Checking Out a Neuromorphic AI Module
What you’ll learn:
- What are neuromorphic computing and spiking neural networks?
- Details about BrainChip’s Akida M.2 platforms.
Neuromorphic computing is designed to mimic the operation of the human brain using a spiking-neural-network (SSN) approach. This is different from the convolutional neural networks (CNNs) that use lots of matrix manipulation to implement multiple layers in an artificial-intelligence (AI) model.
BrainChip has implemented SSN support in hardware with its AKD1000 and AKD1500 family of embedded AI accelerators (Fig. 1). In the video above, I give a hands-on look at the AKD1000 M.2 module that’s useful for developing applications using SSN models.
The M.2 module is targeted at evaluation tasks, although it could be used in some embedded applications. For example, many applications are using Raspberry Pi’s (Fig. 2) as the host. The module can be added using a PCI Express HAT that typically holds a pair of M.2 modules.
The AKD1000 offers 1.5 TOPS of performance, but the AKD1500 is actually more powerful, delivering even at 0.8 TOPS thanks to better efficiency. It also consumes less than 300 mW of power. The AKD1500 M.2 module is finally available, too. However, I tested the AKD1000 that was sent earlier by BrainChip.
The chips support PCIe and SPI interfaces. Which one you use depends on the host. For the modules, it’s PCIe.
BrainChip’s hardware is supported by Edge Impulse (Fig. 3). This is a commercial, web-based AI development platform that supports an array of AI targets, including the AKD1000/1500. It can handle training and optimization. A subscription is required to use higher-end features, such as faster training via a GPU.
The company’s software is command-line driven and based around Python, which tends to be the de facto standard for AI tools these days. It takes a bit of time to set up, but it’s easy to use once configured. Deployment simply requires the runtime drivers and the models.
Spiking neural networks are typically more efficient, and the low-power operation of the AKD1500 can allow always-on operation for many applications. The software will handle most models developed for other hardware platforms with just a recompile and optimization. The platform is definitely worth considering for evaluation — it’s just an M.2 socket away.
About the Author
William G. WongWilliam G. Wong
Senior Content Director - Electronic Design and Microwaves & RF
I am Editor of Electronic Design focusing on embedded, software, and systems. As Senior Content Director, I also manage Microwaves & RF and I work with a great team of editors to provide engineers, programmers, developers and technical managers with interesting and useful articles and videos on a regular basis. Check out our free newsletters to see the latest content.
You can send press releases for new products for possible coverage on the website. I am also interested in receiving contributed articles for publishing on our website. Use our template and send to me along with a signed release form.
Check out my blog, AltEmbedded on Electronic Design, as well as his latest articles on this site that are listed below.
You can visit my social media via these links:
- AltEmbedded on Electronic Design
- Bill Wong on Facebook
- @AltEmbedded on Twitter
- Bill Wong on LinkedIn
I earned a Bachelor of Electrical Engineering at the Georgia Institute of Technology and a Masters in Computer Science from Rutgers University. I still do a bit of programming using everything from C and C++ to Rust and Ada/SPARK. I do a bit of PHP programming for Drupal websites. I have posted a few Drupal modules.
I still get a hand on software and electronic hardware. Some of this can be found on our Kit Close-Up video series. You can also see me on many of our TechXchange Talk videos. I am interested in a range of projects from robotics to artificial intelligence.