Well using FF's security camera example, a normal camera (not a DVS camera) records every pixel on the photoreceptor plate and bundles this information into a packet.
Say each the information from each pixel is 8 bits, and there are L pixels per line, and N lines per frame.
Thus the packet includes 8*L*N bits plus header and check information. Each of these bits must be processed by the computer.
Suppose the computer is programmed to identift objects in the picture. The computer must then compare the bits from the camera packet, the camera must retreive images of objects from a memory so it can compare the image from the camera with thousands of images from the memory.
If Akida receives the same packet from the camera, it identifies which bits have changed from the previous frame (active pixels or events), and only processes those bits. This can reduce the number of bits to be processed by a factor of a thousand or more.
Akida also has the comparison images stored right next to the "neuron" processors which compare the active pixels (events).
Storing the bits next to the neurons eliminates one of the large delays in normal (von Neumann) computers, the von Neumann bottleneck, which results in the traffic jams arising from transferring information from memory to processor on a set of wires called a bus.
So Akida is much faster and uses less power than a normal computer because it processes far fewer bits, and it eliminates the memory/processor traffic jam.
Also, because it processes fewer bits, it uses less power.
(20min delay)