» Current | 2020 | 2019 | 2018 | Subscribe

Linley Newsletter

AImotive Accelerates Deep Learning

September 12, 2017

Author: Mike Demler

Known for its software for advanced driver-assistance systems (ADASs), AImotive has introduced licensable hardware intellectual property (IP) to accelerate such algorithms running on convolutional neural networks (CNNs). The company calls its IP AIware and positions it as a general-purpose CNN accelerator for autonomous vehicles, data analytics, robotics, speech recognition, and other tasks. Whereas most CNN accelerators offer flexibility at the expense of efficiency, designers custom-build AIware to match the parameters of a specific CNN.

The company offers an FPGA development kit to customers interested in evaluating AIware, and the synthesizable RTL is available for licensing now. It’s also working with VeriSilicon to design AIware test chips for GlobalFoundries’ 22nm FD-SOI process. By 1Q18, AImotive plans to build a few chips to validate power and performance.

In the test chip, the AIware core can function independently or pair with VeriSilicon’s VIP8000 ULQ computer-vision core. The AIware developers employed their experience in designing autonomous-vehicle systems to create accelerators for most common CNN layers, but by including the VIP8000, they enabled programmability for new and rare functions. For popular image-classification CNNs such as GoogleNet and VGG-16, for example, AIware can run 99.9% of the calculations at up to 96% utilization. The AIware test chip has a PCIe interface for connecting to a host computer, as well as a DDR4 memory interface. According to AImotive simulations, the accelerator in the test chip will deliver 2,000 billion multiply-accumulate operations per second (GMAC/s).

Subscribers can view the full article in the Microprocessor Report.

Subscribe to the Microprocessor Report and always get the full story!

Purchase the full article

Free Newsletter

Linley Newsletter
Analysis of new developments in microprocessors and other semiconductor products
Subscribe to our Newsletter »


Linley Fall Processor Conference 2020
October 20-22 and 27-29, 2020 (All Times Pacific)
Register Now!
More Events »