PCIe Board with Intel (Altera) Arria 10 GX FPGA Machine Learning Demonstration

BittWare took part in Super Computing 2016 last week, and used the event to showcase their must-see FPGA demos in network packet processing (NPP) and high-performance computing (HPC). If you were at the show you will also have seen BittWare's wide range of the latest Intel (formerly Altera) and Xilinx FPGA-based boards on display.



One of the demonstrations featured BittWare’s A10PL4 board, using Intel’s deep learning IP core loaded on the Arria 10 FPGA.

The A10PL4 from BittWare is an Altera Arria 10 FPGA PCIe board offering plenty of high speed serial I/O interfaces and debug support. Two QSFP cages are available, both connected to the Arria 10 FPGA removing latency in the external PHYs. The A10PL4 supports up to 32GB of DDR4 as well as flash memory with factory default and support for multiple FPGA images.

Find out more - A10PL4

 

Machine Learning

'Machine Learning' is based on the development of computer programs that can learn and create their own rules. In deep learning, a task can be learned by the machine from a vast amount of data, either in supervised or unsupervised manner.

With the Deep Learning IP customers can leverage Intel Caffe in the Data Analytics Acceleration Library (DAAL) and Intel MKL-DNN to build AlexNet- or GoogleNet-like topologies without requiring a recompilation of the FPGA.

Accelerating machine learning applications like these are an excellent fit for FPGAs and OpenCL. Up to 16 A10PL4s can be combined inside a 4U server for a total of 18M logic elements and 512 GBytes of DDR4 memory.

Find out more - Terabox

For more information please speak to one of our FPGA specialists - +44 1672 511166 or send us an email - info@sarsen.net 

www.sarsen.net