BRN Discussion Ongoing

[cosors]

That may well be true, but you are missing my point that it is weird to say that Speck did not come up to Prophesee’s expectations, as that statement sounds (to me at least) as if they had been collaborating on it, whereas Speck was in fact developed with Prophesee’s competitor iniVation.

I have to correct myself in so far, though, as I just noticed that the Speck module is a combination of SynSense’s Dynap-CNN processor with an iniVation DVS camera and hence does not involve a newer processor.

As for iniVation being a current partner of Brainchip, too (beyond that 2016 highway monitoring video): I am very well aware of it, as I was actually the one who shared the info here on TSE that Nandan Nayampally let it slip during the CVPR 2023 Workshop on Event-Based Vision …

But what does this have to do with the fact that SynSense and BMW still seem to be collaborating?

I love(d) BMW more than any car company so far. I was just at the Pace Museum yesterday. On the other hand I don't know of any offensive that makes me confident for BMW about the drift in this sector.
Once it was. If the company BMW survives the next 10a and not just the brand I will be happy and surprised at the same time.

 

[Frangipani]

While we already know that a team from Technische Hochschule (University of Applied Sciences) Nürnberg successfully used Akida to win 2nd prize in this year’s tinyML Hackathon Challenge Pedestrian Detection…

… it is nice to see Brainchip also getting exposure on the institute’s website now (under ‘equipment’), alongside other development boards:

Ausstattung – Technische Hochschule Nürnberg Georg Simon Ohm

Durch die Größe der Fakultät Informatik, mit rund 1000 Studierenden, ist das Angebot vielfältig. Die Praxisorientierung des Studiums zeigt sich insbesondere bei Praxissemestern und Abschlussarbeiten in Kooperation mit den vielen großen und kleineren Unternehmen vor Ort.

www.th-nuernberg.de

FAKULTÄT INFORMATIK


/ Informatik / Forschung / Cognitive Neurocomputing /

Ausstattung​

PROPHESEE Metavision® EVK3 – VGA/HD Event-Based Neuromorphic Vision sensor

SynSense SpeckTM Event-driven neuromorphic visual computing SoC Development Kit

Brainchip Akida with Raspberry Pi Carrier Edge Device

KTH NCS PushBots with event-based cameras

nVidia Jetson TX1 Embedded Development Board

GiantAxon Lu.i Electronic neuron circuit boards

Arduino Tiny Machine Learning (TinyML) Development Kits

Manchester Robotics nVidia Jetson PuzzleBot

 

[TheDrooben]

Nice mention of Akida platform pages 15-16

Tech Briefs Magazine - October 2023 - Sensor-16

Experience our interactive, profoundly engaging digital publication!

www.nxtbook.com

Happy as Larry

 

[Frangipani]

Samsung to unveil 3D AI chip packaging tech SAINT to rival TSMC - KED Global

Samsung Electronics Co., the world’s largest memory chipmaker, plans to unveil an advanced three-dimensional (3D) chip packaging technology next year to

www.kedglobal.com

Korean chipmakers

Samsung to unveil 3D AI chip packaging tech SAINT to rival TSMC​

The race for 3D chip packaging tech is getting fierce as chip shrinking becomes increasingly challenging​

By Jeong-Soo Hwang
13 HOURS AGO
3 Min read

Samsung rapidly advances its 3D chip packaging tech
Samsung Electronics Co., the world’s largest memory chipmaker, plans to unveil an advanced three-dimensional (3D) chip packaging technology next year to compete with foundry leader Taiwan Semiconductor Manufacturing Company (TSMC).

The Suwon, South Korea-based chipmaker will use the technology – SAINT, or Samsung Advanced Interconnection Technology – to integrate memory and processors required of high-performance chips, including AI chips, in much smaller sizes.

Under the SAINT brand, Samsung plans to unveil three types of technologies – SAINT S, which vertically stacks SRAM memory chips and the CPU; SAINT D, which involves vertical packaging of processors such as the CPU and GPU and DRAM memory; and SAINT L, which stacks application processors (APs), people familiar with the matter said on Sunday.

The current 2.5D packaging tech, in most cases, assembles different types of chips side by side horizontally.

Some of Samsung’s new technologies, including SAINT S, have already passed validation tests, however, Samsung will launch commercial services next year after further tests with its clients, sources said.

Samsung's 2.5D H-Cube chip packaging solution

Packaging, one of the final steps in semiconductor manufacturing, places chips in a protective case to prevent corrosion and provides an interface to combine and connect already-made chips.

Leading chipmakers such as TSMC, Samsung and Intel Corp. are fiercely competing for advanced packaging, which integrates different semiconductors or vertically interconnects multiple chips. Advanced packaging allows multiple devices to be merged and packaged as a single electronic device.

Packaging technology can help enhance semiconductor performance without having to shrink the nanometer through ultra-fine processing, which is technologically challenging and requires more time.

According to consulting firm Yole Intelligence, the global advanced chip packaging market is forecast to grow from $44.3 billion in 2022 to $66 billion by 2027. Of $66 billion, 3D packaging is expected to account for about a quarter, or $15 billion.


TSMC, CURRENT 3D PACKAGING LEADER

The technology has been rapidly increasing in line with growth in generative AI such as ChatGPT, which requires semiconductors that can process large data quickly.

TSMC is the industry leader in chip packaging technology

The industry’s current mainstream is 2.5D packaging, which places chips as close together as possible to reduce data bottlenecks.

The world’s No. 1 contract chipmaker TSMC is also the leader in the global advanced packaging market with its decade-old 2.5D packaging technology.

TSMC is spending heavily to test and upgrade its 3D inter-chip stacking tech, SoIC, for its clients, including Apple Inc. and Nvidia Corp. TSMC said in July it will invest 90 billion Taiwanese dollars ($2.9 billion) in a new domestic advanced packaging plant.

Earlier this month, Taiwan's United Microelectronics Corp. (UMC), the world’s No. 3 foundry player, launched its wafer-to-wafer (W2W) 3D IC project to provide its clients with cutting-edge solutions for efficiently integrating memory and processors using silicon stacking technology.

Samsung's 2.5D H-Cube chip packaging solution

UMC said its W2W 3D IC project, in collaboration with packaging firms such as ASE, Winbond, Faraday and Cadence Design Systems, is an ambitious undertaking that seeks to leverage 3D chip integration technology to address the specific requirements of edge AI applications.

Intel uses its next-generation 3D chip packaging tech, Foveros, to make advanced chips.


SAMSUNG’S CHIP PACKAGING ROADMAP

Samsung, the world’s No. 2 foundry company, has been accelerating the development of its chip packaging technology since it unveiled its 2.5D packaging technology H-Cube in 2021.

Samsung Chairman Jay Y. Lee (third from left) visits a Samsung chip packaging line in Korea on Feb. 17, 2023

The 2.5D packaging tech allows logic chips or high-bandwidth memory (HBM) to be stacked on top of a silicon interposer in a small form factor, Samsung said.


The Korean company said in April it is offering a packaging turnkey service, handling the entire process from chip production to packaging and tests.

With its new SAINT technology, Samsung aims to improve the performance of AI chips for data centers and mobile APs with on-device AI functions, sources said.

Write to Jeong-Soo Hwang at hjs@hankyung.com
In-Soo Nam edited this article.

 

[Frank Zappa]

View attachment 49311

Magnus may still mention Akida when discussing voice recognition as he can obviously see that it is a highly sort after topic for him to include .We can only wait and see .

 

[Diogenese]

Funny that ole Speck chip which ended in toys seems to be good enough for BMW:


View attachment 49343

'sfunny they don't mention lidar.

https://www.synsense.ai/synsense-cars-that-think-like-you-eetimes/

SynSense is working with BMW to advance the integration of neuromorphic chips into smart cockpits and explore related fields together. BMW will be evaluating SynSense’s Speck SoC, which combines SynSense’s neuromorphic vision processor with a 128 x 128–pixel event–based camera from Inivation. It can be used to capture real–time visual information, recognize and detect objects, and perform other vision–based detection and interaction functions.

 

[Diogenese]

I agree with you from another angle. I am close to the certification business and have worked through a lot of requirement catalogues. The certification body is the subject of the evaluation and then to a certification as a whole. Imagine a device like a terminal, for example. The device as a whole is certified and not each individual component, input, circuit board, chip, software, etc.
I reckon it's the same here. Anyone who has the standards at their disposal could check. The auditor also has a certain degree of freedom of interpretation. A chip, a circuit board, a system, a device is full of different IPs. It is up to the the company who wants to sell it on a market with regulations to have the device evaluated and certified or to commission this.

I can imagine at one point that a seller of modules for eg the automotive sector who wants to have its device certified approaches Brainchip with some questions from the catalogue and asks how Brainchip fulfils this or that requirement. The company then takes this information and incorporates it into its own ~answer catalogue, if this is not already known and supplied to the customer.
Sometimes you can simplify the work instead of answering such questions every time. And you can do this by certify yourselve to a certain degree. Then only a reference to the binding certificate is necessary. However, this involves a lot of work and costs and raises the question of whether it makes sense. Brainchip does not sell any module or device on the respective market, e.g. US or EU.
That's how I interpret it and that's probably why he didn't get an answer to his question because someone rolled their eyes.
The certification business is complicated. Especially describing it to someone from the outside.

At this moment, for the first time, I see this as one next clear advantage that Brainchip 'only' sells IP, no device, system, module and not its own chip for the respective sector to be certified.

But that's just my guess. As I said, complicated it is.

Valeo has that work.

Thanks cosors,

Akida uses stock-standard CMOS manufacturing so there's no real issues with the hardware side.

Also, Akida 1 does not use any processor based software while performing classification/inference, so again, no issue with ISO certification on this point.

One area which would need to be examined for ISO certification is the accuracy of the results of the hardware working with the image model, and related to this is the use of a processor to load the model and weights and configure the NN layers.

Akida 2 does make some use of the processor for its more advanced functionality, so again some additional scrutiny would be required for ISO.

 

[Bravo]

That may well be true, but you are missing my point that it is weird to say that Speck did not come up to Prophesee’s expectations, as that statement sounds (to me at least) as if they had been collaborating on it, whereas Speck was in fact developed with Prophesee’s competitor iniVation.

I have to correct myself in so far, though, as I just noticed that the Speck module is a combination of SynSense’s Dynap-CNN processor with an iniVation DVS camera and hence does not involve a newer processor.

As for iniVation being a current partner of Brainchip, too (beyond that 2016 highway monitoring video): I am very well aware of it, as I was actually the one who shared the info here on TSE that Nandan Nayampally let it slip during the CVPR 2023 Workshop on Event-Based Vision …

But what does this have to do with the fact that SynSense and BMW still seem to be collaborating?

AKIDA is much betterer.

 

[Bravo]

1 Article dated 10 November 2023

Foxconn, Nvidia joining hands to forge AI industrial revolution​

Taiwan, US tech giants in ambitious tie-up that aims to define the AI future and leave chip-starved Chinese competitors in its dust
By SCOTT FOSTER NOVEMBER 10, 2023

Nvidia and Foxconn have shared AI dreams. Image: Facebook Screengrab

Taiwan’s Hon Hai Technology Group, widely known as Foxconn, and America’s Nvidia plan to join technological forces “to accelerate the AI industrial revolution.” Foxconn, the world’s largest contract manufacturer, is the primary assembler of the Apple iPhone. California-based Nvidia is the world’s leading designer of the graphics processing units (GPUs) used in artificial intelligence applications.

Foxconn, Nvidia joining hands to forge AI industrial revolution - Asia Times

Taiwan’s Hon Hai Technology Group, widely known as Foxconn, and America’s Nvidia plan to join technological forces “to accelerate the AI industrial

asiatimes.com

2. Article dated 18 October 2023

Foxconn and Nvidia are building 'AI factories' to accelerate self-driving cars | TechCrunch

Nvidia and Foxconn are working together to build so-called "AI factories," a new class of data centers that promise to provide supercomputing powers to

techcrunch.com

3 This post from January 2023

 

[Diogenese]

Hi @Diogenese,

I was going to say something along the same lines as @Perhaps.

FYI, SynSense’s website still lists BMW as a partner under ‘Industry Ecology’.

View attachment 49344



I am somewhat puzzled you are claiming Speck did not meet Prophesee’s expectations, as SynSense developed Speck with its sister start-up iniVation, not with Prophesee. Are you by any chance mixing up Speck and Dynap-CNN? This was the SynSense processor used in the 2021 collaboration with Prophesee.

SynSense and Prophesee develop one-chip event-based smart sensing solution

Partnership leverages respective leadership in neuromorphic computing and sensing to realize on-sensor processing integration leading to small form-factor, low-power and cost-effective IoT solutions

www.prophesee.ai

Prophesee continues to be listed as a partner both under ‘Industry Ecology’ (see above) and under ‘Academia cooperation’, by the way.


View attachment 49346

And both SynSense and Brainchip are listed as partners on Prophesee’s website.

View attachment 49347


According to this article dated March 30, 2023

SynSense secures $10M to advance low-power vision processing SoC to mass production | Biometric Update

The company has developed ultra-low-power AI chips, driven by its goal to enable neuromorphic intelligence at the network edge.

www.biometricupdate.com

… SynSense were still working with BMW at the time.

“Last year, SynSense announced a partnership with BMW to explore the integration of their brain-like chips into smart cockpits. This was SynSense’s first venture into the field of smart cockpits.

The company says Speck enables real-time visual information capturing, object recognition and detection, and other vision-based detection and interaction functions for BMW.”

This was published a week before the podcast you linked (about Speck being used in a toy) was released.

And why would the fact that Speck is used in a toy preclude it from being used in the automotive sector as well? How would that be different from Akida being not only used in future MB cars, but also in smart doorbells, hearing aids, wearables etc one (hopefully not too far away) day?

Mind you, I am not saying that Akida wouldn’t be the better choice, but I have yet to see any evidence that BMW and SynSense have since ended their collaboration.

Hi Frangipanni,

You are right. I did use the term Speck loosely. Speck is the combination of Dynap-CNN with Inivation's 128*128 pixel DVS. Clearly 128*128 pixels is far short of the detail required for lidar.

"SynSense had already integrated its Dynap-CNN processor with a dynamic vision sensor – or DVS sensor – from another company called IniVation, in a module known as Speck."

Dylan Muir said in the interview that Speck lacked the accuracy to be used in automotive so they targeted the "low hanging fruit".
When asked by Sally Ward-Foxton about the Synsense/prophesee project, Muir was less than effusive:

"SALLY WARD-FOXTON

Last time we spoke, which admittedly was a while ago, you had partnered with Prophesee, also. How is that collaboration going?

DYLAN MUIR

That’s going very well. We’ve fabricated a device with Prophesee, and this is… we’re testing this, examining this at the moment, in conjunction with them.

SALLY WARD-FOXTON

OK, to be clear, it uses a different processor that you’ve made compared to the Speck module – or the same?

DYLAN MUIR

The processor IP is basically the same, same design, that’s our processor design for Speck. These little DYNAP-CNN cores and that’s what we apply for CNN processing. So we’ve also looked at, for example, an RGB standard CMOS camera interface which we could then also process using the same spiking CNN architecture. We do get asked from customers for quite often about having an RGB standard CMOS interface."

After having tried SynSense, Luca Verre of Prophesee acknowledged that Akida was better able to match the performance of their DVS:

"By combining our Metavision solution with Akida-based IP, we are better able to deliver a complete high-performance and ultra-low power solution to OEMs looking to leverage edge-based visual technologies as part of their product offerings, said Luca Verre, CEO and co-founder of Prophesee."

Given that the discussion was about Valeo lidar, the reference to exploratory investigations with BMW for gesture recognition and other in-cabin applications is a diversionary red herring.

https://www.synsense.ai/synsense-cars-that-think-like-you-eetimes/

SynSense is working with BMW to advance the integration of neuromorphic chips into smart cockpits and explore related fields together. BMW will be evaluating SynSense’s Speck SoC, which combines SynSense’s neuromorphic vision processor with a 128 x 128–pixel event–based camera from Inivation. It can be used to capture real–time visual information, recognize and detect objects, and perform other vision–based detection and interaction functions. 202208220

 

[Diogenese]

Because in toys, it doesn’t have to be accurate 100% of the time whereas in Automotive it has to be accurate all the time, that is what was said by the Synsense employee on that podcast @Diogenese shared.
Automotive goes through at least a 12 month safety check through the regulators and the like.
Nanden has also said back in July 2023 that Inivation was a partner of theirs!

Prophesee also said in a podcast with Brainchip that it wasn’t until Prophesee came across Brainchip that they were able to tell a full story as they kept on hitting bottlenecks.

Hi Chapman,

This is a good example of where @Zeebot needs to introduce a prospective long skip feature so we can see ahead to all the responses to a post someone is about to respond to - could have saved me an hour if I'd seen your and Bravo's responses.

 

[Tothemoon24]

 

[Glen]

Automotive Custom SoC Technologies and Solutions

Socionext’s advanced automotive custom SoC solutions are designed to help OEMs and tier-one automakers achieve differentiating technologies

socionextus.com

 

[Quiltman]

Automotive Custom SoC Technologies and Solutions

Socionext’s advanced automotive custom SoC solutions are designed to help OEMs and tier-one automakers achieve differentiating technologies

socionextus.com

Nice reminder.

It's like as a collective group we forget important partnerships continuing to develop in the background.
If there are no sound bites to feed the masses the relationship to BrainChip tends to drift away from the collective consciousness.

 

[McHale]

A New Era of Galaxy AI is Coming — Here’s a Glimpse

samsungmobilepress.com

How Samsung Electronics uses #GAI to handle data on phones | Patrick Moorhead posted on the topic | LinkedIn

You knew that Samsung Electronics would get into the #GAI game with devices, especially with all the on-device AI acceleration coming from chip companies like…

www.linkedin.com

For a tiny glimpse of one benefit Galaxy AI will enable, look no further than the very thing our phones were originally created to do, calling each other. AI Live Translate Call will soon give users with the latest Galaxy AI phone a personal translator whenever they need it. Because it’s integrated into the native call feature, the hassle of having to use third-party apps is gone.

Audio and text translations will appear in real-time as you speak, making calling someone who speaks another language about as simple as turning on closed captions when you stream a show. Because it’s on-device Galaxy AI, you can trust that no matter the scenario, private conversations never leave your phone.

Coming early next year, Galaxy AI will bring us closer to a world where some of the most common barriers to social connection dissolve, and communication is easier and more productive than ever.

That’s AI changing the world and your life for the better.

View attachment 49273

Well @IlioveLamp, I just listened to the recent interview for a second time, because I wanted to be sure about all this. It is the interview Sean did with Mark Kennis, and Sean said on two occasions in the interview that the most interest from potential customers with regard to adoption of akida 2 was coming from the following industries: 1. hearing aids, 2. industrial use robotics, vibration analysis etc, 3. wearables and 4. the automotive sector (for a range of applications).

Mark asked a specific question about where the interest in Akida 2 was coming from, so Seans response was of great interest to me, but as I said Sean repeated the same 4 sectors again later in the interview.

There were a number of takeaways that were encouraging, Sean said up to approx 75% of customer interest was in adoption of Akida 2, that customers had been incredibly positive about Akida 2, and (most importantly IMO) that a good number of these potential customers were very serious engagements where the customer had to undertake evaluations, had plans to adopt Akida and production schedules they envisaged - this all means the customer is spending considerable amounts of money to further the planned adoption of Akida.

Sean also said that he felt there would be new contracts forthcoming, but that 2024 would be make or break for BRN, so of course you would expect your CEO to be upbeat - that is his job after all - I hope he can deliver.

However at no point did Sean make any mention of mobile phones, of course we all know if BRN could crack into that sector it would make a massive impact on S/P, so I follow all the speculation here (for years) about connections or possible connections to phone makers and players related to phones. So maybe Sean wants to keep possibilities there under his hat, or maybe he thinks talk about connections to mobile phones would generate too much hype which could become problematic.

I speculate, but regardless, no talk about phones, but having said that there are any number of other things Sean didn't speak to in the interview either.

 

[Esq.111]

Afternoon Chippers ,

Some large Sell orders floating through, .......Fuck i'm looking forward to a Big entity with a long portfolio stepping in ,...Start wiping the floor with these bastards.

Regards,
Esq.

 

[mrgds]

Afternoon Chippers ,

Some large Sell orders floating through, .......Fuck i'm looking forward to a Big entity with a long portfolio stepping in ,...Start wiping the floor with these bastards.

Regards,
Esq.

Gday @Esq.111
I may be pissing into the wind, ...................... but, i still can't understand the reasoning why the Q3 Shareholder Podcast has been delayed due to
"scheduling issues". Can anyone here please let me know if a "recorded audio podcast" can have scheduling issues?

Also the "un-announced investor presentation".

Im thinking there will be a "price sensitive ann" this week, whether that be positive or not, i don't know.

Anyhoo, am waiting on TDs reply to these questions.

 

[TECH]

Gday @Esq.111
I may be pissing into the wind, ...................... but, i still can't understand the reasoning why the Q3 Shareholder Podcast has been delayed due to
"scheduling issues". Can anyone here please let me know if a "recorded audio podcast" can have scheduling issues?

Also the "un-announced investor presentation".

Im thinking there will be a "price sensitive ann" this week, whether that be positive or not, i don't know.

Anyhoo, am waiting on TDs reply to these questions.

Welcome to Seoul Sean....great that you travelled to meet with us in person, our Board prefers to do business face-face...now lets sit down and
discuss the future direction we "could" pursue together, I see that you have brought AKD 2.0 with you, fantastic Sean, the floor is all yours.

Tech thinking positive as always...

 

[mrgds]

Welcome to Seoul Sean....great that you travelled to meet with us in person, our Board prefers to do business face-face...now lets sit down and
discuss the future direction we "could" pursue together, I see that you have brought AKD 2.0 with you, fantastic Sean, the floor is all yours.

Tech thinking positive as always...

Hi @TECH
I, like all here, hope you are spot on.

 

[mrgds]

Welcome to Seoul Sean....great that you travelled to meet with us in person, our Board prefers to do business face-face...now lets sit down and
discuss the future direction we "could" pursue together, I see that you have brought AKD 2.0 with you, fantastic Sean, the floor is all yours.

Tech thinking positive as always...

@TECH ...........Sorry if ive missed this, but, is Sean definately travelling to Seoul? Or was that tongue in cheek?
And if so when?
Cheers