Acceleration for Financial Technology

By Shivangi Kharoo


Adoption of AI in trading is picking pace.

AMD has introduced the AMD Alveo UL3524 accelerator card for electronic trading. This card is designed for customizing trading algorithms with its 64 low-latency transceivers, 780K LUTs of FPGA fabric, and 1,680 DSP slices of compute.

It offers significantly reduced latency compared to previous FPGA technology, with less than 3 nanoseconds of FPGA transceiver latency. It supports traditional FPGA workflows using Vivado Design Suite, providing reference designs and performance benchmarks for FPGA designers to develop custom trading strategies.

“UL3524 accelerator card is powered by the lowest latency FPGA transceiver from AMD purpose-built to give our fintech customers an unprecedented competitive advantage in financial markets.” said Hamid Salehi, director of product marketing at AMD.

To aid algorithmic trading, AMD provides the open-source FINN development framework, allowing developers to reduce AI model size while retaining accuracy for low latency. This initiative offers flexibility and accessibility to developers as projects evolve.

AMD’s Alveo UL3524 collaborates with partners like Alpha Data, Exegy, and Hypertec to create solutions for the fintech market, including rack-mount appliances and software/hardware development environments for ultra-low latency applications in finance.