brainchip AKD1000

[ali_asadzadeh]

Hi,
I was searching keil packs, and the brainchip AKD1000 catch my attention, It seems we would get some Cortex M4 part with some accelerators for doing AI in IOT devices.
But I could not find any datasheet or user manual yet,

This is their website,
https://brainchipinc.com/

Do you know any info on this core or company?

[iMo]

Each Akida NSoC has effectively 1.2 million neurons
and 10 billion synapses, representing orders of
magnitude better efficiency than other neural
processing devices on the market.

How all those neurons and synapses fit into that chip??

[ali_asadzadeh]

I don't know maybe a typo, Also they say it has USB3.0, PCIe  etc...

[uer166]

Each Akida NSoC has effectively 1.2 million neurons
and 10 billion synapses, representing orders of
magnitude better efficiency than other neural
processing devices on the market.

How all those neurons and synapses fit into that chip??

Probably not a single chip, but

scalable solution utilizing a built-in serial chip-to-chip connectivity to allow up to 64 devices to be arrayed as a single solution.

With 64 chips that number might be vaguely achievable?

[WatchfulEye]

Each Akida NSoC has effectively 1.2 million neurons
and 10 billion synapses, representing orders of
magnitude better efficiency than other neural
processing devices on the market.

How all those neurons and synapses fit into that chip??

The individual components may be quite small. The technology appears to be a variant of spiking neural network, where instead of RAM/multiply-add units for storing data and performing the matrix multipications digitally, the technology is effectively analog - albeit using pulse frequency modulation (hence the "spiking" name) rather than a continuously varying voltage.

The basic operation of a neuron can be modeled as the dot product (a∙b) of a array of pulse modulated inputs (a) and an array of pre-programmed memristors (b). The current delivered by each input is given by a_ib_i, where a_i is the pulse frequency of input i, and b_i is the programmed conductivity of memristor i. By connecting all inputs to a summing capacitor, the dot product operation is complete. So, you can have a neuron with 1k inputs consisting of a capacitor, sense amp/pulse frequency modulator and 1000 memristor cells which receive inputs from 1000 other neurons. The memristor cells can be manufactured on the scale of flash memory cells, and multilayer manufacturing (similar to 3D NAND flash) has also been demonstrated, so 10 G memristor cells on one chip doesn't seem completely unreasonable.

[MadMayHam]

Hi all, you can actually buy the development kits with the actual production version of the Akida 1000 processor in silicon at https://shop.brainchipinc.com/ .

[ali_asadzadeh]

MadMayHam are you working there?