Groq is recognized for its unique Tensor Streaming Processor (TSP) architecture that integrates into existing systems and provides the means to address big math and data challenges for accelerated science, autonomous, and industry solutions.
Recently, Frost & Sullivan performed research and analysis of the global artificial intelligence (AI) processors market for the data centers industry, and for 2022, recognizes Groq with the Technology Innovation Leadership Award for its innovation, best-in-class methodologies, experienced internal teams, and strong overall performance.
Jonathan Ross, CEO and founder of Groq, said, “The hard work of our talented teams, focused on solving machine learning’s increasingly complex needs, has driven Groq to innovate. We started out knowing software is limited by hardware, so we built a new processor architecture to advance software for machine learning workloads. We’re excited to receive Frost & Sullivan’s Technology Innovation Leadership Award as evidence of our accomplishments so far.”
Frost & Sullivan notes that the accelerating change in AI and machine learning models and the legacy need to custom code each new model iteration presents a significant challenge as the custom coding-based industry struggles to keep pace with the latest technology. Moreover, rising complexity stemming from hardware scaling with multiple processor cores, on-chip, modules, and threads present significant hurdles. With simplifying development and delivering real-time results being critical steps for vendors, Groq is well-positioned to address these complexities.
“With nearly seven times the compute density per transistor, Frost & Sullivan recognizes how Groq’s approach leverages a unique combination of easy-to-program systems that enables developer velocity and offers low latency and massive throughput in a single compute core that scales near-linearly across multiple chips,” said Sushrutha Sadashiva, Industry Analyst at Frost & Sullivan. “GroqChip™ integrates into data centers and is dedicated to handling AI workloads, where it can operate independently or as an integrated solution to help accelerate parts of a system that take advantage of the TSP architecture,” added Samantha Fisher, Best Practices Research Analyst at Frost & Sullivan.
To this end, the company delivers differentiated features (e.g., sub-millimeter low latency and predictable performance) and outperforms traditional tactics in critical areas, such as performance, power, and accuracy.
Groq customer, Argonne National Laboratory, has realized Groq-enabled performance firsthand. Argonne computational scientist Tom Brettin said, “Using the Groq platform at Argonne, we were able to accelerate our efforts to identify promising COVID-19 drug candidates from a vast number of small molecules. The system’s AI capabilities enabled us to achieve significantly more inferences a second, reducing the time needed for each search from days to minutes.”
The innovative Groq technology stack results from the startup’s talent density focus. Groq recently added to its talent and solutions portfolio by acquiring Maxeler Technologies. This combines Maxeler’s expertise in dataflow solutions with the Groq TSP architecture, helping customers build a cohesive AI strategy in a high-performance compute environment.
The latest technical information on Groq solutions will be presented by Groq Chief Architect and Fellow Dennis Abts at the 2022 International Symposium on Computer Architecture (ISCA) based on their accepted paper, “A Software-defined Tensor Streaming Multiprocessor for Large-scale Machine Learning.”