» Current | 2020 | 2019 | 2018 | Subscribe

Linley Newsletter

Blaize Ignites Edge-AI Performance

September 8, 2020

Author: Mike Demler

Ten-year-old edge-AI startup Blaize is sampling board-level systems that integrate its El Cano inference processor. The chip integrates 16 of the company’s graph-streaming processors (GSPs), delivering peak throughput of 16 trillion operations per second (TOPS). El Cano uses dual Cortex-A53 CPUs to run an operating system and application software, but the programmable GSPs handle all neural-network operations independently, as well as image signal processing, sensor fusion, and other tasks. The first products target commercial and enterprise computer-vision systems, such as retail analytics and antitheft systems. El Cano is also well suited to automotive and industrial applications, including camera/lidar sensor fusion and robotics.

The new processor, named for the often overlooked Spanish explorer who completed the first circumnavigation of the Earth after Magellan died at sea, improves hardware utilization by combining data, instruction, and task parallelism. It can dynamically manage multiple workloads according to their bandwidth and compute requirements, in addition to supporting conditional processing. The chip’s hardware scheduler runs multiple neural networks in parallel or sequentially, or it can stagger execution to optimize use of available resources. In one of Blaize’s demos, El Cano ran five independent Yolo v3 networks on five separate HD-video streams, delivering inference results on 416x416 subframes at 10fps. It can also run a single copy at 50fps. The startup plans to begin volume production next quarter. Samsung manufactures the chip in 14nm technology.

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 »

Events

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