Terasic's DE5a-Net-DDR4 has been purpose-built to accelerate certain workloads that involve AI, high-frequency trading, and high-performance computing. By doing this, the DE5a-Net-DDR4 accelerator helps meet the demands of compute-intensive applications and provides an industry-leading platform for industries that rely on HPC to maintain a competitive edge.
The DE5a-Net-DDR4 Development Kit provides an ideal hardware solution for any and all designs that require high capacity and bandwidth memory interfacing, as well as the lowest latency communication—all while still being power efficient.
Designed for Demanding High-End Applications
Using Intel's powerful Arria 10 FPGA (field-programmable interface), this accelerator enables data processing at higher speeds and includes 16GB of DDR4 memory that runs at over 75 GB/s. This delivers up to 7.876GB/s of data transfer via PCIe Gen 3 x8 edge between FPGA and the host PC.
The accelerator also has 4 onboard QSFP+ (quadi—i.e. 4-channel—small form-factor pluggable) connectors, plus, the onboard QDRII (aka quad-data rate 2) memory and DDR4 (aka Double Data Rate 4) module can achieve lower application latency and higher throughput.
The DE5a-Net-DDR4 (Double Data Memory 4) board. Image Credit: Intel.
Additionally, the DE5a-Net-DDR4 supports Intel’s Open VINO toolkit for optimal computer vision and deep learning solutions. End users can achieve the best computing performance at the lowest cost for their AI applications, all by using the Arria 10 FPGA on DE5a-Net-DDR4.
Additional features include:
Intel Arria 10 GX FPGA
Ultra-low latency and straight connections to 4 external 40 G QSFP+ modules
High-speed parallel flash memory
2 DDR4 SO-DIMM socket banks, up to 16GB 1,066 MHz, or 8GB 1,200 MHz
As mentioned, the DE5a-Net-DDR4 accelerator is ideal for a large range of compute-intensive applications that are required to maintain a competitive edge. Some of these applications include Fintech, video and broadcast, AI and deep learning, data centre acceleration, data filtering, and algorithmic acceleration.