… and we become the AK-1000s.
BRN Discussion Ongoing
- Thread starter TechGirl
- Start date
BaconLover
Founding Member
stuart888
Regular
I was not aware that SiFive are collaborating with Intel on a ‘powerful tool for developers’
Is BrainChip also involved?
View attachment 27066
HiFive Pro P550
RISC-V is inevitable, and the HiFive Pro P550 development system exemplifies that.
In partnership, Intel and SiFive are excited to introduce the highest performance RISC-V development board, which is scheduled to be available Summer 2023.
The soul of the machine is the Intel Horse Creek SoC, built on the Intel 4 process, that includes a SiFive Performance™ P550 Core Complex, a quad-core application processor featuring a thirteen-stage, triple-issue, out-of-order pipeline with the RISC-V RV64GBC ISA, and on-board DDR5-5600 and PCIe Gen5.
Board features (subject to change) include; 16GB DDR5, 2x PCIe expansion slots, 1/10GbE Networking, USB 3, on-board graphics and a remote management ready interface (OCP DC-SCM).
This is a premium software development system ideal for developer desktop machines and rack-based build/test/deploy servers for RISC-V software development. RISC-V has no limits.
View attachment 27069
View attachment 27071
HiFive™ Premier P550 - SiFive Boards
RISC-V is inevitable, and the HiFive Premier P550 development system exemplifies that. The soul of the RISC-V development board is a SiFive Performance™ P550 Core Complex, fabricated in a 12nm process technology.www.sifive.com
BrainChip and SiFive Partner to Deploy AI/ML Technology at the Edge
LAGUNA HILLS, Calif., April 5, 2022 — BrainChip Holdings Ltd, a commercial producer of ultra-low power neuromorphic AI chips and IP, and SiFive, Inc., the founder and leader of RISC-V computing, have […]www.hpcwire.com
Intel at 18:25
A high-class AI forum with people like @thelittleshort.
Amazing to me, and it is people from all over the world investigating and reporting the findings to the 1000-eyes-team. Every twitter, Instagram, every YouTube, every financial statement, every everything.Crazy. Brainchip has a great shot is to take for instance $25k to $250k, within 36 months.
Nothing certain, but the dots say likely.I actually feel like this is not just feasible, but a darn-good-chance. This company is freaky, they have the secret snn spiking key. Time, and low power smarts will prove out.
Akida 2.0 and 2.1 and 2.2 are in the works. Big catalyst as the BRN income stream goes from trickle (expected 2023) and a notch higher month after month ongoing.
Happy to be on the Brainchip Bus!
Thanks thelittleshort!hamilton66
Regular
Rise, enjoyed that a lot. And the best bloke in the room was - the Aussie!
stuart888
Regular
I watch every week. So good, the big picture. The first few weeks as a new viewer, I would pause to look at the charts he was looking at. Yeah @Dhm!
After a few weeks or longer, I was right on board. Could not say enough good things about this no commercial market breakdown. Once a month or every two weeks might be the sweet spot.
First thing I watch every Saturday morning. Technical charts, history, but easily digestible.
FJ-215
Regular
Good morning,
The company won't be making any comments about our new chip/s until it's official launch, which I would "assume" would be around June 2023, that is after the foundry has finished it's run and testing, before being handed over to our team to evaluate, before we in turn make the reference chips available to our clients.
Great to see Nikunj speaking in glowing terms with regards to our partners Edge Impulse:
"I am so delighted to share with you all that yesterday was the official launch of the support for #BrainChip #Akida technology on the #EdgeImpulse Studio. I am very excited with this launch as it allows anyone to build and deploy machine learning models on #Akida, which is an advanced technology that is designed based on Neuromorphic principles to deliver low power and low latency for ML inferences at the #Edge. This partnership aims to advance developments of millions of products that are looking to have the capability of the cloud at the #Edge where problems such as power, latency, privacy, etc are the biggest challenges to overcome."
"We are already seeing the excitement behind this launch by many, and I cannot wait to see what the community will accomplish with this integration."
Note the passion, millions of products. Nikunj knows the real potential of what is about to commence over the short/medium term.
I'd like to see our university program keep growing, it's a very powerful tool in getting young students up to speed in the Neuromophic computing space, as when all is said and done, they and their ideas are the future.
My own opinion/s.
Tech
Haven't even taped out the chips and you are talking launch dates!!!
Settle petal.
stuart888
Regular
Toss in the massive reduction in the M2 Money Supply @BaconLover. When money is flowing out of the system, a typical standard down cycle. Just for the short term, as innovation will blast back.
Enjoy the wild cycles, the chance to buy lower and wait for SNN low-power IoT with periodic cloud future to arrive. The difference between a year ago and today is night and day.
BaconLover
Founding Member
Toss in the massive reduction in the M2 Money Supply @BaconLover. When money is flowing out of the system, a typical standard down cycle. Just for the short term, as innovation will blast back.
Enjoy the wild cycles, the chance to buy lower and wait for SNN low-power IoT with periodic cloud future to arrive. The difference between a year ago and today is night and day.
View attachment 27126
View attachment 27125
No doubt semiconductor will get back into the favourite sector again.
I'm not sure about Tech sector in general though. Once Edge AI becomes standard, those data leaking dependent companies like Facebook, Snapchat etc will go out of fashion and new players emerge.
At the moment, insiders are still selling their shares in Tech sector. (This is a fact, everything else above is my opinion)
stuart888
Regular
That fantastic article 4 Bits Are Enough got me digging into a little deeper into some of PvDM's writings.
https://www.mdpi.com/1424-8220/19/22/4831
I was thinking about the input spike data. For more simple pattern detection, my assumption it is typically an array of numbers. I could be way off with the voice or text, but video pixels would be numbers. Sound would be numbers, and all packed into an array, I believe. A big array, lots of numbers.
So I thought Akida version one was targeting use-cases where the events happen rarely. For example, the use-case of an endangered animal walking in front of a battery powered wilderness video camera. Only when it sees the very rare (less than 300 alive) Florida Panther does it trigger the event.
So lots of zeros, until some sort of animal passes past the camera, where motion detection triggers the video camera.
Then with the 1000fps video camera, I suppose that is Akida making inference and machine learning on 1000 fps is awfully special. On the edge, where only real events get captured and passed to the cloud on some timetable.
From sparse to fastest of all time!
Perhaps only the Brainchip SNN Team really knows, with their secret JAST sauce. Perhaps (low event bits, and after the first 4 events, Akida will demote all to zero). i==BigfanHere!
Just trying to always learn more.
https://www.mdpi.com/1424-8220/19/22/4831
I was thinking about the input spike data. For more simple pattern detection, my assumption it is typically an array of numbers. I could be way off with the voice or text, but video pixels would be numbers. Sound would be numbers, and all packed into an array, I believe. A big array, lots of numbers.
So I thought Akida version one was targeting use-cases where the events happen rarely. For example, the use-case of an endangered animal walking in front of a battery powered wilderness video camera. Only when it sees the very rare (less than 300 alive) Florida Panther does it trigger the event.
So lots of zeros, until some sort of animal passes past the camera, where motion detection triggers the video camera.
Then with the 1000fps video camera, I suppose that is Akida making inference and machine learning on 1000 fps is awfully special. On the edge, where only real events get captured and passed to the cloud on some timetable.
From sparse to fastest of all time!
Perhaps only the Brainchip SNN Team really knows, with their secret JAST sauce. Perhaps (low event bits, and after the first 4 events, Akida will demote all to zero). i==BigfanHere!
Just trying to always learn more.
Izzzzzzzzzzy
Regular
Sorry what does edge AI have to do with Facebook and Snapchat leaking data. Your data on this website TSEX is also being sold to provide advertisements through cookies. That statement cannot be just thrown around without context. Unless you are talking about being hacked then that’s something very different as well.
Izzzzzzzzzzy
Regular
Just wanted to pick some minds here, what is everyone’s realistic expectation or opinion on the up coming 4c? As for the past couple of ones i have had no expectations, I will keep the same sentiment for the up coming one. I just really really wish to be proven wrong.
stuart888
Regular
Google has been life changing. The best company in massive innovation (google maps, android, youtube) where everything is free to the user.
Big data has its place for sure. I own google stock from years back, but my new money is going to Brainchip.
Edge IoT via SNN is Brainchip's to grab the spotlight over time. Lots of AI/ML companies are going to win.
Fullmoonfever
Top 20
@Diogenese
Further info I posted on Horse Creek, SiFive etc beginning of Jan.
Have zoomed in and snipped a couple pics from presso which were interesting I thought.
Additional IP such as AI accelerators and I see they using Xilinx on the board too.
Not sure if any of this assists our opps of getting involved at some point?
BRN Discussion Ongoing
You stirred up some emotions with that post. https://www.maritimequest.com/daily_event_archive/2008/07_july/07_smeraldo.htm An uncle I never got to meet. Sorry about your uncle Rise, was not my intention to upset you, hope you are ok ❤️
thestockexchange.com.au
BaconLover
Founding Member
How AI Can Help With Data Privacy - BairesDev
The more digital our lives get, the more exposed we become. That’s why companies are working on ways to ensure privacy while using AI.
As I very clearly mentioned in my post, that was my opinion.
This is an article I have come across, hope that shines some light into what I was saying. Again, the article doesn't specify edge AI, but we know BRN could be a key player in cybersecurity.
As to Facebook and Snapchat, well, those were examples, and when I thought about data leaks, those were the companies that came to my mind. Again, it is my opinion.
Rise from the ashes
Regular
Awesome post Mr.Bacon
Mccabe84
Regular
Im not expecting anything in the 4C till the 2nd half of this year
TheUnfairAdvantage
Regular
From memory Megachips is the only company that has signed and sealed IP licence with payment for such licence outstanding. It has been accrued but not paid. From memory this is because there were milestones for product sales in the Ip licence. The half yearly recognised product sales and engineering associated to that licence. I am hoping that those milestones have been met and payment has been made ie recognised as cash receipted in this coming 4c. Hoping for $4m from Megachips. Bonus if Socionext or renasus has paid something.
Hi Stuart,
I'm not sure I follow your drift.
"I was thinking about the input spike data. For more simple pattern detection, my assumption it is typically an array of numbers. I could be way off with the voice or text, but video pixels would be numbers. Sound would be numbers, and all packed into an array, I believe. A big array, lots of numbers."
There's quite a lot I don't understand about how Akida works, but here is a summary of my understanding.
Sensor data is analog. For instance, a pixel puts out a voltage signal whose amplitude is proportional to the strength of light hitting the pixel.
A digital camera will then have an Analog to Digital Converter (ADC) to convert the analog pixel signal to digital bits, say 16 bits. As you say, these digital numbers are arranged in an array corresponding to the pixel array.
The matrix of these digital bits is what CNN operates on to classify the object being photographed. When the CNN compares small segments (sliding windows) of the matrix with the images stored in the 16-bit image model library, it performs a Multiply Accumulate computation (MAC) by multiplying the two 16-bit numbers together for each pixel. The processing is sequential as the sampling window "slides" across the larger matrix.
Akida uses N-of-M coding and the JAST rules directly on the analog pixel signal.
The strength of the light impinging on a pixel determines both the amplitude of the analog signal and the time when the signal is generated, the stronger the light input, the faster the analog signal is generated. N-of-M coding utilizes this timing to select the strongest incoming signals (N signals) and to ignore the later arriving (M-N) signals because most of the image information is contained in the stronger signals. The pixels are processed in parallel so the timing of the input analog signals can be compared, hence the millions of neurons.
In addition, where the content of adjacent pixels is the same (within a predetermined threshold value), only the first pixel is processed. This provides sparsity in processing so only edges of uniform image segments are processed.
Thus
CNN requires the ADC process for every pixel - Akida does not need ADC.
CNN processes every pixel - Akida only processes N-of-M pixels
CNN processes every pixel - Akida only processes changes in pixel strength.
CNN uses MAC for every pixel - Akida does not use MAC.
CNN runs entirely in software on a CPU. Akida runs in a SoC almost without CPU involvement*
CNN requires the fetching of instructions and image data from memory (von Neumann bottleneck) - Akida stores the data with the NPUs.
*Akida 2-bit and 4-bit operation now uses of the order of 3% CPU operation and 97% silicon for recurrent NNs.
https://brainchip.com/4-bits-are-enough/
Last edited:
Getupthere
Regular
I’m not expecting too much from the Dec 4c.
Particularly with renesas only Taping out last month.
Looks like Renesas were 6 months behind schedule.
Bring on 2023!!
Particularly with renesas only Taping out last month.
Looks like Renesas were 6 months behind schedule.
Bring on 2023!!
Similar threads
