Humble Genius
Regular
Anyone seen this?
BRN Discussion Ongoing
- Thread starter TechGirl
- Start date
DingoBorat
Slim
The IP changes to AKIDA 1.0 IP "were" taped out, with the subsequent runs of AKD1000 chips (which there "appear" to be 4 different types) and the AKD1500 In FD-SOI (which makes 5 that we know of, if the previous is correct).
I really don't know, what would constitute the need for a tape-out, to confirm a small tweak, which surely could be run in simulation these days.
With AKIDA 2.0 for example, I'm pretty sure Dr Anthony Lewis, said a tape-out wasn't specifically required, because the core technology, the "novel" aspect of AKIDA, had already been well proven.
Whispering Bob
Member
Hi there, I'm sorry but what kind of nonsense is this? anti-gravity technology??
After reading this how is one supposed to trust anything you write if you are so ready to believe such conspiracies? You say you work for Brainchip but what's not to say that this isn't also one of your delusions?
For the record I am a big believer in the company, just not non-sensical shit like what is written here.
Bravo
If ARM was an arm, BRN would be its biceps💪!
jtardif999
Regular
I would think that any change to the RTL blue prints would require a tape out.
jtardif999
Regular
The subsequent runs of AKD1000 were all improvements/fixes to the original Akida1.0 architecture after the initial run. There are no different types of the architecture? AKD1500 uses the Akida1.0 architecture but with changes to the SOC itself, around removing the reliance on the ARM preprocessor chip to allow the preprocessor to be the customers choice and the fab process for the SOC as the FDSOI process - the preference for space applications.
Just because humanity has not succeeded in developing anti-gravity technology doesn’t mean we also can’t do it. There are concepts that go far beyond human imagination. If people collectively realized this and stopped making such a mystery and drama out of it, as if the world would stop spinning if such things were officially acknowledged, we could take a massive step toward a new era of understanding and progress—an era where we redefine the boundaries of science and revolutionize our perspective on the universe and our place within it. But just my opinion
overpup
Regular
I just saw one of those “New Jersey” drones above my neighbourhood (central Victoria) and at the same time the power just went out - there is something weird going down that’s for sure!
JDelekto
Regular
Sorry, I didn't think that the seemingly unused light switch in my garage did anything. I promise not to toggle it again.
So how long have you lived in Castle Frankenstein?
DingoBorat
Slim
I think you must have misconstrued what I've said in my posts, as I never mentioned anything about different types of architecture?..
The words I used were "improvements" "subtle changes" and "tweaks" which are congruent with your "improvements/fixes" to the original run..
While the chips produced by/for BRN have a fixed number of nodes/NPUs, the IP licenced to customers can have any number of nodes/NPUs, up to 256 nodes for Akida 1.
BRN has produced different tapeouts for Akida 1 and Akida 1500. Akida 1 has 80 nodes. I can't recall seeing the number of nodes in 1500, but since 1500 is a Cortexless clone of Akia1, I guess it has 80 nodes, but that may not be correct.
IP exists for Akida 2 and Pico. Each node has 4 NPUs. Pico has a single NPU. We went through a stage where Akida 2 was going to be taped out, first under BRN's auspices, and then by a presumed customer - but that went cold.
The customers are responsible for taping out their specific architecture for the number of nodes they choose to license. We do know that initially Renesas only licensed 2 nodes (8 NPUs) for "low end work", preferring their in-house DRPAi for heavy lifting.
There are several examples of Akida 1 being used in projects, but, to date, we have not seen any user-customized SoCs produced by licencees.
Fullmoonfever
Top 20
Nice to see BAE across the benefits of neuromorphic. Thoughts by their Chief Technologist at #2 for neuromorphic & quantum.
Neuromorphic and quantum tech to bring real-world benefits
Rob Wythe, Chief Technologist
Neuromorphic vs. quantum computing
When it’s mature enough to be applied to real-world problems, quantum computing will be a game changer. But this reality is years away. In the shorter term, there is another form of high-performance computing we should be looking at: neuromorphic computing.
This is an alternative approach to computer architectures that tries to mimic how the brain works and is currently closer to practical application than quantum computing. One key application of this is for Spiking Neural Networks (SNNs), which are different from the neural networks we use today in most AI. Instead of processing information in a steady flow, like current AI models, SNNs send information in short bursts, or “spikes,” similar to how neurons in the brain communicate. This method is more energy efficient as it only works when there’s something to process. It’s also more powerful because it can handle the timing of signals, in terms of when they arrive and how that varies, making them closer to how real brains work. This leads to new abilities, like organising themselves into patterns that make their decisions easier to understand—a big improvement over current AI, which can be a “black box” and hard to explain.
Neuromorphic computing will become increasingly important in 2025 and beyond, particularly as AI demands more from computing than ever before. An area in which neuromorphic technology will play a key role is the ability to use AI at the edge – environments which don’t have supercomputers to run large language models. Neuromorphic chips can provide an answer here, by delivering a low power and highly constrained weight, size, power and cost environment.
Quantum and neuromorphic computer architectures have different strengths that make them more or less suitable for certain applications. Ultimately a hybrid architecture, just like conventional computers which use dedicated graphics processors, is likely to provide the best long term solution.
Quantum sensing
While quantum computing is a way off from practical implementation, the testing and development of quantum sensing is racing ahead. In 2025, this technology will become increasingly mature, especially in its ability to enhance Position, Navigation and Timing (PNT).
GPS jamming and spoofing presents a real-world threat, meaning the need to futureproof and enhance the resilience of PNT systems is vital. Quantum sensors have the potential to overcome recognised weaknesses in these systems by enabling us to calculate the position of aircraft without fully depending on satellite signals. For defence operations, this is important for a number of reasons. It ensures accurate navigation even in GPS-denied environments, significantly reduces risk to personnel and enhances mission readiness.
With the UK Government announcing its first test flight of quantum-based navigation systems, supported by BAE Systems, this technology is far from just a concept. As threats to society advance and become more complex on all fronts, quantum sensing can help keep vital infrastructure running, national security protected, and armed forces prepared.
2025: Our top five emerging tech predictions
Experts take a close look at the rise of AI-generated threats; neuromorphic and quantum computing; situational awareness in multi-domain environments; CEMA dominance across the connected battlespace and the role of space-derived intelligenceNeuromorphic and quantum tech to bring real-world benefits
Rob Wythe, Chief Technologist
Neuromorphic vs. quantum computing
When it’s mature enough to be applied to real-world problems, quantum computing will be a game changer. But this reality is years away. In the shorter term, there is another form of high-performance computing we should be looking at: neuromorphic computing.
This is an alternative approach to computer architectures that tries to mimic how the brain works and is currently closer to practical application than quantum computing. One key application of this is for Spiking Neural Networks (SNNs), which are different from the neural networks we use today in most AI. Instead of processing information in a steady flow, like current AI models, SNNs send information in short bursts, or “spikes,” similar to how neurons in the brain communicate. This method is more energy efficient as it only works when there’s something to process. It’s also more powerful because it can handle the timing of signals, in terms of when they arrive and how that varies, making them closer to how real brains work. This leads to new abilities, like organising themselves into patterns that make their decisions easier to understand—a big improvement over current AI, which can be a “black box” and hard to explain.
Neuromorphic computing will become increasingly important in 2025 and beyond, particularly as AI demands more from computing than ever before. An area in which neuromorphic technology will play a key role is the ability to use AI at the edge – environments which don’t have supercomputers to run large language models. Neuromorphic chips can provide an answer here, by delivering a low power and highly constrained weight, size, power and cost environment.
Quantum and neuromorphic computer architectures have different strengths that make them more or less suitable for certain applications. Ultimately a hybrid architecture, just like conventional computers which use dedicated graphics processors, is likely to provide the best long term solution.
Quantum sensing
While quantum computing is a way off from practical implementation, the testing and development of quantum sensing is racing ahead. In 2025, this technology will become increasingly mature, especially in its ability to enhance Position, Navigation and Timing (PNT).
GPS jamming and spoofing presents a real-world threat, meaning the need to futureproof and enhance the resilience of PNT systems is vital. Quantum sensors have the potential to overcome recognised weaknesses in these systems by enabling us to calculate the position of aircraft without fully depending on satellite signals. For defence operations, this is important for a number of reasons. It ensures accurate navigation even in GPS-denied environments, significantly reduces risk to personnel and enhances mission readiness.
With the UK Government announcing its first test flight of quantum-based navigation systems, supported by BAE Systems, this technology is far from just a concept. As threats to society advance and become more complex on all fronts, quantum sensing can help keep vital infrastructure running, national security protected, and armed forces prepared.
DingoBorat
Slim
Also, I think where I said "(which there "appear" to be 4 different types)" is where you misunderstood me..
I was referring to the silver ES B00/01/02 and the gold B00 chip, which I'm assuming are the "batch codes" I wasn't implying, that they were completely different AKIDA "types" of chip, just different runs, where small changes could have been made, which after all, is the purpose of Engineering Samples.
I still consider those small changes, to be changes, or improvements of the IP, whereas you may see the IP as fixed already, for AKIDA 1.0.
Last edited:
Apologies if shared already
www.eenewseurope.com
Low-power neuromorphic processing boosts cyberthreat detection
The Akida™ processor is integrated into the Quantum Ventura cyberthreat intelligence tool to provide protection for a variety of devices.
www.eenewseurope.com
Airbus and HENSOLDT revolutionize Germany's electronic warfare center with cutting-edge tech upgrades
In a groundbreaking partnership, Airbus Defence and Space has teamed up with German defense technology leader HENSOLDT to transform the “German Airborne Weapon Systems Electronic Warfare Center” into a state-of-the-art hub for electronic warfare (EW). This collaboration will arm the German Armed...
motorcyclesports.net
TheDrooben
Pretty Pretty Pretty Pretty Good
Another mention for us alongside IBM and Intel........
The Potential of Neuromorphic Computing: A New Frontier for Web Development
www.linkedin.com
Happy as Larry
The Potential of Neuromorphic Computing: A New Frontier for Web Development
The Potential of Neuromorphic Computing: A New Frontier for Web Development
Neuromorphic Computing: Revolutionizing AI In the rapidly evolving world of technology, neuromorphic computing is emerging as one of the most fascinating and transformative innovations. Still relatively unknown to many, neuromorphic computing offers a profound shift in how machines process informati
Happy as Larry
Bennysmadness
Regular
BRN potential market cap in the next 5 years?
Low - Med & High estimates.
For me I think it will look something like this
5B , 10B, 25B. How does this compare to others estimates?
Low - Med & High estimates.
For me I think it will look something like this
5B , 10B, 25B. How does this compare to others estimates?
Good morning Benny.....
The "B" part is correct in my opinion, things appear to be picking up slowly, lots more talk about Neuromorphic worldwide, Brainchip's
name getting published a lot more frequently, if things pan out the way most of us think that they will, in the 5 year time frame you
have mentioned, well, how longs a piece of string ?
I still personally feel a major will offer to buy us out or partner up with us in some form within the next 3 years, I would like to see our
company (through) clever marketing, start yelling out a major asset that we possess, Power Reduction at the Edge.
Over the weekend I read (not all) of a 79 page document published in December 2024 by the DOE with regards power consumption at the Data Centre, I have attached it below, we really are a "green company", and I feel that our company needs to publish an article highlighting why
Brainchip's technology is so beneficial for our planet (environment)....yes there's lots of great advantages with the AKIDA ever developing
technology, but a green, healthy environment resonates with the masses worldwide, lets get out there and sing to our strengths, I don't feel it's been pushed enough.
Regards....Tech.
Smoothsailing
Regular
BrainChip’s Edge AI Box include:
Edge AI Model training and development with Edge Impulse, demonstrating how easy it is to build and deploy custom machine learning models directly on the Akida Edge AI box.
Gesture recognition by BeEmotion, an innovator at the intersection of AI and human behavior that provides cutting-edge solutions to enhance interaction, prevent accidents, and bring enjoyment to the everyday use of technology.
Climate change solutions to forecast and gain visibility into contributors to climate change such as emissions, developed by AI Labs, a pioneer in AI, robotic process automation, and software development.
Model evaluation to streamline the AI workflow from experts at DeGirum, a leader in AI innovation with comprehensive solutions that simplify and accelerate the adoption of AI at the edge.
Cybersecurity engineered by Quantum Ventura Inc., provider of advanced R&D to the federal government in fields such as machine learning, hyperspectral imaging, cybersecurity, computer vision, and synthetic aperture radar, is demonstrating their CyberNeuro-RT (CNRT) technology, an AI/ML-driven, highly-scalable, real-time network defense & threat intelligence tool.
Computer vision analysis by Vedya Labs, an AI accelerator in edge, embedded systems, Gen AI, SDKs/toolchains, and developer of AI models, applications, optimization, libraries, and more will be demonstrating video object detection models.
Edge AI Model training and development with Edge Impulse, demonstrating how easy it is to build and deploy custom machine learning models directly on the Akida Edge AI box.
Gesture recognition by BeEmotion, an innovator at the intersection of AI and human behavior that provides cutting-edge solutions to enhance interaction, prevent accidents, and bring enjoyment to the everyday use of technology.
Climate change solutions to forecast and gain visibility into contributors to climate change such as emissions, developed by AI Labs, a pioneer in AI, robotic process automation, and software development.
Model evaluation to streamline the AI workflow from experts at DeGirum, a leader in AI innovation with comprehensive solutions that simplify and accelerate the adoption of AI at the edge.
Cybersecurity engineered by Quantum Ventura Inc., provider of advanced R&D to the federal government in fields such as machine learning, hyperspectral imaging, cybersecurity, computer vision, and synthetic aperture radar, is demonstrating their CyberNeuro-RT (CNRT) technology, an AI/ML-driven, highly-scalable, real-time network defense & threat intelligence tool.
Computer vision analysis by Vedya Labs, an AI accelerator in edge, embedded systems, Gen AI, SDKs/toolchains, and developer of AI models, applications, optimization, libraries, and more will be demonstrating video object detection models.
Similar threads
