BrainChip, SiFive partner to bring AI and ML to edge computing
We’re excited to bring Transform 2022 back in person on July 19 and virtually July 20 – August 3. Join AI and data leaders for insightful conversations and exciting networking opportunities. Learn more about Transform 2022
AI processor maker brain chipthat makes ultra-low-power neuromorphic chips and supporting software, and SiFivefounder of the RISC-V computing genre, today announced that they have combined their respective technologies to provide chip designers with optimized artificial intelligence (AI) and machine learning (ML) for edge computing
BrainChip’s AI engine, Akida, is an advanced neural network processor architecture that brings AI functionality to edge and cloud computing in a way not previously possible with its high performance and ultra-low power consumption, the company said. SiFive Intelligence solutions, with their highly configurable multi-core, multi-cluster capable design, integrate software and hardware to accelerate AI/ML applications, BrainChip CMO Jerome Nadel told VentureBeat.
The integration of BrainChip’s Akida technology and SiFive’s multi-core capable RISC-V processors is expected to provide an efficient solution for integrated edge AI computing, Nadel said.
RISC-V (pronounced “risk-five”) is an open instruction set computing architecture based on established RISC (Reduced Instructions Set Computing) principles. It is an open source project that is available to anyone who wants to use it. RISC-V represents a major step forward in data processing — speed required by all the new and much “heavier” applications (such as machine learning, AI, and high-resolution video) used on a daily basis. RISC-V appears to be a natural fit for BrainChip’s architecture for neural network processors.
RISC-V, with the addition of 5G broadband wireless connectivity, will give a major boost to all areas of IT here in 2022. WD has become one of the largest manufacturers of RISC-V processors and other products.
AI engine mimics the human brain
SiFive Intelligence-based processors have a highly configurable multi-core, multi-cluster-capable design optimized for a range of applications requiring high-throughput, single-thread performance while being under tight power and area constraints, Nadel said. .
Akida from BrainChip mimics the human brain to analyze only essential sensor inputs at the time of acquisition, processing data with efficiency, precision and energy savings, Nadel said. Keeping AI/ML local on the chip and independent of the cloud will reduce latency while improving privacy and data security, he said.
BrainChip’s technology is based on SNAP (spiking neuron adaptive processor) technology and licenses SNAP with technology partners. SNAP provides a development solution for companies entering the neuromorphic semiconductor chips market. It is a core technology in neuromorphic semiconductor chips that enables various applications such as gaming, cybersecurity, robotics, and stock market forecasting, among others.
“As we expand our ecosystem of portfolio partners, we want to ensure that our relationships are built on complementary technologies, capabilities, and a wide variety of environments so we can expand the opportunities to as many potential customers as possible,” said Nadel. “Driving our technology in a SiFive-based subsystem is exactly the kind of partnership that fulfills these goals.”
3 questions for BrainChip
VentureBeat asked Jack Kang, senior vice president of Business Development, Customer Experience (CX), Corporate Marketing at SiFive, a few specific questions about the news and the relevance of the partnership.
VentureBeat: What’s the #1 corporate takeaway from this announcement?
Jack Kang: For SiFive, this announcement demonstrates the continued adoption of the SiFive Intelligence family of RISC-V based processor IPs. More companies are choosing RISC-V as part of their product roadmap strategy, and SiFive is the leading provider of commercial RISC-V IP. In the emerging greenfield markets of AI/ML-enabled platforms, such as the edge processing market targeted by BrainChip, the performance by area and efficiency benefits of the SiFive processor architecture make the SiFive Intelligence family a competitive choice.
VentureBeat: Does BrainChip use any of Arm’s IP addresses in its chips? Arm is known for its low power consumption and high performance.
Kang: BrainChip discussed Arm IP for their product line. Arm processors have earned a reputation for low power consumption, based on comparisons to x86-based products. SiFive Intelligence products compare well to Arm products by offering improved performance per area of up to 30% combined with a single ISA for easier programming, and a modular approach well suited to working with hardened AI IP, as developed by BrainChip.
VentureBeat: Can you elaborate on this statement: “(Brainchip) mimics the human brain to analyze only essential sensor inputs at the time of acquisition.”
Kang: This statement refers to people’s ability to focus on what is important. For example, listening to a conversation in a coffee shop while still recording and acknowledging background noise. The BrainChip solution mimics this ability to reduce power and increase efficiency by focusing on the important data being processed. This is similar to, but a step beyond, applying mixed and lower precision data types (INT8 vs FP16) to speed up and improve the efficiency of AI/ML processing.
Competitors in the Market
Based in Aliso Viejo, California, Brainchip competes in the burgeoning intelligent-edge chip market with Nvidia Deep Learning GPU, Keras, TFLearn, Clarifai and Microsoft Cognitive Toolkit, AWS Deep Learning AMIs, and Torch. Nvidia owns about 80 percent of the global GPU (graphics processing unit) market. G2.com has market information here.Availability of the new SiFive/BrainChip solutions will be: announced later†
VentureBeat’s mission is to be a digital city square for tech decision makers to learn about transformative business technology and transactions. Learn more about membership.