Groq is a California-based machine learning systems company that has created hardware that acts as an accelerator.The company has stated that it has taken a compiler-led design approach so that its customers can deploy its systems with ease. Its core product is the Tensor Streaming Processor.
The company's Tensor Streaming Processor (TSP) is an application-specific integrated circuit designed for machine learning. It is a single processor with hundreds of functional units, which is a novel architecture within the processor industry. The execution planning is conducted by the compiler in the processor. It has 220 megabites of static random-access memory (SRAM) and 1,000 tera operations per second. It has a single core single instruction, multiple data (SIMD) engine that is run by software and an on-die memory bandwidth of 80 terabytes per second.Groq states that the The TSP is designed to exploit parallelism inherent in machine-learning workloads, including instruction-level, memory concurrency, data and model parallelism.
TSP becomes available on the Nimbix Cloud.
Groq releases a TSP capable of 1 Peta operations per second.
Groq is founded.
A Look Inside the Groq Approach to AI Inference
October 23, 2019
Groq, A Stealthy Startup Founded By Google's TPU Team, Is Raising...
Jason D. Rowley
September 6, 2018
Documentaries, videos and podcasts
December 11, 2019