A new ultra-low latency fintech accelerator targeting the electronic trading market where nanoseconds matter has been launched. The solution, which can accelerate customized trading algorithms with flexible FPGAs, is already being actively adopted by global financial services companies.

AMD today announced the availability of the AMD Alveo UL3524 accelerator card, a new fintech accelerator designed for ultra-low latency electronic trading applications.

The Alveo UL3524 is already deployed by leading trading firms and supported by multiple solution partner offerings, providing proprietary traders, investment banking institutions, hedge funds, brokerages and exchanges with a state-of-the-art FPGA platform for electronic trading at nanosecond speeds.

AMD emphasized that it can accelerate trading execution speeds by achieving FPGA transceiver latency of less than 3ns, which is a 7x improvement in latency compared to previous-generation FPGA technology, as tested by AMD Performance Labs.

“In ultra-low latency trading, a single nanosecond can make the difference between winning or losing a trade,” said Hamid Salehi, AMD’s director of product marketing.“The Alveo UL3524 accelerator card is specifically developed to provide a competitive edge to fintech customers in the financial markets,” he said.

■ Hardware flexibility and AI-based trading strategies

The Alveo UL3254 features 64 ultra-low latency transceivers, an FPGA fabric of 780,000 LUTs, and 1,680 DSP slices. The flexible hardware allows traders to adapt their designs based on evolving strategies and market conditions, accelerating custom trading algorithms.

Supported by the existing FPGA design flow based on the Vivado Design Suite, the product provides global support from AMD domain experts, a set of reference designs and performance benchmarks to enable FPGA designers to develop customized trading strategies according to their specifications.

AMD is providing developers with an open-source, community-supported FINN development framework to streamline the adoption of AI in algorithmic trading markets.

Leveraging PyTorch and neural network quantization techniques, the FINN project allows developers to reduce the size of AI models while maintaining accuracy, compile to hardware IP, and integrate network models into the algorithm data path for low-latency performance. It is an open-source framework that allows developers to flexibly take advantage of the latest improvements.

■ Expansion of ultra-low latency fintech solution ecosystem

Alveo UL3524 and FPGA technology developed specifically for fintech enable strategic partners to implement customized solutions for the fintech market. Currently, partner solutions are available from Alpha Data, Exegy, and Hypertec.

“AMD’s new Virtex UltraScale+ FPGAs represent a game-changing advancement in ultra-low latency trading and networking,” said David Miller, General Manager at Alpha Data. “To enable our customers to easily harness the full potential of these new AMD FPGA devices, Alpha Data has developed the ADA-R9100 rack-mount appliance.”

Exergy, a provider of front-office trading solutions, is supporting the Alveo UL3524 card with nxFramework, a custom software and hardware development environment for creating and maintaining ultra-low latency FPGA applications for the financial industry.

“Combining AMD’s leading ultra-low latency FPGA technology with Exergy’s expertise in capital markets allows us to provide a comprehensive solution to address the ever-growing optimization needs and power the trading infrastructure of the future,” said Olivier Cousin, Director of FPGA Solutions at Exergy.

The AMD Alveo UL3524 accelerator card is reportedly currently in production and shipping to global financial services customers.