandwidth of 1TB/sec to speed up even the most challenging tasks. In addition, its energy efficiency is 46 GFLOPS/W.
The graphics card is equipped with a pair of Fiji cores with 4096 stream processors and a new type of HBM memory with low power consumption but wide bandwidth. Overall, the device has a capacity of 8GB with a throughput of two times 512 GB/sec. Both cores are clocked at 850MHz and their TDP does not exceed 300W. The card features a passive dual-slot cooler and uses the airflow coming through the server rack (reverse airflow). This means that the graphics card is designed exclusively for servers to perform complex calculations. It also lacks any image outputs.
The most complex calculations can be utilised in a variety of fields, such as seismic data processing, which is mainly used to search for oil and gas sites, or for neural networks or data analyses. The card can be run on a Linux-based AMD GPUOpen software package that is optimised for computing and supports both C++ and OpenCL 1.2.
System Requirements:
PCI Express 3.0 ×16
2× PCIe 8pin power
Air flow through the graphics card of at least 25CFM with a maximum temperature of 45°C
System memory of at least 16GB of DDR3/DDR4
This product is not an independently functional unit and may require professional installation.
Specifications can be changed without notice. Images are for illustrative purposes only.