No I don't, but I had the benefit of seeing the demo after this years AGM put on by Tim Llewellyn.I need some help here with the wording circled in red. To me it sounds like we are being used within the Nvidia Jetson nano chip. Is that how others interpret that sentence. TIA View attachment 21801
Thanks FF for the clarification.No I don't, but I had the benefit of seeing the demo after this years AGM put on by Tim Llewellyn.
Tim had the Nvidia Jeston Nano as well as the AKIDA board to show the performance comparison and I believe that is all he is alluding to here.
I believe he is saying that AKIDA running at 300MHz is accelerating the performance you can achieve with an ARM Cortex A57 by 18.1x.
Then for a reference as to what this means he is telling readers that the ARM Cortex A57 he is comparing AKIDA against is in fact the same semiconductor that powers the Nvidia Jetson Nano.
Imagine that AKIDA 1.0 is blowing Nvidia Jetson Nano away how can this be say all those who have never heard of AKIDA technology.
My opinion only DYOR
FF
AKIDA BALLISTA
PS: This find you have generously shared here just shows what a magnificent BEAST AKIDA 1.0 is compared with GPU's and CPU's running the Nviso Apps. AKIDA 1.0 just too cool for school. What a great weekend find many thanks @TopCat
... and in the immortal words of the Bard himself ...No I don't, but I had the benefit of seeing the demo after this years AGM put on by Tim Llewellyn.
Tim had the Nvidia Jeston Nano as well as the AKIDA board to show the performance comparison and I believe that is all he is alluding to here.
I believe he is saying that AKIDA running at 300MHz is accelerating the performance you can achieve with an ARM Cortex A57 by 18.1x.
Then for a reference as to what this means he is telling readers that the ARM Cortex A57 he is comparing AKIDA against is in fact the same semiconductor that powers the Nvidia Jetson Nano.
Imagine that AKIDA 1.0 is blowing Nvidia Jetson Nano away how can this be say all those who have never heard of AKIDA technology.
My opinion only DYOR
FF
AKIDA BALLISTA
PS: This find you have generously shared here just shows what a magnificent BEAST AKIDA 1.0 is compared with GPU's and CPU's running the Nviso Apps. AKIDA 1.0 just too cool for school. What a great weekend find many thanks @TopCat
I’d love to know how much or if any of this is our IP? So far we’ve only seen Renesas mention us in one article and without a specific use case. I spose the question we need to consider is could these use cases be accomplished with the 2 nodes that they’ve licensed?Rob Telson is a big FAN of fans.
My opinion only DYOR
FF
AKIDA BALLISTA
If I have understood what Peter van der Made, Anil Mankar and Kristopher Carlson have presented the answer is yes because a single node can do everything 80 nodes can do but you will sacrifice latency and training time.I’d love to know how much or if any of this is our IP? So far we’ve only seen Renesas mention us in one article and without a specific use case. I spose the question we need to consider is could these use cases be accomplished with the 2 nodes that they’ve licensed?
Brainchip Turner Overdrive? Hope you are faring well, FF. All the best from Queensland,... and in the immortal words of the Bard himself ...
"You ain't seen nuthin' yet!"
Brainchip Turner Overdrive? Hope you are faring well, FF. All the best from Queensland,
Adam aka vcis123
So, crudely, $463 million royalty revenue from 7.5 billion chips? 6 cents per chip? Am I reading this correctly?Arm on LinkedIn: Arm technology is defining the future of computing
We achieved record royalties in Q2! 🎉 Our investment across all lines of business is continuing to pay off, with impressive growth in automotive and IoT in…www.linkedin.com
Just look at this as one partner, 7.5 billion chips in One Quarter ……….
Wow, say no more for Chippa to be a fraction of this would be phenomenal !!
View attachment 21802
Glad to see you here VCIS123.Brainchip Turner Overdrive? Hope you are faring well, FF. All the best from Queensland,
Adam aka vcis123
Exactly, a clogged air filter, anomaly detection or a sensor looking for a human entering a room doesn’t require a millisecond response and could be achieved with a “multi pass”. Safety (time) critical applications may need more nodes.If I have understood what Peter van der Made, Anil Mankar and Kristopher Carlson have presented the answer is yes because a single node can do everything 80 nodes can do but you will sacrifice latency and training time.
Which is why it is incredibly difficult to rule in or out with Renesas unless they make a deliberate or accidental reveal.
My opinion only DYOR
FF
AKIDA BALLISTA
Those figures ae presumably $US, but still not startling.So, crudely, $463 million royalty revenue from 7.5 billion chips? 6 cents per chip? Am I reading this correctly?
That's a great article. Thanks for posting.Those figures ae presumably $US, but still not startling.
The other thing is that many of ARM's customers would be established clients, so would be paying royalties only and not the licence fees. BrainChip have an open order book.
ARM licence fees ten years ago:
https://www.anandtech.com/show/7112/the-arm-diaries-part-1-how-arms-business-model-works/2#:~:text=The upfront fee generally ranges from $1M -,2% of the selling price of the chip.
The ARM Diaries, Part 1: How ARM’s Business Model Works
by Anand Lal Shimpion June 28, 2013 12:06 AM EST
The upfront fee generally ranges from $1M - $10M, although there are options lower or higher than that (I’ll get to that shortly). The royalty is on a per chip basis. Every chip that contains ARM IP has a royalty associated with it. The royalty is typically 1 - 2% of the selling price of the chip.
https://www.neowin.net/news/arm-reportedly-hiking-licensing-fees-of-its-chips-by-four-times-for-some-customers/
Arm reportedly hiking licensing fees of its chips by four times for some customers
Rajesh Pandey @ePandu · Jul 16, 2020 02:58 EDT ·
A couple of days after reports first emerged of SoftBank looking to sell Arm Holdings partially or do a public offering, a new report sheds light on how Arm has increased the licensing fees for some of its customers. The Reuters report claims that in recent negotiations, Arm representatives have hiked license costs by as much as four times for some customers. The move has led some of Arm's licensees to consider non-Arm alternatives.
“It’s created a lot of tension for us,” one Arm licensee told Reuters, saying the hikes seemed out of proportion to the improvements in the technology.
Arm's instruction set and CPU designs play a major role in almost every consumer product out there today, including smartphones, laptops, tablets, wearables, and more. The importance of Arm has only increased over the last decade with the rising popularity of smartphones. They are going to become even more important in the future as Apple has made its intention clear of switching its Mac lineup to custom Arm-based cores as they will offer better performance and efficiency to consumers. Some companies like Qualcomm license the instruction set as well as CPU core designs from Arm and pay a royalty on the latter. Others like Apple only pay the licensing fees for the instruction set and design their own CPU cores.
Arm already rakes in millions of dollars every year from licensing fees and billions of dollars in royalty fees from chips designed by it. A steep rise in Arm licensing fees could negatively impact consumers as well as companies could be forced to increase the final product price to make up for the increased fees.
I wonder how much the customers are prepared to pay proportionally for improvements in the technology.