Hi Iseki,
You’re a valued poster and obviously a bit passionate about the delays in Renesas producing a product with Akida. Obviously we all want Akida nodes put on a Renesas product asap. Don’t feel humiliated if you get something wrong; we all do; it’s the best way to learn.
I’ve gone back to your original post to try to understand your frustrations better. Forgive me if I am misunderstanding the issue or don’t explain it very well as I’m not a computer person and what I am trying to explain could be wrong anyway so take it with a pinch salt
I’m trying to learn also, so I can understand what the company is doing as my future and retirement is invested in it.
The way I understand how it all works is similar to baking a cake. There are many different recipies depending on what result you want to achieve. You can make big cakes, like the Arm M85, or small cakes like the Arm M33. The different families of devices can do different workloads. Using an Arm chip for example is like choosing a brand of flour for the cake. There are many brands. In this instance Renesas are using Arm.
Then the different components are specific to the task the chip is to perform, or like the cake analogy; nuts, raisens or chocolate as per personal preference which might correspond to having for example; bluetooth capabilities, several sensor types and maybe a voice activation system.
Renesas are not going to make an Akida Chip alone unless someone pays them to. The chip/iP needs to be married up with a sensor of some type to recieve information, to then make a decision on that information which is passed onto another part of the device to perform a task. This is my understanding of a System on Chip (SOC). When Brainchip made a chip they also used an Arm chip to put their IP onto it also.
My understanding is that Renesas have paid a million dollars to licence 2 nodes to use in a product (SOC) they are developing. The entire Akida chip had 80 nodes but having 80 nodes when you only need 2 is the equivelant to getting Arnold Swartznagger to lift a feather: not necessary.
We will get royalties from the product as it is sold. Renesas have stated they are going to put 2 nodes onto a System on Chip (SOC) which includes the Arm Cortex M33.
The M33 is a very low powered chip so I am guessing that is why the Akida nodes will be used on it AND why only 2 are required.
If they use something else it will require too much power and therefore will defeat the purpose of getting a low powered chip. An example is the smart wrist watch type applications which is battery powered which is why there was excitement about their latest product release. That was why I was looking into the Voice Wake Activation System but was dissapointed I couldn’t find any details pointing to Akida.
However this is where ultimately Akida will eventually win. As a result of the brilliant neuromorphic cabilities Akida is extremely low powered and yet high performance, plus there is on chip learning. Akida won’t be everywhere, there was always going to be competition. But Akida is the best in class and therefore it will find it’s way into products as it can do what others can’t. So where it is necessary or advantagous to use it I’m sure Renesas will.
If you’re still reading this far, apologies for the length of the post: I hope it makes sense and is helpful.
Cheers