FPGAs are the Right Solution for Machine Learning and AI | Symmetry Blog

Jul 3, 2018
 

Back in March, NVIDIA CEO Jensen Huang made the claim that using FPGAs for self-driving vehicle development “is not the right answer.” He makes the point that FPGAs are not purpose-built for machine learning, and ASICs should be used to create a focused design when developing machine-learning applications.

While there are many pros and cons about using FPGAs vs ASICs, Jensen’s argument on FPGAs in machine learning is focused on the wrong topic. While his point is made on processing power, the reason why FPGA is so important to machine learning isn’t about the power, but about the way that data is processed.

 

Understanding Parallel Processing

Machine learning involves multiple different inputs that could come from different sources, and a computer must be able to process all of them in parallel. It used to be that all computing was accomplished using CPUs, but as we reach for more advanced technology and higher processing requirements, sequential processing will not be enough. With the many different signals and processes involved in machine learning, parallel processing is important to insuring a low latency and adequate power.

While ASICs also provide parallel processing, the field of machine learning undergoes rapid change. The amount of time and money to develop an ASIC solution for a specific design is impractical and often times risky for any company. The most obvious issue with an ASIC solution would be that it is inflexible, and in a field that is constantly learning, FPGAs are well positioned to handle design changes in a practical manner.

Besides learning, a major consideration for AI engineers are practical solutions for inferencing. This is where edge computing comes in.

 

Pushing Machine Learning to the Edge

When it comes to edge computing, the choice of FPGAs does matter.

Deepak Boppana of Lattice Semiconductor writes:

One way designers can quickly bring more computational resources to the network edge without re-tuning existing devices is to use the parallel processing capabilities inherent in FPGAs to accelerate neural network performance. Moreover, by employing lower density FPGAs optimized for low power operation and available in compact packages, designers can meet the stringent power and footprint limitations associated with fast-growing consumer and industrial applications.

While simple applications may be able to use any FPGA, complex applications require only the best FPGAs. That is why Lattice FPGAs are the best solution for AI development. Not only do they lead the technology of FPGA development, their FPGA development software is a cut above the rest.

To accommodate the growing field of AI and machine learning use FPGAs, Lattice has released their new SensAI stack, helping engineers create inferencing technology that works with edge applications in little time. Lattice has developed an ecosystem that gives your FPGA solutions the best performance in machine learning development.

Full-featured Lattice sensAI stack includes modular hardware platforms, neural network IP cores, software tools, reference designs, and custom design services from eco-system partners.

Flexible inferencing solutions optimized for power consumption from under 1mW-1W, package sizes starting at 5.5mm2, and priced from $1-$10 for high volume production.

Accelerates deployment of AI into a range of Edge applications including mobile, smart home, smart city, smart factory, and smart car products.

By using Lattice FPGAs, engineers get access to a complete array of innovative tools, include the new Lattice Radiant (learn more about Radiant in our interview with Lattice Semiconductor).

5 new reference designs and demos, including:

  • Face Detection
  • Key Phrase Detection
  • Face Tracking
  • Object Counting
  • Speed Sign Detection

 

Check out the Lattice Embedded Vision Development Kit offered on Symmetry’s website.

 

 


 

Symmetry Electronics Logo

 

Why partner with Symmetry Electronics? Symmetry's technical staff is specially trained by our suppliers to provide a comprehensive level of technical support. Our in-house Applications Engineers provide free design services to help customers early in the design cycle, providing solutions to save them time, money and frustration. Contact Symmetry for more information.



Subscribe

Sign up to get tech news, product updates and store promos from Symmetry Electronics.

Subscribe

Blog Archive

  2019
  2018
  2017
  2016
  2015
  2014
  2013
  2012
  2011
  2010
  2009

Previously Viewed Products