Myrtle.ai Reduces Latency in Financial ML with VOLLO

Myrtle.ai has announced significant improvements in machine learning inference latency with its VOLLO product. During the latest STAC Summit in London, an audit by the leading benchmark authority, STAC, showed that VOLLO, featuring an FPGA-based solution, achieved latencies as low as 2 microseconds in the 99th percentile. This result represents a halving of the previous latency benchmark.

Designed by top financial firms, the STAC-ML Markets benchmark highlighted VOLLO's efficiency, particularly in throughput, enabling users to process complex trading models faster. The system uses Silicom's PCIe accelerator card, fitted with AMD's Versal Premium series SoC, marking a step forward in low-latency inference for the financial sector.

Myrtle.ai’s collaboration with AMD, Silicom, and Supermicro underscores its commitment to ultra-low latency AI inference in trading environments. The improvements in latency enable faster and more intelligent decisions in trading and risk analysis, offering a competitive edge in financial markets.