Achieving unmatched efficiency and performance
Menu
Volga has developed a truly unique AI compute platform that enables effective surveillance systems, smart appliances, powerful machine vision systems, commercial drones, brilliant robotics, and more. With unmatched performance and power efficiency, the Volga platform enables AI designers to deploy in form factors that were previously out of reach.
Key features include:
Always running with batch size of 1
Best-in-class TOPS/W
Single-chip to high-performance PCIe cards
Supports standard platforms, frameworks, and DNN’s
The Volga Analog Matrix Processor (Volga AMP™) is based on a unique tile-based AI compute architecture that features three fundamental hardware technologies – Compute-in-Memory, Dataflow Architecture, and Analog Computing.
For AI developers, Volga delivers the hardware, software tool kit, and trained neural networks to ease deployment in edge devices.
The Volga AMP features an array of tiles that blend familiar concepts and new breakthrough technology to deliver unmatched performance, power, and flexibility. Each tile has a large Analog Compute Engine (Volga ACE™) to store bulky neural network weights, local SRAM memory for data being passed between the neural network nodes, a single-instruction multiple-data (SIMD) unit for processing operations not handled by the ACE, and a nano-processor for controlling the sequencing and operation of the tile. The tiles are interconnected with an efficient on-chip router network, which facilitates the dataflow from one tile to the next. On the edge of the Volga AMP™, off-chip connections provide an interface to the host system.
Volga software provides the necessary tools to bridge the gap between popular training frameworks like PyTorch and highly constrained edge and server deployments on the Volga AMP™ that must optimize for cost, power, and performance. The Volga Optimization Suite solves a significant customer pain point by converting the neural network to an 8-bit representation, while preserving accuracy in the analog compute domain. The Volga Graph Compiler automatically generates machine code for our AMP, so developers do not have to worry about low-level implementation or optimization. Volga host drivers are pain-free and lightweight, with support planned for most popular embedded and server operating systems.
Maximizing inference performance through careful architecture design
Our team ushers in a new era of AI innovation