Myrtle.ai's VOLLO Sets New Standards for Gradient-Boosted Tree Inference Performance
Myrtle.ai has made waves in the AI and financial sectors with its VOLLO inference accelerator, recently announcing record-breaking achievements in gradient-boosted tree inference performance. This news was unveiled during the STAC Summit held in London, showcasing VOLLO's impressive stats and capabilities.
The VOLLO accelerator has successfully set new standards on the STAC-ML Markets (Inference) benchmarks, revealing a more than 30% reduction in the 99th-percentile latency. Additionally, it demonstrated a staggering throughput increase of at least 5 times over previous high-performance results. With a p99 latency below 2 microseconds across three different models, VOLLO reaches unprecedented operational capacities, sustaining up to 50 million inferences per second with just 1.77 microseconds of latency on its smallest model.
In the high-stakes world of electronic trading, the speed at which market data is processed directly impacts profit margins. Deterministic low latency allows trading firms to utilize more complex and accurate models without sacrificing speed. This capability means that firms can harness the full potential of their data-driven models without having to compromise on performance. As stated by Peter Baldwin, CEO of Myrtle.ai, these results enable trading firms to experiment with advanced modeling techniques while maintaining rapid execution speeds.
Myrtle.ai's VOLLO doesn’t just set records; it is already proving its reliability in practical applications. The technology has been in operation for hundreds of thousands of hours in live trading, successfully generating alpha for numerous leading trading entities globally. Its versatility extends beyond finance, finding a growing preference among telecom companies, network security professionals, and defense sectors due to its adaptation capabilities.
The STAC-ML Markets (Inference) benchmark is recognized as the gold standard for evaluating the performance of technologies that handle inference on real-time market data. Designed by quant analysts and tech professionals from the top financial institutions, the benchmark rigorously assesses the speed, resource efficiency, and accuracy of various technology platforms that utilize the models provided.
Developers interested in how their models would perform on VOLLO can now evaluate without needing extensive FPGA expertise or tools. Myrtle.ai has made it easy to access and test this leading-edge technology, inviting machine learning developers to explore its capabilities via its dedicated website. Once evaluated, the results can spark insights into how to further optimize models for maximum performance on their platforms.
Myrtle.ai continues to push the boundaries of what's possible with AI and machine learning, creating valuable opportunities for sectors that depend on speed and precision. As industries evolve and demand increases for faster, more reliable technologies, its VOLLO inference accelerator stands out as a transformative solution in the ever-competitive tech landscape. With the combination of ultra-low latency capabilities and reliable results, Myrtle.ai solidifies its position as a leader in AI inference technology, setting a new benchmark that others will strive to meet.
For those curious about delving deeper into VOLLO's capabilities, further information is readily available through Myrtle.ai’s dedicated web platform, inviting a new wave of developers and firms to stay at the forefront of technological advancements in AI-driven sectors.