BRN Discussion Ongoing

[Perhaps]

Nvidia use 16-bit or 32-bit weights/actuations.

3rd Generation Tensor Cores and Sparsity NVIDIA Tensor cores provide the performance necessary to accelerate next generation AI applications. Tensor cores are programmable fused matrix-multiply-and-accumulate units that execute concurrently alongside the CUDA cores. Tensor cores implement floating point HMMA (HalfPrecision Matrix Multiply and Accumulate) and IMMA (Integer Matrix Multiple and Accumulate) instructions for accelerating dense linear algebra computations, signal processing, and deep learning inference


They do try to fudge a bit of sparsity by ignoring zeros if 2 out of every 4 bits are non-zero (or do they mean if 2 out of every 4 bits are zero?). Anyhow, they have to count the bits (fine-tuning weights), and add the address (value) of the non-zero bits (non-zero indices, ie, is its value 1, 2, 4, 8, 16, 32, ...?). Rather inelegant!

View attachment 10272


https://developer.nvidia.com/blog/nvidia-jetson-agx-xavier-32-teraops-ai-robotics/

The latest generation of NVIDIA’s industry-leading Jetson AGX family of embedded Linux high-performance computers, Jetson AGX Xavier delivers GPU workstation class performance with an unparalleled 32 TeraOPS (TOPS) of peak compute and 750Gbps of high-speed I/O in a compact 100x87mm form-factor. Users can configure operating modes at 10W, 15W, and 30W as needed for their applications. Jetson AGX Xavier sets a new bar for compute density, energy efficiency, and AI inferencing capabilities deployable to the edge, enabling next-level intelligent machines with end-to-end autonomous capabilities.

View attachment 10274

There is an opportunity to plug Akida in in place of the Vision Accelerator in the green Xavier SoC box - but they have Watts to burn.

So, on the disclosed information ...

View attachment 10275

Xavier is the old platform, build in actual Mercedes models. Thought there might be some progress in the new Orin architecture, as it is much faster with less power consumption..

 

[MDhere]

Seems Magik-eye is still alive and well, apologies if already posted but found this Japanese article about Magik-eye.

Looks like "Shikino" "knows" a few things or two And just when i thought Magik-eye was in sleepy town, sneeky find

3Dセンシングに先進的な技術を保有するMagikEye Incとの資本業務提携のお知らせ

www.dreamnews.jp

Announcement of capital and business alliance with Magik Eye Inc, which owns advanced technology for 3D sensing​

Accelerating the development of camera modules targeting the growing 3D sensing market
Shikino High Tech Co., Ltd. (hereinafter, Shikino High Tech) is a part of MagikEye Inc. (Headquarters: Connecticut, USA, Magic Eye),
which possesses advanced technology for 3D sensing. On June 7, 2022, we announced that we
have acquired shares and signed an MOU for joint development of camera modules that utilize this technology . This alliance is the first step towards future collaboration between the two companies developing products and technologies , especially with regard to the commercialization of cameras for 3D sensing, which is growing rapidly in the sensing camera market, where demand has been increasing in recent years . After discussions between the two companies, we plan to conclude a joint development agreement for commercialization. Takeo Miyazawa, CEO of MagikEye Inc., said: "We have developed a unique 3D measurement algorithm: ILT (Invertible Light Technology) (registered trademark). Simply add an ultra-compact laser projector designed with this technology to an existing camera module to achieve up to 600fps. It can achieve higher speed and lower delay response within 2ms at the shortest. It does not require major changes or additional resources to the CPU or SoC. By advancing, we are confident that we will be able to market the 3D sensor products required by various businesses at the fastest speed. ” Shikino Hi-Tech has introduced Magic Eye technology to commercialize advanced and high-performance 3D cameras. Aim

And it also seems they are associated with Socionexts SC1320A o and wait for it... PANASONIC

As members of the HD-PLC Alliance, Socionext and Shikino High-Tech will continue to work closely together and propose a variety of applications that utilize HD-PLC, in addition to delivering stand- alone ICs and communication modules, thereby contributing to the spread of this technology and the expansion of the market.

Shikino High-Tech communication module P-TMFSU-041 information:

https://www.shikino.co.jp/eng/products/P-TMFSU-041_specifications_E.pdf

Socionext HD-PLC IC SC1320A information:

https://www.socionext.com/en/products/assp/HD-PLC/

About HD-PLC IP core (Panasonic Corporation):

https://www.panasonic.com/global/business/hd-plc/en.html

Seems like Shikino is singing -

A little bit of Socionext here i am, A little bit of Panasonic by my side, a little bit of Magic-eye is what i see and A little bit of AKIDA is what i need. You and me a gonna touch the sky.

Ok my singing voice is not the best but it seems all in tune above.

 

[Diogenese]

Xavier is the old platform, build in actual Mercedes models. Thought there might be some progress in the new Orin architecture, as it is much faster with less power consumption..

Hi Perhaps,

I was a little distracted and went down the Xavier rabbit hole following Fn 2.

However, as you can see, the discussion re Fig. 5 from my previous response applies to Orin. It seems nvidia wouldn't know a spike if they sat on one. They are using 16-bit or 32-bit weights and activations, with their 2-out-of-4 zero cancellation with associated values for their CNN on a chip.

The NVIDIA Deep Learning Accelerator, or DLA, is a fixed-function accelerator optimized for deep learning operations. It is designed to do full hardware acceleration of convolutional neural network inferencing. The Orin SoC brings support for the next generation NVDLA 2.0 with 9X the performance of NVDLA 1.0. DLA 2.0 provides a highly energy efficient architecture. With this new design, NVIDIA increased local buffering for even more efficiency and reduced DRAM bandwidth. DLA 2.0 additionally brings a set of new features including structured sparsity, depth wise convolution, and a hardware scheduler. This enables up to 105 INT8 Sparse TOPs total on Jetson AGX Orin DLAs compared with 11.4 INT8 Dense TOPS total on Jetson AGX Xavier DLAs.

PS: While it is against Ogres Inc policy to accept returned ogres, the company may have been prepared to make a exception in this case. However, it would then be necessary to issue a new one ...

 

[Rise from the ashes]

Done

 

[Rise from the ashes]

Apologies

 

[Diogenese]

Done

Thanks.

 

[Fox151]

Hey Izzzzy, perhaps you didn't know this but using the C word is actually incredibly misogynistic and insulting to women. Nothing shows greater contempt for women than using the vernacular for vagina as the ultimate term of abuse or insult. Just saying.

Unless he meant dont be a

 

[Bravo]

Unless he meant dont be a

We all know what he meant because he's done it before.

 

[jtardif999]

Thank you for that. Sounds like AWS has got good security but is still cloud based so not suitable in areas of dodgy internet connection.
Also the latency issue may still be a problem?
Cheers
Boab

Just need to unwind the acronym really, AWS stands for Amazon Web Services - says it all, their edge can only be the edge of the cloud.

 

[Fullmoonfever]

Just found this pop up on a surf.

Not sure if been posted previously but anyway, little 2pg intro to Akida

Doc attached.

 

[Diogenese]

Just found this pop up on a surf.

Not sure if been posted previously but anyway, little 2pg intro to Akida

Doc attached.


View attachment 10288


View attachment 10290

Great find Fmf,

Here's a link to ARM:

https://www.arm.com/partners/ai-ecosystem-catalog

If you follow the link and click on "Learn More" in the brainchip box, there's a lot more info.

 

[Slade]

BrainChip and Palantir with strong rebound - How does BioNTech react?

The correction on the stock markets has been going on for months. Technology stocks, in particular, have been hit hard. Due to the end of the ultra-loose monetary policy and possible larger interest rate steps, even market leaders lost more than 80% of their ground in some cases. Whether...

news.financial

 

[stockduck]

That's right - look at the stuff up we've made with coal and oil.

The option of concentrated solar thermal is looking better all the time. Linear parabolic solar reflectors focused on water/steam pipes - a steam powered generator powered by solar steam, and when the sun goes down, a hydrogen boiler.

Diogenese this is exactly where humankind should travel along to get energy for industrial prosperity with also underdeveloped countries to help them to prosper and built up a democratic lifestyle for their own culture.

I read today something new important technical revolution for producing h2 with an mindblowing efficiency of 95-98%.
It`s a technical invention from an australian university called:

• AIIM Building, Innovation Campus
• North Wollongong
• Wollongong, New South Wales 2500
• Australia

News & Media – Hysata

hysata.com

Please feel free to tell the brainchip team to wake up their intrests to support the production process of these new ‘capillary-fed electrolysis cell’ by using Akida Technology or something else, to make it the efficientest h2 technology.

Hysata’s electrolyser breaks efficiency records – Hysata

hysata.com

Unfortunetly, we need a lot of gas until now for producing solar glass....
Everyone with good vibrations to the technical team here feel free for doing some reseach ....for better support.

 

[Rise from the ashes]

Micron Delivers ASIL-D ISO 26262 Certified LPDDR5 for Safety Requirements

Micron LPDDR5 is certified ISO 26262 ASIL-D compliant for automotive applications with the strictest safety requirements. Our commitment to functional safety includes designing products with random fault detection capabilities as well as a Safety Office staffed with dedicated engineering talent.

www.micron.com

 

[Rise from the ashes]

Cerence Inc. on LinkedIn: Cerence Cloud Services in MBUX

The latest version of #MBUX -- Mercedes-Benz User Experience -- powered by Cerence, features seamlessly integrated cloud services and flexible natural language…

www.linkedin.com

 

[Rise from the ashes]

What does “Lead in Car Software” mean to us? Our customers come first - LinkedIn

What does “Lead in Car Software” mean to us? Our customers come first! So, with software, we want to save them time and offer the highest driving comfort...

www.linkedin.com

 

[thelittleshort]

BRAINCHIP HOLDINGS - CLASSIC SETBACK OFFERS OPPORTUNITY​

Unlike Netflix, Zoom or other tech industry giants, the developer of the Akida chip, Australia's BrainChip Holdings, is showing resilience. The share price performance of the emerging IP company, which is the world's first commercial manufacturer of neuromorphic AI chips, has been in the green since the beginning of the year. After starting the year at EUR 0.45, it went up to EUR 1.66 on the back of news related to a cooperation with car giant Mercedes. It was announced that Akida technology would make the "Hey, Mercedes" voice control in the EQXX five to ten times more efficient than conventional voice control. Since then, however, the share has corrected by almost 70% due to the weak market environment. However, a bottom is currently being formed in the EUR 0.60 area. In addition, MACD and RSI are close to a buy signal.

In the long term, BrainChip's novel AI chip could become the new standard in the semiconductor market. The advantages are that Akida is extremely low-power, high-performance, and promotes the growth of edge AI technology by using neuromorphic architecture, a type of artificial intelligence inspired by the biology of the human brain. BrainChip enables the universal use of efficient edge computing in real-world applications such as connected cars, consumer electronics and the Internet of Things. It is no coincidence that BrainChip has been accepted into chip giant ARM's partner program, enabling research into the next generation of AI solutions.

Another new partnership is with the inventor of the world's most advanced neuromorphic vision systems, Prophesee. Prophesee's technology is inspired by human vision. It uses a patented sensor design and AI algorithms that mimic the eye and brain to detect what was previously invisible with standard image-based technology. Prophesee's computer vision systems, like Akida, target segments such as autonomous vehicles, industrial automation, IoT, security and surveillance, augmented reality, and virtual reality.

"We have successfully ported Prophesee's neuromorphic-based camera sensor data to process inferences on Akida with impressive performance," said Anil Mankar, Co-Founder and CDO of BrainChip. "This combination of intelligent vision sensors with Akida's ability to process data with unprecedented efficiency, precision and power savings at the point of capture is a true advancement in AI and provides manufacturers with an immediately actionable solution."

 

[Rise from the ashes]

Embedded Insiders Podcast: Verifying the CHIPS Act and Emerging Security Standards - Embedded Computing Design

On this episode, the Insiders are joined by Andreas Kuehlmann, Executive Chairman and CEO of Cycuity, formerly known as Tortuga Logic, to discuss the importance of security at the chip level amidst the release of new standards and recent government involvement in the semiconductor industry via...

embeddedcomputing.com

 

[Sirod69]

does somebody knows something about the meaning of this twitter post?

 

[Rise from the ashes]

does somebody knows something about the meaning of this twitter post?

who is M&S services GMBH?