BRN Discussion Ongoing
- Thread starter TechGirl
- Start date
Beyond GPUs: Innatera and the quiet uprising in AI hardware — VentureBeat
While much of the tech world remains fixated on the latest large language models (LLMs) powered by Nvidia GPUs, a quieter revolution is brewing in AI hardware.
Frangipani
Regular
The following article is based on an interview with Innatera founder and CEO Sumeet Kumar. It doesn’t mention BrainChip, but proves once again how neuromorphic technology in general is getting more and more exposure in the media, reflecting “the growing excitement around neuromorphic computing”.
The article ends with an outlook that predicts the high probability of a rising neuromorphic tide that will lift all seaworthy boats, no matter who the owner is:
“As AI continues to diffuse into every facet of our lives, the need for more efficient hardware solutions will only grow. Neuromorphic computing represents one of the most exciting frontiers in chip design today, with the potential to enable a new generation of intelligent devices that are both more capable and more sustainable (…) As these brain-inspired chips make their way into consumer devices and industrial systems, we may be on the cusp of a new era in artificial intelligence – one that’s faster, more efficient, and more closely aligned with the remarkable abilities of biological brains.”
Take a moment to recall what Sean Hehir said a few weeks ago at the AGM (from 59:50 min - the transcript of the middle part may not be 100% correct, though, as it was hard to understand)
“We welcome competition, because it certainly signals the interest in the market, right? You would be worried if you didn’t have competition, ‘cause you would say, “well, we’re the only ones seeing this market”, then there must be something wrong. So we welcome the competition. The idea with our benchmarking is to ensure that we’re always better than the competition.”
I personally think that in addition there should be standardised benchmarking conducted by other entities than the competing companies themselves, as potential customers would surely consider such a comparison to be more objective. That way Akida could prove its uniqueness in specific aspects.
Turns out, Innatera has meanwhile also partnered with one of the companies we have a relationship with, which shouldn’t really come as a surprise, though, given we are targeting the same Edge AI market (although Innatera’s spiking neural processor is of a different architecture and the company does not appear to be aiming for an IP-centred model to date):
“Innatera has partnered with Socionext, a Japanese sensor vendor, to develop an innovative solution for human presence detection. This technology, which Kumar demonstrated at CES in January, combines a radar sensor with Innatera’s neuromorphic chip to create highly efficient, privacy-preserving devices.”
The other day I pointed out that in recent months at least two Mercedes AI engineers have liked LinkedIn posts by Innatera, and so has Sounak Dey from TCS Research - IMO it is naive to just keep on bashing the competition and to repeat the mantra that we are years ahead of them:
I believe that consulting companies such as TCS or Accenture will happily offer their customers solutions with various providers if they see profitability for themselves - we should move away from the romantic notion that they will work with us exclusively. If they still end up doing so, great! But you shouldn’t get your hopes up that they will ignore our present or future competitors.
venturebeat.com
July 6, 2024 6:30 AM
While much of the tech world remains fixated on the latest large language models (LLMs) powered by Nvidia GPUs, a quieter revolution is brewing in AI hardware. As the limitations and energy demands of traditional deep learning architectures become increasingly apparent, a new paradigm called neuromorphic computing is emerging – one that promises to slash the computational and power requirements of AI by orders of magnitude.
“Neuromorphic processors are designed to mimic the way biological brains process information,” Kumar explained. “Rather than performing sequential operations on data stored in memory, neuromorphic chips use networks of artificial neurons that communicate through spikes, much like real neurons.”
This brain-inspired architecture gives neuromorphic systems distinct advantages, particularly for edge computing applications in consumer devices and industrial IoT. Kumar highlighted several compelling use cases, including always-on audio processing for voice activation, real-time sensor fusion for robotics and autonomous systems, and ultra-low power computer vision.
“The key is that neuromorphic processors can perform complex AI tasks using a fraction of the energy of traditional solutions,” Kumar noted. “This enables capabilities like continuous environmental awareness in battery-powered devices that simply weren’t possible before.”
Innatera’s flagship product, the Spiking Neural Processor T1, unveiled in January 2024, exemplifies these advantages. The T1 combines an event-driven computing engine with a conventional CNN accelerator and RISC-V CPU, creating a comprehensive platform for ultra-low-power AI in battery-powered devices.
“Our neuromorphic solutions can perform computations with 500 times less energy compared to conventional approaches,” Kumar stated. “And we’re seeing pattern recognition speeds about 100 times faster than competitors.”
Kumar illustrated this point with a compelling real-world application. Innatera has partnered with Socionext, a Japanese sensor vendor, to develop an innovative solution for human presence detection. This technology, which Kumar demonstrated at CES in January, combines a radar sensor with Innatera’s neuromorphic chip to create highly efficient, privacy-preserving devices.
“Take video doorbells, for instance,” Kumar explained. “Traditional ones use power-hungry image sensors that need frequent recharging. Our solution uses a radar sensor, which is far more energy-efficient.” The system can detect human presence even when a person is motionless, as long as they have a heartbeat. Being non-imaging, it preserves privacy until it’s necessary to activate a camera.
This technology has wide-ranging applications beyond doorbells, including smart home automation, building security and even occupancy detection in vehicles. “It’s a perfect example of how neuromorphic computing can transform everyday devices,” Kumar noted. “We’re bringing AI capabilities to the edge while actually reducing power consumption and enhancing privacy.”
To meet this growing demand, Innatera is ramping up production. The Spiking Neural Processor is slated to enter production later in 2024, with high-volume deliveries starting in Q2 of 2025. This timeline reflects the rapid progress the company has made since spinning out from Delft University of Technology in 2018. In just six years, Innatera has grown to about 75 employees and recently appointed Duco Pasmooij, former VP at Apple, to their advisory board.
The company recently closed a $21 million Series A round to accelerate the development of its spiking neural processors. The round, which was oversubscribed, included investors like Innavest, InvestNL, EIC Fund and MIG Capital. This strong investor backing underscores the growing excitement around neuromorphic computing.
Kumar envisions a future where neuromorphic chips increasingly handle AI workloads at the edge, while larger foundational models remain in the cloud. “There’s a natural complementarity,” he said. “Neuromorphics excel at fast, efficient processing of real-world sensor data, while large language models are better suited for reasoning and knowledge-intensive tasks.”
“It’s not just about raw computing power,” Kumar observed. “The brain achieves remarkable feats of intelligence with a fraction of the energy our current AI systems require. That’s the promise of neuromorphic computing – AI that’s not only more capable but dramatically more efficient.”
Innatera’s SDK uses PyTorch as a front end. “You actually develop your neural networks completely in a standard PyTorch environment,” Kumar noted. “So if you know how to build neural networks in PyTorch, you can already use the SDK to target our chips.”
This approach significantly lowers the barrier to entry for developers already familiar with popular machine learning frameworks. It allows them to leverage their existing skills and workflows while tapping into the power and efficiency of neuromorphic computing.
“It is a simple turnkey, standard, and very fast way of building and deploying applications onto our chips,” Kumar added, highlighting the potential for rapid adoption and integration of Innatera’s technology into a wide range of AI applications.
This move is telling. Despite Altman’s public statements about scaling up current AI technologies, his investment suggests a recognition that the path to more advanced AI may require a fundamental shift in computing architecture. Neuromorphic computing could be one of the keys to bridging the efficiency gap that current architectures face.
While large language models capture the headlines, the real future of AI may lie in chips that think more like our own brains. As Kumar put it: “We’re just scratching the surface of what’s possible with neuromorphic systems. The next few years are going to be very exciting.”
As these brain-inspired chips make their way into consumer devices and industrial systems, we may be on the cusp of a new era in artificial intelligence – one that’s faster, more efficient, and more closely aligned with the remarkable abilities of biological brains.
The article ends with an outlook that predicts the high probability of a rising neuromorphic tide that will lift all seaworthy boats, no matter who the owner is:
“As AI continues to diffuse into every facet of our lives, the need for more efficient hardware solutions will only grow. Neuromorphic computing represents one of the most exciting frontiers in chip design today, with the potential to enable a new generation of intelligent devices that are both more capable and more sustainable (…) As these brain-inspired chips make their way into consumer devices and industrial systems, we may be on the cusp of a new era in artificial intelligence – one that’s faster, more efficient, and more closely aligned with the remarkable abilities of biological brains.”
Take a moment to recall what Sean Hehir said a few weeks ago at the AGM (from 59:50 min - the transcript of the middle part may not be 100% correct, though, as it was hard to understand)
“We welcome competition, because it certainly signals the interest in the market, right? You would be worried if you didn’t have competition, ‘cause you would say, “well, we’re the only ones seeing this market”, then there must be something wrong. So we welcome the competition. The idea with our benchmarking is to ensure that we’re always better than the competition.”
I personally think that in addition there should be standardised benchmarking conducted by other entities than the competing companies themselves, as potential customers would surely consider such a comparison to be more objective. That way Akida could prove its uniqueness in specific aspects.
Turns out, Innatera has meanwhile also partnered with one of the companies we have a relationship with, which shouldn’t really come as a surprise, though, given we are targeting the same Edge AI market (although Innatera’s spiking neural processor is of a different architecture and the company does not appear to be aiming for an IP-centred model to date):
“Innatera has partnered with Socionext, a Japanese sensor vendor, to develop an innovative solution for human presence detection. This technology, which Kumar demonstrated at CES in January, combines a radar sensor with Innatera’s neuromorphic chip to create highly efficient, privacy-preserving devices.”
The other day I pointed out that in recent months at least two Mercedes AI engineers have liked LinkedIn posts by Innatera, and so has Sounak Dey from TCS Research - IMO it is naive to just keep on bashing the competition and to repeat the mantra that we are years ahead of them:
I believe that consulting companies such as TCS or Accenture will happily offer their customers solutions with various providers if they see profitability for themselves - we should move away from the romantic notion that they will work with us exclusively. If they still end up doing so, great! But you shouldn’t get your hopes up that they will ignore our present or future competitors.
Beyond GPUs: Innatera and the quiet uprising in AI hardware
The brain-inspired architecture gives neuromorphic systems distinct advantages, particularly for edge computing applications in consumer devices and industrial IoT.
Beyond GPUs: Innatera and the quiet uprising in AI hardware
James Thomason@jathomasonJuly 6, 2024 6:30 AM
While much of the tech world remains fixated on the latest large language models (LLMs) powered by Nvidia GPUs, a quieter revolution is brewing in AI hardware. As the limitations and energy demands of traditional deep learning architectures become increasingly apparent, a new paradigm called neuromorphic computing is emerging – one that promises to slash the computational and power requirements of AI by orders of magnitude.
Mimicking nature’s masterpiece: How neuromorphic chips work
But what exactly are neuromorphic systems? To find out, VentureBeat spoke with Sumeet Kumar, CEO and founder of Innatera, a leading startup in the neuromorphic chip space.“Neuromorphic processors are designed to mimic the way biological brains process information,” Kumar explained. “Rather than performing sequential operations on data stored in memory, neuromorphic chips use networks of artificial neurons that communicate through spikes, much like real neurons.”
This brain-inspired architecture gives neuromorphic systems distinct advantages, particularly for edge computing applications in consumer devices and industrial IoT. Kumar highlighted several compelling use cases, including always-on audio processing for voice activation, real-time sensor fusion for robotics and autonomous systems, and ultra-low power computer vision.
“The key is that neuromorphic processors can perform complex AI tasks using a fraction of the energy of traditional solutions,” Kumar noted. “This enables capabilities like continuous environmental awareness in battery-powered devices that simply weren’t possible before.”
From doorbell to data center: Real-world applications emerge
Innatera’s flagship product, the Spiking Neural Processor T1, unveiled in January 2024, exemplifies these advantages. The T1 combines an event-driven computing engine with a conventional CNN accelerator and RISC-V CPU, creating a comprehensive platform for ultra-low-power AI in battery-powered devices.
“Our neuromorphic solutions can perform computations with 500 times less energy compared to conventional approaches,” Kumar stated. “And we’re seeing pattern recognition speeds about 100 times faster than competitors.”
Kumar illustrated this point with a compelling real-world application. Innatera has partnered with Socionext, a Japanese sensor vendor, to develop an innovative solution for human presence detection. This technology, which Kumar demonstrated at CES in January, combines a radar sensor with Innatera’s neuromorphic chip to create highly efficient, privacy-preserving devices.
“Take video doorbells, for instance,” Kumar explained. “Traditional ones use power-hungry image sensors that need frequent recharging. Our solution uses a radar sensor, which is far more energy-efficient.” The system can detect human presence even when a person is motionless, as long as they have a heartbeat. Being non-imaging, it preserves privacy until it’s necessary to activate a camera.
This technology has wide-ranging applications beyond doorbells, including smart home automation, building security and even occupancy detection in vehicles. “It’s a perfect example of how neuromorphic computing can transform everyday devices,” Kumar noted. “We’re bringing AI capabilities to the edge while actually reducing power consumption and enhancing privacy.”
Doing more with less in AI compute
These dramatic improvements in energy efficiency and speed are driving significant industry interest. Kumar revealed that Innatera has multiple customer engagements, with traction for neuromorphic technologies growing steadily. The company is targeting the sensor-edge applications market, with an ambitious goal of bringing intelligence to a billion devices by 2030.To meet this growing demand, Innatera is ramping up production. The Spiking Neural Processor is slated to enter production later in 2024, with high-volume deliveries starting in Q2 of 2025. This timeline reflects the rapid progress the company has made since spinning out from Delft University of Technology in 2018. In just six years, Innatera has grown to about 75 employees and recently appointed Duco Pasmooij, former VP at Apple, to their advisory board.
The company recently closed a $21 million Series A round to accelerate the development of its spiking neural processors. The round, which was oversubscribed, included investors like Innavest, InvestNL, EIC Fund and MIG Capital. This strong investor backing underscores the growing excitement around neuromorphic computing.
Kumar envisions a future where neuromorphic chips increasingly handle AI workloads at the edge, while larger foundational models remain in the cloud. “There’s a natural complementarity,” he said. “Neuromorphics excel at fast, efficient processing of real-world sensor data, while large language models are better suited for reasoning and knowledge-intensive tasks.”
“It’s not just about raw computing power,” Kumar observed. “The brain achieves remarkable feats of intelligence with a fraction of the energy our current AI systems require. That’s the promise of neuromorphic computing – AI that’s not only more capable but dramatically more efficient.”
Seamless integration with existing tools
Kumar emphasized a key factor that could accelerate the adoption of their neuromorphic technology: developer-friendly tools. “We’ve built a very extensive software development kit that allows application developers to easily target our silicon,” Kumar explained.Innatera’s SDK uses PyTorch as a front end. “You actually develop your neural networks completely in a standard PyTorch environment,” Kumar noted. “So if you know how to build neural networks in PyTorch, you can already use the SDK to target our chips.”
This approach significantly lowers the barrier to entry for developers already familiar with popular machine learning frameworks. It allows them to leverage their existing skills and workflows while tapping into the power and efficiency of neuromorphic computing.
“It is a simple turnkey, standard, and very fast way of building and deploying applications onto our chips,” Kumar added, highlighting the potential for rapid adoption and integration of Innatera’s technology into a wide range of AI applications.
Silicon Valley’s stealth game
While LLMs capture the headlines, industry leaders are quietly acknowledging the need for radically new chip architectures. Notably, OpenAI CEO Sam Altman, who has been vocal about the imminent arrival of artificial general intelligence (AGI) and the need for massive investments in chip manufacturing, personally invested in Rain, another neuromorphic chip startup.This move is telling. Despite Altman’s public statements about scaling up current AI technologies, his investment suggests a recognition that the path to more advanced AI may require a fundamental shift in computing architecture. Neuromorphic computing could be one of the keys to bridging the efficiency gap that current architectures face.
Bridging the gap between artificial and biological intelligence
As AI continues to diffuse into every facet of our lives, the need for more efficient hardware solutions will only grow. Neuromorphic computing represents one of the most exciting frontiers in chip design today, with the potential to enable a new generation of intelligent devices that are both more capable and more sustainable.While large language models capture the headlines, the real future of AI may lie in chips that think more like our own brains. As Kumar put it: “We’re just scratching the surface of what’s possible with neuromorphic systems. The next few years are going to be very exciting.”
As these brain-inspired chips make their way into consumer devices and industrial systems, we may be on the cusp of a new era in artificial intelligence – one that’s faster, more efficient, and more closely aligned with the remarkable abilities of biological brains.
Frangipani
Regular
To further prove my point about Accenture keeping a two (or more)-pronged strategy re collaboration with neuromorphic players: Ten days ago, a paper about a resonator network (whatever that is) implemented on Loihi was published, co-authored by one of the co-inventors of the Accenture patent published on March 21, 2024 and previously posted here, namely Lavinia Andreea Danielescu (who used to work for Intel from 2014-2018), as well as her Accenture Labs colleague Lazar Supic, alongside six researchers from other institutions, including Intel Labs.
Bravo
If ARM was an arm, BRN would be its biceps💪!
Edge AI and its impact on network infrastructure | Communications Today
Important referral for telecom, networking, IT solution equipment buyer, and broadcasters since the last 25 years.
www.communicationstoday.co.in
Extract 1
Extract 2
7für7
Top 20
Wow the last market valuation was predicted around 30 billion !! I’m wondering where we will end upEdge AI and its impact on network infrastructure | Communications Today
Important referral for telecom, networking, IT solution equipment buyer, and broadcasters since the last 25 years.
View attachment 66109
Extract 1
View attachment 66110
Extract 2
View attachment 66112
Mt09
Regular
Hopefully with a license or two that produces regular income..
7für7
Top 20
I don’t! Better expect nothing…then you can not get disappointed
Frangipani
Regular
Strictly speaking, Jonathan Tapson is not a new hire, as he has in fact been working for us at least since April.
Three months ago, I had spotted him in a picture posted on LinkedIn by Frontgrade Gaisler, showing him standing next to their CEO Sandi Habinc at their booth at the Space Symposium’s Exhibit Center in Colorado Springs, wearing a lanyard with a conference badge that identified him as a representative of our company. The caption above the photo also stated he was from BrainChip and that he was visiting Frontgrade Gaisler’s booth to “discuss neuromorphic processing for space applications.”
(Less than a month later, BrainChip and Frontgrade Gaisler officially announced their collaboration “to explore the integration of BrainChip’s Akida TM neuromorphic processor into Frontgrade Gaisler’s next-generation fault-tolerant, radiation-hardened microprocessors”.
https://thestockexchange.com.au/threads/brn-discussion-ongoing.1/post-421614)
To (re)read my original and the ensuing posts:
https://thestockexchange.com.au/threads/brn-discussion-ongoing.1/post-418644
While some users here and elsewhere speculated that I could be wrong and that Jonathan Tapson was possibly just a visitor or guest of BrainChip (some even failed to notice that the photo was taken in Colorado and mistook the Frontgrade Gaisler booth there for our booth at the 2024 embedded world - a totally different conference concurrently taking place in Germany), @mrgds kindly enquired with Investor Relations whether or not Jonathan Tapson was indeed working for us, which was subsequently confirmed:
So from our outward perspective pretty much the only thing that has since changed is that Jonathan Tapson stepped down from his position as CTO at Iona Tech last month and is now working for us full-time as a Senior R&D Consultant.
However, I am afraid I can’t answer the legitimate question why his previous part-time consultancy engagement with BrainChip (at least April - May 2024, if not longer) is not at all reflected in his LinkedIn work experience.
Given that his LinkedIn profile now says he is “Exploring options in Hardware and AI” and that he continues to be engaged as a consultant rather than having become a member of the BrainChip C-suite could in my eyes point to him seeing himself working full-time for us only temporarily. Either because he already has another adventure in mind - likely as a CEO or CTO somewhere, if not returning to academia - or alternatively it could signify that we are currently not able to pay him a competitive compensation package for a more permanent position with us (I am pretty sure our company could otherwise have come up with a creative title and role for him).
Anyway, taking into account his decades-long expertise in things neuromorphic as well as his large professional network reaching all around the globe, I love having him aboard the BrainChip bus, regardless of how short or long his ride with us will be!
Attachments
Last edited:
Fullmoonfever
Top 20
Obviously we know about the edge box but this is a decent summary article written on a Chinese site.
makerpro.cc
Login|Registe
Author: Lu Xiangyang
Figure 1 Akida Edge AI Box (left) and artistic AKD1000 chip (right) (Image source: brainchip)
BrainChip is an Australian technology company. The company's main business is to design and sell AI acceleration circuits. The target customers are chip manufacturers. It is expected that chip manufacturers will adopt the AI hardware acceleration circuit design developed by BrainChip and pay technology licensing fees and subsequent chip mass production. The business model is similar to the well-known Arm.
In order to accelerate the chip manufacturer's evaluation and adoption of the company's acceleration circuit design, BrainChip actually launched Akida1000 Ref SoC (AKD1000 for short), which is a reference system-on-chip. The main body in the chip is BrainChip's AI acceleration circuit, which is called Akida.
The so-called "reference" means that this is not the main business of BrainChip, but just a technology demonstration to facilitate potential buyers to evaluate the company's technology. There is no intention to mass-produce and sell the chip. ( BrainChip officially emphasized that the chip uses TSMC’s 28nm process
Figure 2 Main functions inside the AKD1000 chip (Image source: brainchip)
However, BrainChip seems to feel that it has not done enough. It should give more terminal AI application developers the opportunity to come into contact with BrainChip's technology, experience its advantages and value, and then help promote its technology licensing business. Therefore, BrainChip Corporation, in collaboration with VVDN Technologies, a software and hardware design and manufacturing service provider, launched the Akida Edge AI Box.
It is worth noting that BrainChip actually already has a second-generation Akida acceleration circuit, but it has not yet been made into Ref SoC, so it is not used in the Akida Edge AI Box. Second-generation acceleration circuits have some new features, such as event-based temporal neural networks (Temporal Event-based Neural Nets, TRNNs), or visual transformation (Vision Transformer, ViT).
In addition to reference chips, BrainChip also provides other technical supporting facilities, such as launching a software development environment called MetaTF, which can convert ready-made TensorFlow models, build and train new models, convert CNNs into SNNs, etc. It also provides converted software execution environment, called Akida Runtime.
In order to promote Akida Edge AI Box, BrainChip has once again entered the software field like MetaTF. It promises to launch an operating system based on the Linux 6.1 core for Akida Edge AI Box to use. It also provides multiple Edge AI demonstration applications to potential users. Engineers who want to develop Edge AI can evaluate and use BrainChip's circuit technology faster.
Figure 3 MetaTF development tool, schematic diagram of environmental functions (Image source: BrainChip)
As for VVDN, it also pays attention to the system design of Akida Edge AI Box. For example, the machine uses passive heat dissipation entirely from heat sinks, which saves power, stability, cost, and silence compared to active electric fans.
In addition, Akida Edge AI Box has a USB 3.0 Type-A interface and a Micro-USB interface, which can be used to update firmware or debug; then it has an HDMI video output, 4GB LPDDR memory, and 32GB eMMC storage space, plus a Micro-SDXC specification card slot, which can access memory cards with a maximum capacity of 1TB; then there are two sets of Wi-Fi wireless networks、 For two sets of GbE cable networks, VVDN stated that GbE is mainly used to connect cameras, but basic sales do not include cameras.
In terms of power supply, the Akida Edge AI Box uses a 12V DC power supply. Even if the official does not come with a power supply or the power supply fails, it is easy to find an alternative transformer.
The second is that Akida Edge AI Box can only be regarded as one member hardware of Akida Enablement Platforms in the eyes of BrainChip. Other member hardware is Akida PCIe Board, Akida Development Kit Shuttle PC, and Akida Development Kit Raspberry Pi.
To put it more simply, software developers who want to evaluate BrainChip technology have four hardware options. The Akida Edge AI Box in open pre-order is one of the four. In fact, the Akida Edge AI Box is similar to the Akida Development Kit Raspberry Pi. The difference is that the Akida Development Kit Raspberry Pi is equipped with an Akida PCIe Board with an acceleration chip, while the Akida Edge AI Box is equipped with two on the system circuit board， It has higher acceleration, and the entire body is relatively compact and petite.
In addition, an Akida PCIe Board alone costs US$499, and both the Akida Development Kit Shuttle PC and Akida Development Kit Raspberry Pi are currently sold out (late February). In this way, the US$799 Akida Edge AI Box is quite attractive.
Figure 4 The PCIe Board with one AKD1000 chip costs $499, and the Akida Edge AI Box with two chips costs $799 (Image source: brainchip)
(Editor: Xie Jiaxun.)
Lu Xiangyang
After graduating from the Department of Electronics, he continued this indissoluble bond with "electronic technology" as a media person. He has successively served as an author of electronic technology books, electronic media reporters, analysts, etc., and continues to write. Recently, I have invested in research and tutorial introduction of open hardware such as Arduino and Raspberry Pi.
不只是評估套件？BrainChip推出Akida Edge AI Box閘道器
BrainChip與VVDN Technologies合作，推出了Akida Edge AI Box，本篇文章...
makerpro.cc
Login|Registe
Not just an evaluation kit? BrainChip launches Akida Edge AI Box Gateway
Author: Lu Xiangyang
Figure 1 Akida Edge AI Box (left) and artistic AKD1000 chip (right) (Image source: brainchip)
BrainChip is an Australian technology company. The company's main business is to design and sell AI acceleration circuits. The target customers are chip manufacturers. It is expected that chip manufacturers will adopt the AI hardware acceleration circuit design developed by BrainChip and pay technology licensing fees and subsequent chip mass production. The business model is similar to the well-known Arm.
In order to accelerate the chip manufacturer's evaluation and adoption of the company's acceleration circuit design, BrainChip actually launched Akida1000 Ref SoC (AKD1000 for short), which is a reference system-on-chip. The main body in the chip is BrainChip's AI acceleration circuit, which is called Akida.
The so-called "reference" means that this is not the main business of BrainChip, but just a technology demonstration to facilitate potential buyers to evaluate the company's technology. There is no intention to mass-produce and sell the chip. ( BrainChip officially emphasized that the chip uses TSMC’s 28nm process
Figure 2 Main functions inside the AKD1000 chip (Image source: brainchip)
However, BrainChip seems to feel that it has not done enough. It should give more terminal AI application developers the opportunity to come into contact with BrainChip's technology, experience its advantages and value, and then help promote its technology licensing business. Therefore, BrainChip Corporation, in collaboration with VVDN Technologies, a software and hardware design and manufacturing service provider, launched the Akida Edge AI Box.
Akida Edge AI Box
The Akida Edge AI Box is a complete computer with a circuit board and an organic shell. It is equipped with 2 AKD1000s mentioned above. However, after all, the AKD1000 focuses on AI acceleration, and the internal control processing unit is weak, so the overall system still needs a master control chip. For this VVDN uses an i.MX 8M Plus processor chip from NXP. In order to make AI computing as fast as possible, AKD1000 has a PCI Express interface, which is used to communicate with i. MX 8M Plus processor chip.It is worth noting that BrainChip actually already has a second-generation Akida acceleration circuit, but it has not yet been made into Ref SoC, so it is not used in the Akida Edge AI Box. Second-generation acceleration circuits have some new features, such as event-based temporal neural networks (Temporal Event-based Neural Nets, TRNNs), or visual transformation (Vision Transformer, ViT).
In addition to reference chips, BrainChip also provides other technical supporting facilities, such as launching a software development environment called MetaTF, which can convert ready-made TensorFlow models, build and train new models, convert CNNs into SNNs, etc. It also provides converted software execution environment, called Akida Runtime.
In order to promote Akida Edge AI Box, BrainChip has once again entered the software field like MetaTF. It promises to launch an operating system based on the Linux 6.1 core for Akida Edge AI Box to use. It also provides multiple Edge AI demonstration applications to potential users. Engineers who want to develop Edge AI can evaluate and use BrainChip's circuit technology faster.
Figure 3 MetaTF development tool, schematic diagram of environmental functions (Image source: BrainChip)
As for VVDN, it also pays attention to the system design of Akida Edge AI Box. For example, the machine uses passive heat dissipation entirely from heat sinks, which saves power, stability, cost, and silence compared to active electric fans.
In addition, Akida Edge AI Box has a USB 3.0 Type-A interface and a Micro-USB interface, which can be used to update firmware or debug; then it has an HDMI video output, 4GB LPDDR memory, and 32GB eMMC storage space, plus a Micro-SDXC specification card slot, which can access memory cards with a maximum capacity of 1TB; then there are two sets of Wi-Fi wireless networks、 For two sets of GbE cable networks, VVDN stated that GbE is mainly used to connect cameras, but basic sales do not include cameras.
In terms of power supply, the Akida Edge AI Box uses a 12V DC power supply. Even if the official does not come with a power supply or the power supply fails, it is easy to find an alternative transformer.
Akida Edge AI Box Hardware Selection
Finally, there are two points to explain. First, the Akida Edge AI Box was announced as early as December 2023, and pre-orders were only open in February. It is initially speculated that the official release will not be until July 2024. The specific shipment time will be announced in April. The current price is $799.The second is that Akida Edge AI Box can only be regarded as one member hardware of Akida Enablement Platforms in the eyes of BrainChip. Other member hardware is Akida PCIe Board, Akida Development Kit Shuttle PC, and Akida Development Kit Raspberry Pi.
To put it more simply, software developers who want to evaluate BrainChip technology have four hardware options. The Akida Edge AI Box in open pre-order is one of the four. In fact, the Akida Edge AI Box is similar to the Akida Development Kit Raspberry Pi. The difference is that the Akida Development Kit Raspberry Pi is equipped with an Akida PCIe Board with an acceleration chip, while the Akida Edge AI Box is equipped with two on the system circuit board， It has higher acceleration, and the entire body is relatively compact and petite.
In addition, an Akida PCIe Board alone costs US$499, and both the Akida Development Kit Shuttle PC and Akida Development Kit Raspberry Pi are currently sold out (late February). In this way, the US$799 Akida Edge AI Box is quite attractive.
Figure 4 The PCIe Board with one AKD1000 chip costs $499, and the Akida Edge AI Box with two chips costs $799 (Image source: brainchip)
Summary
The author believes that the most suitable role of the Akida Edge AI Box is the IoT gateway. In addition to transmitting, converting, and temporarily storing the data uploaded by the sensor, it also conducts immediate AI inference and judgment on the sensed data, that is, it exerts the power of Edge AI, but I believe that in addition There are many more wonderful possibilities. Let’s see how creative makers can play it.(Editor: Xie Jiaxun.)
Lu Xiangyang
After graduating from the Department of Electronics, he continued this indissoluble bond with "electronic technology" as a media person. He has successively served as an author of electronic technology books, electronic media reporters, analysts, etc., and continues to write. Recently, I have invested in research and tutorial introduction of open hardware such as Arduino and Raspberry Pi.
Bravo
If ARM was an arm, BRN would be its biceps💪!
It would be weird if Jonathan Tapson knew how to tap dance. It would be like being a plumber who’s name was Lui Drain.
Last edited:
Thanks Fmf,
A very positive writeup.
Things that could have been better expressed -
"the internal control processing unit is weak"
How about:
"While Akida has an internal microprocessor to configure the NN, it is capable of functioning together with an external CPU to utilize the AI inference/classification carried out by Akida".
I'd missed this news:
"It promises to launch an operating system based on the Linux 6.1 core for Akida Edge AI Box to use."
Also interesting that he quotes VVDN:
"For two sets of GbE cable networks, VVDN stated that GbE is mainly used to connect cameras, but basic sales do not include cameras."
A very positive writeup.
Things that could have been better expressed -
"the internal control processing unit is weak"
How about:
"While Akida has an internal microprocessor to configure the NN, it is capable of functioning together with an external CPU to utilize the AI inference/classification carried out by Akida".
I'd missed this news:
"It promises to launch an operating system based on the Linux 6.1 core for Akida Edge AI Box to use."
Also interesting that he quotes VVDN:
"For two sets of GbE cable networks, VVDN stated that GbE is mainly used to connect cameras, but basic sales do not include cameras."
Fullmoonfever
Top 20
Yeah, thought there was some potentially additional info in there and to add to the comment about the control processing, need remember this was written in Chinese and run via a translator so never know how exactly it was translated I guess.
DingoBorat
Slim
"Given that his LinkedIn profile now says he is “Exploring options in Hardware and AI” and that he continues to be engaged as a consultant rather than having become a member of the BrainChip C-suite could in my eyes point to him seeing himself working full-time for us only temporarily. Either because he already has another adventure in mind - likely as a CEO or CTO somewhere, if not returning to academia - or alternatively it could signify that we are currently not able to pay him a competitive compensation package for a more permanent position with us (I am pretty sure our company could otherwise have come up with a creative title and role for him)"
I'm not sure what's wrong with the title of "Senior R&D Consultant"?
I don't think it's got anything to do with us not paying him what he's worth..
Maybe it's different in Germany, but my understanding, is that anything with "Consultant" attached to the title, usually means they are getting very well paid, for their input..
Although it does mean he may only hold that position until he's fulfilled some particular purpose, it could be a few months, or years..
Last edited:
DingoBorat
Slim
The Brainchip website does say it supports Linux 6.1..
But what about..
"For example, the machine uses passive heat dissipation entirely from heat sinks, which saves power, stability, cost, and silence compared to active electric fans"
I thought they did end up adding a fan, or is that just a vent?
I recall an earlier version had 4 holes in a fan shaped arrangement and the first images we saw, of the Edge Box, had no openings at all?..
Last edited:
Frangipani
Regular
Ah, I see.
I was under the impression that a Senior Consultant was more of an upper mid-level role, which would mean that this now being his full-time job, it would be a few steps down the career ladder for him, after two decades of being a university professor both in South Africa and then in Australia (as well as Director of the MARCS institute at WSU for two years), followed by company positions as CSO (GrAI Matter Labs) and CTO (IONA Tech).
Somehow I would have expected a gentleman of his calibre to fill a position with a fancier title - existing or newly created, such as CNO (Chief Neuromorphic Officer).
After a quick Google search, it appears the corporate hierarchy is indeed not standardised. Most of the examples of Senior Consultants I found referred to positions in consulting companies, but even they differ:
Last edited:
Frangipani
Regular
Was it someone else who suggested you better edit in some “joking” comment half an hour after posting the video on its own, in order to somewhat distract from your thinly veiled attack on me, or was it your own idea?
To save others from googling, here are the charming lyrics of the song to which those five Korean ladies are performing their choreographed dance moves:
Here To Dance (by The Veronicas)
Blood on the floor, look what they started
Hungry for more, they're spitting garbage
I'm getting bored and broken hearted
They won't ever get the best of me
You all need to stop
Not gonna listen to that blah blah blah
Someone's always talking - I don't wanna hear it
All that noise is toxic
I just want to feel this bass inside my body
And this moment in my hands
I don't care who's watching, I'm just here to dance
I'm just here to dance
I'm just here to dance
No time for tears, I'm busy sweating
Come over here, teach you a lesson
If you need a sign, I'll be your heaven
Yeah you'll only get the best of me
You all need to stop
Not gonna listen to that blah blah blah
(I'm not gonna listen)
Someone's always talking - I don't wanna hear it
All that noise is toxic
I just want to feel this bass inside my body and this moment in my hands
I don't care who's watching, I'm just here to dance
I'm just here to dance
I'm just here to dance
I'm just here to dance
I'm just here to dance
Yeah, whatever…
Why don’t you simply get yourself noise-cancelling headphones, then?
And just for the record - I didn’t start the fire.
Instead of replying to my post with the sole intention of insulting me, you could have easily used the time to educate yourself about the difference between tap dancing and line dancing. Let me thus suggest the following video, in which a cat that I find heaps cooler than you will show you how to tap dance (from 1:20 min):
Similar threads
