BRN Discussion Ongoing

Esq.111

Fascinatingly Intuitive.
I believe that SNNs have some inate resilience to radiation errors because they opertae on probability rather than mathematical precision used in CPU/GPU. A one bit error in a CPU will produce an incorrect answer, whereas in an SNN, it will make a small difference in the probability, but would not affect the result .

I think the fully depleted silicon on insulator also improves radhardness.

That said, I'm not aware that BRN has itself done a radhard design.
Evening Diogenese ,

From a layman , belive GLOBAL FOUNDRIES dabble on depleted silicone, ....... thay knocked out our AKD1500 in depleted silicone if memory serves correct.

Regards,
Esq.
 
  • Like
  • Love
Reactions: 8 users

cosors

👀
Someone is probably alluding to Richard Montañez, the son of Mexican immigrants, who initially worked as a caretaker for the snack manufacturer Frito-Lay. One day, he came up with the idea for Flamin' Hot Cheetos. At least that's the story of the movie I saw recently.
 
  • Haha
  • Like
Reactions: 3 users

Guzzi62

Regular
Nothing new, not sure if this has been posted:

https://www.design-reuse.com/news/57632/frontgrade-gaisler-grain-neuromorphic-ai-space.html


Frontgrade Gaisler Launches New GRAIN Line and Wins SNSA Contract to Commercialize First Energy-Efficient Neuromorphic AI for Space Applications​

Gothenburg, Sweden -- April 3, 2025 The Swedish National Space Agency (SNSA) has awarded Frontgrade Gaisler, a leading provider of radiation-hardened microprocessors for space missions, a contract to commercialize the first neuromorphic System on Chip (SoC) device for space applications. Already in development at Frontgrade Gaisler, the device is part of the company’s new GRAIN (Gaisler Research Artificial Intelligence NOEL-V) product line.
The first GRAIN device that Frontgrade Gaisler will premier – the GR801 SoC – integrates AkidaTM neuromorphic technology from BrainChip, the world’s first commercial producer of ultra-low power, fully digital, event-based, neuromorphic AI. The GR801 combines Gaisler’s NOEL-V RISC-V processor and the Akida™ neuromorphic AI processor into a single integrated circuit to enable energy-efficient AI applications in the space environment. Sweden’s Royal Institute of Technology (KTH) is contributing to this development by designing a demonstration application that uses a neuromorphic sensor directly connected to Gaisler’s new GR801 device
.

If Frontgrade is bringing out a processor containing Akida, is it under a licence or a partnership?

Frontgrade are designing a complete SoC including a processor and Akida IP (so we don't need to worry about where the chips are coming from).


This looks like a licence as it has a specific Frontgrade product number, but the article refers to "collaboration".

Our continued collaboration with Frontgrade Gaisler to incorporate Akida IP into space SoCs showcases the importance of having environmentally hardened solutions, already proven to perform in the most extreme condition,” said Sean Hehir, CEO of BrainChip. “We have worked hard to ensure our neuromorphic technology can meet the low-energy,low-latency, high-performance needs of GRAIN and other space-based devices in order to provide AI at – and beyond – the edge.”

... "already proven in the most extreme conditions"

We've already had a couple of radhard developments:

https://brainchip.com/edgx-announce...sruptive-data-processing-solutions-for-space/



Then there's EdgeX and Ant61.

While space is a high prestige application, it is relatively low volume, but there is often a crossover from space to military.
It's on F.G. website.

Status: Development.

I doubt they pay an IP license fee but pure guesswork on my side.

They will be sold in very limited numbers (obviously) those radiation-hardened microprocessors, but they are very expensive. I vaguely recall I saw the price on similar items somewhere, but I will be dammed if I remember where, and they are something like 10-20kUS$ a pop.

Not impossible BRN will be getting over $US1000 per processor, more money per chip if no IP deal is signed.

Not something that will make BRN loads of money, but the commercial value is worth gold IMO.

All the above is mostly guesswork from a layman, so take it with a bit of salt.

 
  • Like
  • Love
Reactions: 8 users

Rach2512

Regular


Good bit of advertising, front and centre stage, is that banner in the middle the same as the one in the last photo?


Screenshot_20250403_230052_Samsung Internet.jpg
Screenshot_20250403_230108_Samsung Internet.jpg
Screenshot_20250403_225653_Samsung Internet.jpg
 
  • Like
  • Love
  • Fire
Reactions: 28 users

Rach2512

Regular
  • Like
  • Love
  • Fire
Reactions: 34 users

cosors

👀
Here is an example of a really disastrously bad investment.
Boeing invested 1M in Trump's election campaign. Today they reckon that thanks to Trump they expect additional costs of 5B.
Perhaps some of them will now take off their red cap and great beautiful colouring sunglasses and think about it.

______
The Airfoce One is also a financial disaster for Boeing. They will not be finished it during his term of office (this one, you never know what he will do). The costs have already exploded.

___
I recommend

and
 
Last edited:
  • Like
  • Love
  • Wow
Reactions: 10 users

Frangipani

Top 20


View attachment 79017


View attachment 79020



View attachment 79018

View attachment 79019

Another video clip from our booth at Embedded World 2025, slightly different from the one already posted last month:


3CE627E5-16E3-44AD-AC6F-0B7257A79ED9.jpeg



„[Our Generation 2 IP] It’s all implemented in our box that we have here. It’s running on an FPGA, and so our customers can actually try it out in reality here, in real hardware, and the next step may or may not be to put this into a chip.”
 
Last edited:
  • Like
  • Fire
  • Love
Reactions: 29 users

Frangipani

Top 20

530B4C49-54D0-4711-84CC-0B36340DB054.jpeg
 
  • Like
  • Love
  • Fire
Reactions: 30 users
View attachment 81255

From Room-Sized Machines to Brain-Like Chips"

Remember when computers filled entire rooms? Then came microchips that fit in your palm. Today, we are witnessing another revolution with neuromorphic chips - hardware designed to work like the most powerful computer ever created: the human brain. 🧠

Traditional computers follow the Von Neumann architecture - a design where information moves back and forth between memory and processing units. But your brain? It processes everything at once, learning as it goes, using tiny amounts of energy.

Neuromorphic chips break this bottleneck by mimicking how our brains process information. Take Intel Corporation's Loihi 2 chip as an example. Unlike conventional processors, Loihi 2 features:
🔹 1 million artificial neurons connected by synapses
🔹 Spike-based communication (neurons only "fire" when needed)
🔹 On-chip learning that happens in real-time
🔹 Asynchronous circuits that eliminate the need for a central clock

This approach could transform embedded systems in several practical ways:
🔹 Smart Hearing Aids that filter out background noise in busy environment
🔹 Security Cameras that detect patterns of suspicious activities
🔹 Industrial Sensors that run on minimal power while monitoring systems
🔹 Medical Devices that adapt to individual patient patterns

These changes are not just theoretical. Companies like BrainChip are already integrating neuromorphic processors in industries like automotive and healthcare are exploring real-world applications.

What is more exciting is the impact on AI. Today’s AI often relies on massive cloud-based models that demand constant updates and huge data centers. Neuromorphic chips could shift this to edge computing allowing AI to learn and adapt in real-time, using far less power and reducing dependency on the cloud.

By 2027, these brain-like chips could become as common as ARM processors are today. What embedded system in your life would benefit most from brain-like processing?
View attachment 81256
Guess I'll wait until 2027 then👴
 
  • Like
Reactions: 2 users

Bravo

If ARM was an arm, BRN would be its biceps💪!
I could be reading too much into things here, but "using new technology" has certainly piqued my interest!


giphy (2).gif





Ask the Developer Vol. 16: Nintendo Switch 2 — Part 4​


EXTRACT ONLY -

Screenshot 2025-04-04 at 9.22.43 am.png



 
  • Like
  • Thinking
  • Love
Reactions: 46 users
The US markets reaction to the tarrifs aren't so bad..
I thought they would be much worse 🙄..

 
  • Haha
  • Like
Reactions: 11 users
  • Like
  • Love
  • Fire
Reactions: 29 users

Bravo

If ARM was an arm, BRN would be its biceps💪!
The US markets reaction to the tarrifs aren't so bad..
I thought they would be much worse 🙄..




Do you think Donald might have mistaken the word "boom" for another word starting with B? 💣

elmo-fire.gif




Screenshot 2025-04-04 at 10.07.25 am.png
 
  • Haha
  • Like
  • Sad
Reactions: 10 users

7für7

Top 20
Great news flow… this should push us to 15cent ! Wohoooo🤡
 
  • Sad
  • Like
  • Haha
Reactions: 5 users
  • Like
  • Sad
Reactions: 4 users

manny100

Top 20
I would expect that the DOD transistion to the Edge and our connection to Bascom Hunter (Navy) and US AFRL and Raytheon coupled with our client and Tech events since Sept'24 would give our patent portfolio value a material boost.
 
  • Like
  • Love
  • Fire
Reactions: 14 users

perceptron

Regular
Another video clip from our booth at Embedded World 2025, slightly different from the one already posted last month:


View attachment 81286


„[Our Generation 2 IP] It’s all implemented in our box that we have here. It’s running on an FPGA, and so our customers can actually try it out in reality here, in real hardware, and the next step may or may not be to put this into a chip.”
Thanks for posting. I watched the video and was impressed by Dr Lewis. Further, the mind boggles at what a 1 billion parameter model for LLM run on low power at the edge with Akida 2.0 can be used for. There needs to be more videos of what our sales and engineers are doing at these conferences and some tours behind the scenes of what the team do at HQ. In saying that, this video is great.
 
  • Like
  • Fire
  • Love
Reactions: 16 users
Although I disagree with John Maynard Keynes philosophy on government stimulus (but he certainly wouldn't have done it to the extreme levels that occurred during the GFC and Covid, which is much of the source of today's inflationary forces).

He was a staunch advocate of tariffs.


"According to Keynesian theory, trade deficits are harmful.The countries that import more than they export weaken their economies. When the trade deficit increases, unemployment rises and gross domestic product (GDP) slows down. Furthermore, surplus countries exert a "negative externality" on their trading partners. They get richer at the expense of others and destroy the output of their trading partners. John Maynard Keynes believed that the products of surplus countries should be taxed to avoid trade imbalances.

At the beginning of his career, Keynes was an economist close to Alfred Marshall, deeply convinced of the benefits of free trade. From the crisis of 1929 onwards, noting the commitment of the British authorities to defend the gold parity of the pound sterling and the rigidity of nominal wages, he gradually adhered to protectionist measures.

On 5 November 1929, when heard by the Macmillan Committee to bring the British economy out of the crisis, Keynes indicated that the introduction of tariffs on imports would help to rebalance the trade balance. The committee's report states in a section entitled "import control and export aid", that in an economy where there is not full employment, the introduction of tariffs can improve production and employment. Thus, the reduction of the trade deficit favours the country's growth.

In January 1930, in the Economic Advisory Council, Keynes proposed the introduction of a system of protection to reduce imports. In the autumn of 1930, he proposed a uniform tariff of 10% on all imports and subsidies of the same rate for all exports. In the Treatise on Money, published in the autumn of 1930, he took up the idea of tariffs or other trade restrictions with the aim of reducing the volume of imports and rebalancing the balance of trade.

On 7 March 1931, in the New Statesman and Nation, he wrote an article entitled Proposal for a Tariff Revenue. He pointed out that the reduction in wages led to a reduction in national demand which constrained markets. Instead, he proposed the idea of an expansionary policy combined with a tariff system to neutralise the effects on the balance of trade. The application of customs tariffs seemed to him "unavoidable, whoever the Chancellor of the Exchequer might be". Thus, for Keynes, an economic recovery policy is only fully effective if the trade deficit is eliminated. He proposed a 15% tax on manufactured and semi-manufactured goods and 5% on certain foodstuffs and raw materials, with others needed for exports exempted (wool, cotton).

In 1932, in an article entitled The Pro- and Anti-Tariffs, published in The Listener, he envisaged the protection of farmers and certain sectors such as the automobile and iron and steel industries, considering them indispensable to Britain."



Perhaps Trump isn't so "dumb" or "foolish" after all? 🤔...
 
Last edited:
  • Like
  • Haha
  • Thinking
Reactions: 12 users

7für7

Top 20
  • Like
  • Fire
Reactions: 3 users

Diogenese

Top 20
Thanks for posting. I watched the video and was impressed by Dr Lewis. Further, the mind boggles at what a 1 billion parameter model for LLM run on low power at the edge with Akida 2.0 can be used for. There needs to be more videos of what our sales and engineers are doing at these conferences and some tours behind the scenes of what the team do at HQ. In saying that, this video is great.
As Tony Lewis said, it can know a little about a lot or a lot about a little. The second option would be ideal for RAG (Retrieval Augmented Generation), where different models can be loaded depending on the required subject matter. The additional models could be downloaded or they could be stored on a co-located memory.
 
  • Like
  • Fire
  • Love
Reactions: 12 users
Top Bottom