BRN Discussion Ongoing

[Perhaps]

TSE is a reliable archive.
View attachment 49395

View attachment 49396


Jan 5, 2023,
Now, at CES 2023 for those who don't want to believe it like I did back then.

Top secret!
BMW has outsourced the design department a while ago:

But to be true to the competition, when I see something like this for 2024, I feel like living in a time capsule:

 

[Fullmoonfever]

Nah...Sean gonna drop in and catch up with SK Hynix and chew the fat on their neuromorphic AI thoughts.

I wish

Become a Semiconductor Expert with SK hynix – AI Semiconductors

As AI technology and services continue to rapidly develop, SK hynix’s AI semiconductors will also evolve to meet the market and consumer needs.

news.skhynix.com

[We Do Future Technology] Become a Semiconductor Expert with SK hynix – AI Semiconductors​

May 9, 2023

(Snippet)....

Ultimately, AI semiconductors need to combine the functions of a processor and memory while providing more enhanced qualities than the Von Neumann architecture. SAPEON Korea, an AI startup jointly founded by SK hynix, recently developed an AI semiconductor for data centers named after the company. The SAPEON AI processor offers a deep learning computation speed which is 1.5 times faster than that of conventional GPUs and uses 80% less power. In the future, SAPEON will be expanded to other areas like autonomous cars and mobile devices. SK hynix’s commitment to developing technologies to support AI is further highlighted by its establishment of the SK ICT Alliance alongside SK Telecom and SK Square.

The alliance invests and develops in diverse ICT areas such as semiconductors and AI to secure global competitiveness. Furthermore, SK hynix is also developing a next-generation CIM3 with neuromorphic semiconductor4 devices

SK hynix Commercializes World’s Fastest Mobile DRAM LPDDR5T

SK hynix Inc. announced that it has started supplying its customers with 16 GB packages of LPDDR5T, the world's fastest mobile DRAM.

news.skhynix.com

SK hynix Commercializes World’s Fastest Mobile DRAM LPDDR5T​

November 13, 2023

News Highlights​

• Starts supplying 16 GB mobile DRAM package to global smartphone makers
• LPDDR5T to be adopted in the latest smartphones along with MediaTek Dimensity 9300 mobile processor
• “Company to continue developing high-performance DRAM products to bring ‘On-Device AI’ technology to smartphones”

Seoul, November 13, 2023​

SK hynix Inc. (or “the company”, www.skhynix.com) announced today that it has started supplying its customers with 16 gigabyte (GB) packages of Low Power Double Data Rate 5 Turbo (LPDDR5T)1, the fastest mobile DRAM available today that can transfer 9.6 gigabits per second (Gbps).

1LPDDR: Low power DRAM for mobile devices, including smartphones and tablets, which aims to minimize power consumption and features low voltage operation. The latest specifications are for the 7th generation, succeeding the series that end with 1, 2, 3, 4, 4X, 5 and 5X. LPDDR5T is a newly developed version by SK hynix, and an upgraded product of the 7th generation (5X), prior to the development of the 8th generation LPDDR6.

Since the successful development of its LPDDR5T in January, SK hynix has been preparing to commercialize the product by conducting performance verification with global mobile application processor (AP)2 manufacturers.

2Mobile AP: A semiconductor chip that serves as the ‘brain’ of mobile devices, such as smartphones or tablets. It is also the central processing unit in a system-on-chip (SoC) format that contains a complete set of products and systems, including computation, graphics, digital signal processing, etc., in a single chip.

SK hynix explained that LPDDR5T is the optimal memory to maximize the performance of smartphones, with the highest speed ever achieved. The company also emphasized that it would continue to expand the application range of this product and lead the generation shift in the mobile DRAM sector.

The LPDDR5T 16 GB package operates in the ultra-low voltage range of 1.01 to 1.12V set by the Joint Electron Device Engineering Council (JEDEC), and can process 77 GB of data per second, which is equivalent to transferring 15 full high-definition (FHD) movies in one second.

The company recently started shipping its product to a global smartphone manufacturer, Vivo, which also announced that its latest smartphones, X100 and X100 Pro, will be equipped with SK hynix’s up-to-date memory packages.

These devices will also be packed with MediaTek’s newest flagship mobile AP ‘Dimensity 9300’. In August, SK hynix confirmed that it has completed the performance verification for application with MediaTek’s next-generation mobile APs.

“Smartphones are becoming essential devices for implementing On-Device AI3 technology as the AI era kicks into full swing,” said Myoungsoo Park, Vice President and Head of DRAM Marketing at SK hynix. “There is a growing demand for high-performing, high-capacity mobile DRAMs in the market.”

3On-Device AI: Artificial intelligence technology that implements AI capabilities on the device itself, instead of going through computation in a physically separated server. AI functions can become more responsive and provide more customized AI services as smart devices are capable of collecting and computing information on its own.

“We will continue to lead the premium DRAM market based on our technological leadership in AI memories, while staying in tune with market demands,” Park added

 

[Bravo]

An addition:
It's a very simplified olympic idea to think the more neurons the better. Depending on the application there's always an right amount of neurons, overpowering leads to overly complex training algorithms needed and can have a negative effect on accuracy and latency.
So a scalable amount of neurons might be the way to go:

View attachment 49392

How many neurons for a neural network? | Your Data Teacher

Neural networks are a fascinating field of machine learning. Let's see how to find the best number of neurons of a neural network for our dataset.

www.yourdatateacher.com

View attachment 49393

IEEE Xplore Full-Text PDF:

ieeexplore.ieee.org

View attachment 49394

https://hal.science/hal-01633418/document

Well in that case you better get on the blower ASAP to Edge Impulse and tell them they’re backing the wrong horse and see what they have to say! At the moment they’re happily endorsing our IP along with Arm’s to their customers. But I’m sure they’ll see the error of their ways once you contact them and set them straight. He-he-he! Neigh!

 

[Perhaps]

Well in that case you better get on the blower ASAP to Edge Impulse and tell them they’re backing the wrong horse and see what they have to say! At the moment they’re happily endorsing our IP along with Arm’s to their customers. But I’m sure they’ll see the error of their ways once you contact them and set them straight. He-he-he! Neigh!

Look at the financials.

 

[cosors]

A look at the models for the volume market leads to one question, who in the world will pay such completely crazy prices for design accidents?
Chinese car industry enters massively the European market, right now it looks like a change in technological leadership occurs.

No one

"VW and BMW react to electric flops in China - and take a drastic step
...
Both Volkswagen and BMW have now drastically reduced the prices of their electric models in China:
VW offers EVs in China several thousand euros cheaper"

VW und BMW reagieren auf Elektro-Flops in China – und gehen drastischen Schritt

Deutsche Autobauer haben Mühe, auf dem größten Absatzmarkt der Welt ihre E-Autos abzusetzen. VW und BMW sehen sich daher zu einer durchgreifenden Maßnahme gezwungen.

www.hna.de

 

[Bravo]

Look at the financials.

I’m too busy looking at who’s listed below Arm on the bottom right hand corner, as suggested IP for customers such as Infineon, Renesas, NXP, Qualcomm and NVIDIA and it ain’t Synsense I can assure you.

No offence if you work at Synsense because I’m sure there’s going to be enough room for a few players and there is certainly a demand for toys so there will be a market for that product I’m sure.

 

[cosors]

Top secret!
BMW has outsourced the design department a while ago:

View attachment 49398

But to be true to the competition, when I see something like this for 2024, I feel like living in a time capsule:

View attachment 49407

 

[Perhaps]

I’m too busy looking at who’s listed below Arm on the bottom right hand corner, as suggested IP for customers such as Infineon, Renesas, NXP, Qualcomm and NVIDIA and it ain’t Synsense I can assure you.

No offence if you work at Synsense because I’m sure there’s going to be enough room for a few players and there is certainly a demand for toys so there will be a market for that product I’m sure.



View attachment 49409

You really don't understand my posts. But hey, who are all those researchers of the most respected institutions in the world against some positive thinking. I'm talking about challenges which explain the delays in commercialisation, not more, not less.

 

[Perhaps]

View attachment 49410

More and more I come to the conclusion I will drive my 1996 Merc 'til I die.

 

[Dallas]

https://www.linkedin.com/posts/bluw...m_source=share&utm_medium=member_androidF Anil Mankar

 

[Dallas]

Sign Up | LinkedIn

500 million+ members | Manage your professional identity. Build and engage with your professional network. Access knowledge, insights and opportunities.

www.linkedin.com

 

[Dallas]

Sign Up | LinkedIn

500 million+ members | Manage your professional identity. Build and engage with your professional network. Access knowledge, insights and opportunities.

www.linkedin.com

 

[cosors]

https://www.linkedin.com/posts/bluwireless_motorsport-connectivity-telemetry-activity-7129763337759768576-fjev?utm_source=share&utm_medium=member_androidF Anil Mankar

"This solution has recently been tested in Europe and North America, ready for imminent deployment in motorsport racing series."

Anil Mankar
CO-Founder and CDO BrainChip Inc
38m

"Congrates Mark Barrett"

 

[Bravo]

You really don't understand my posts. But hey, who are all those researchers of the most respected institutions in the world against some positive thinking. I'm talking about challenges which explain the delays in commercialisation, not more, not less.

Ok. Maybe I don’t understand your posts and I’m sorry about that and I’ll try harder next time but also I don’t think you understand some other peoples posts, like @Diogenese posts for example, who is an extremely generous contributor here, whose knowledge and insight is invaluable and to whom you were actually quite a rude git to, incase you didn’t know. And that’s not cool because it makes people not want to participate and then we all lose as a consequence.

 

[cosors]

https://www.linkedin.com/posts/bluwireless_motorsport-connectivity-telemetry-activity-7129763337759768576-fjev?utm_source=share&utm_medium=member_androidF Anil Mankar

I can't see the link yet, but I can hardly imagine Anil liking it out of purely private interest.
Unqualified statement. I apologise.

The forum here is full of links, most posts from FF.

chapman89 :

"Have we looked at a company called BluWireless before?
https://www.bluwireless.com/

Blu Wireless offers leading mmWave wireless technology to empower our partners to innovate through 5G connectivity.

Brainchip staff regularly like their posts and Sean Hehir has said they we’re working closely with a Telecommunications company, and just in the last 24 hours Anil Mankar has congratulated one of their staff.

Renesas is one of BluWireless partners.

A snippet here about BluWireless-

“Blu Wireless is disrupting the market with our ultra-fast and seamless 5G mmWave solutions. We are committed to developing cutting-edge network technology that helps set our customers apart as market leaders and innovators”"

https://thestockexchange.com.au/threads/brn-discussion-ongoing.1/post-168352


and this:

https://www.tessolve.com/board-of-directors/

But the products McLaren offers are really interesting! I didn't know that. I only knew about the sports cars and the fantastic red wine from near Adelaide.
https://www.mclarenapplied.com/products

 

[Perhaps]

Ok. Maybe I don’t understand your posts and I’m sorry about that and I’ll try harder next time but also I don’t think you understand some other peoples posts, like @Diogenese posts for example, who is an extremely generous contributor here, whose knowledge and insight is invaluable and to whom you were actually quite a rude git to, incase you didn’t know. And that’s not cool because it makes people not want to participate and then we all lose as a consequence.

When we start to discuss rudeness, we should talk about the reactions on posts of people, who are just tired of those everlasting speculations and dot joining.

 

[jtardif999]

An addition:
It's a very simplified olympic idea to think the more neurons the better. Depending on the application there's always an right amount of neurons, overpowering leads to overly complex training algorithms needed and can have a negative effect on accuracy and latency.
So a scalable amount of neurons might be the way to go:

View attachment 49392

How many neurons for a neural network? | Your Data Teacher

Neural networks are a fascinating field of machine learning. Let's see how to find the best number of neurons of a neural network for our dataset.

www.yourdatateacher.com

View attachment 49393

IEEE Xplore Full-Text PDF:

ieeexplore.ieee.org

View attachment 49394

https://hal.science/hal-01633418/document

I think this description is referring to analogue SNNs which are inherently unstable and unreliable when compared to digital SNNs. Akida is a fully digital SNN architecture built on CMOS logic - making for very repeatable, reliable behaviour. AIMO.

 

[Diogenese]

While we may have every reason to believe that Akida is the superior technology, we should not resort to making factually incorrect statements about Brainchip’s competitors:



True, automotive takes a long time, but no, they are not starting to explore only now - they announced their partnership in mid-April 2022, so that was more than 1.5 years ago.

Also, the announcement about the SynSense-BMW partnership was primarily about smart cockpits and not about LiDAR only. That’s why I compared it to MB (‘Hey, Mercedes!’) and not to Valeo.

While the article I quoted (which I had picked as one of the most recent ones I could find about SynSense + BMW) does indeed literally say “to explore the integration of their brain-like chips into smart cockpits”, I have since double-checked the news around the original press release (which strangely I cannot find anywhere) that consistently quote the partnership announcement being about advancing the integration of neuromorphic chips into smart cockpits and exploring related fields with BMW. (See for example Sally Ward-Foxton referring to the partnership in her “Cars That Think Like You” EETimes article)

https://www.eetimes.com/cars-that-think-like-you/#:~:text=Car%20makers%20are%20checking%20out,attention%2C%20and%20passenger%20behavior%20monitoring.

That’s also the wording I recall. So it sounds as if they had in fact already collaborated quite a bit regarding smart cockpits before announcing their partnership, similar to Valeo and Brainchip having been involved months before announcing their partnership.



Also not true.
“SynSense focuses on the commercialization of neuromorphic intelligence, based on 20+ years of world-leading experience of University of Zürich and ETH Zürich. Centering on edge computing applications, SynSense provides ultra-low power consumption, ultra-low latency inference ASICs and IP blocks, as well as full-stack application development services, and is globally the only neuromorphic technology company that involves both sensing and computing.”

About Us | SynSense

SynSense focuses on neuromorphic intelligence and takes the lead in commercialization of neuromorphic chips; SynSense provides application solutions and services.

www.synsense.ai

Again, factually incorrect: You are comparing apples and oranges. Your juxtaposition probably refers to some other SynSense chip, but definitely not to the DYNAP-CNN (which is the one involved in the partnership with Prophesee), which according to their website has up to 1 million neurons. No mention of synapses, I am afraid. But up to 1000 x more neurons than you claimed!

DYNAP™-CNN: Neuromorphic Processor With 1M Spiking eurons

DYNAP™-CNN is world's first fully scalable, event-driven neuromorphic processor with 1M configurable spiking neurons and direct interface with external DVS.

www.synsense.ai

“DYNAP™-CNN, a fully scalable, event-driven neuromorphic processor, redefines sensory processing. With up to 1M configurable spiking neurons and a direct interface with external DVS, this technology is ideal for ultra-low power, ultra-low latency event-driven applications.”

It is not my intention to spruik a competitor’s product here, but when making statements about competitors or their products, we - me included, of course - should stick to the facts and possibly fact-check before posting.
And also accept the reality that Brainchip will never have 100 % of the market share.

On a more positive note: BMW having partnered with SynSense in 2022 does not automatically exclude BMW also exploring things with Brainchip, if you ask me. Right now (behind an NDA) or in the future.

Now the discussion about whether Scala 3 includes Akida has been totally sidetracked and we're splitting hairs about Synsense.

Spec (DYNAP-CNN+Inivation 128*128 pixel DVS) has a response time of 50 ms,. Prophesee processes 10,000 fps equivalent, so it's not surprising DYNAP-CNN struggled to keep up with Prophesee.

There is also Dynap SEL, Dynap SE2, and XYLO.

XYLO is a chip of a different colour (audio).
https://www.synsense.ai/products/xylo/


DYNAP™-SEL

Our next-generation neuromorphic chip, DYNAP™-SEL, features 1k analog low-power spiking neurons and up to 80k configurable synaptic connections, including 8k synapses with integrated spike-based learning rules.

The fan-in and fan-out capabilities of DYNAP™-SEL are comparable to the connectivity architectures observed in mammalian cerebral cortex, and permit networks with biologically realistic connectivity to be emulated using DYNAP™-SEL. Our technology enables implementation of spiking neural networks with on-chip learning and large fan-in/out network connectivity at an extremely low power-budget in real-time on DYNAP™-SEL.

DYNAP™-SE2

Our latest SE family neuromorphic chip DYNAP™-SE2 combines the impeccable energy efficiency of its previous generations with new features for low latency, real-time end-to-end applications using general purpose feed-forward, recurrent and reservoir networks.

Its integrated bio-signal amplifiers make the DYNAP™-SE2 a perfect fit for mobile health and robotic applications. Each chip features 1k redesigned adaptive exponential integrate-and-fire analog ultra low-power spiking neurons and 65k enhanced synapses with configurable delay, weight and short term plasticity. The innovative asynchronous low latency communication infrastructure (patented) enables each neuron to communicate with up to 230k surrounding neurons and infinite scalability via relay neurons for truly large scale networks.

Doigeeknees Sitter I for one appreciate your, at times, layman posts.

Agree. Sometimes I don't understand my posts myself.

 

[Diogenese]

Now the discussion about whether Scala 3 includes Akida has been totally sidetracked and we're splitting hairs about Synsense.

Spec (DYNAP-CNN+Inivation 128*128 pixel DVS) has a response time of 50 ms,. Prophesee processes 10,000 fps equivalent, so it's not surprising DYNAP-CNN struggled to keep up with Prophesee.

There is also Dynap SEL, Dynap SE2, and XYLO.

XYLO is a chip of a different colour (audio).
https://www.synsense.ai/products/xylo/


DYNAP™-SEL

Our next-generation neuromorphic chip, DYNAP™-SEL, features 1k analog low-power spiking neurons and up to 80k configurable synaptic connections, including 8k synapses with integrated spike-based learning rules.

The fan-in and fan-out capabilities of DYNAP™-SEL are comparable to the connectivity architectures observed in mammalian cerebral cortex, and permit networks with biologically realistic connectivity to be emulated using DYNAP™-SEL. Our technology enables implementation of spiking neural networks with on-chip learning and large fan-in/out network connectivity at an extremely low power-budget in real-time on DYNAP™-SEL.

DYNAP™-SE2

Our latest SE family neuromorphic chip DYNAP™-SE2 combines the impeccable energy efficiency of its previous generations with new features for low latency, real-time end-to-end applications using general purpose feed-forward, recurrent and reservoir networks.

Its integrated bio-signal amplifiers make the DYNAP™-SE2 a perfect fit for mobile health and robotic applications. Each chip features 1k redesigned adaptive exponential integrate-and-fire analog ultra low-power spiking neurons and 65k enhanced synapses with configurable delay, weight and short term plasticity. The innovative asynchronous low latency communication infrastructure (patented) enables each neuron to communicate with up to 230k surrounding neurons and infinite scalability via relay neurons for truly large scale networks.

Agree. Sometimes I don't understand my posts myself.

Getting back to the Valeo discussion after my computer had a wardrobe malfunction*, the point is, Synsense do not promote any of their processors as being suitable for lidar.

* If you copy text from a document, you can paste it into Word as "Unformatted Text" by using "Alt E S'', but if you do it in TSEx (or email) it closes the application.

 

[Perhaps]

I think this description is referring to analogue SNNs which are inherently unstable and unreliable when compared to digital SNNs. Akida is a fully digital SNN architecture built on CMOS logic - making for very repeatable, reliable behaviour. AIMO.

The challenge is more about NNs in general and the changing amount of neurons needed to deliver the best results depending on the tasks.