Feature: Edge AI
systolic array that maps naturally to a neural network. This fundamental new approach includes a new type of compute unit, the analogue MAC – the block that does the work of the von Neumann architecture’s ALU and memory units; see Figure 3. The analogue MAC is optimised for AI systems, where MAC operations represent 95% of the compute workload. Thanks to another Ambient Scientific innovation – the HyperPort 3D memory architecture, which enables vertical stacking of memory elements at each MAC unit – the analogue MAC implements in-memory computing, solving the von Neumann problem of the physical separation between memory and compute blocks. The second breakthrough in the
Figure 3: The analogue MAC enables in-memory computing
analogue computing approach to AI processing is to combine analogue MAC blocks into a structure – a matrix computer – which mirrors the topology of a neural network. This solves the second weakness of the von Neumann architecture in neural network operations: the vastly inefficient way in which a neural networking model is compiled into instructions for a conventional compute system. Each DigAn unit is a single monolithic
circuit that computes an entire layer of neurons in a single cycle. Multiple layers of DigAn circuits can be scaled up into a matrix computer which mirrors the structure of a neural network; see Figure 4. The practical result of this is the big improvement in instruction cycle efficiency: just a single DigAn compute block can compute a 1x32x8 matrix in one cycle, compared to 38,600 cycles for an AI processor based on a conventional silicon architecture. Multiple layers of these DigAn blocks
Figure 4: Ambient Scientifi c’s matrix computer, built from multiple layers of DigAn circuits
compute blocks causes the same problem of memory access – reducing throughput and increasing power consumption – as with the von Neumann architecture. The systolic array is an effective
solution to the problem of mapping the compute architecture to a neural network’s topology, but what is required
to implement it is innovation at the silicon level. Ambient Scientific has achieved this with its DigAn technology.
The configurable matrix computer The DigAn technology enables Ambient Scientific to create at chip level a configurable matrix computer, a kind of
28 February 2026
www.electronicsworld.co.uk
form a matrix computer: 32 layers of a typical 1x32x8 neural network matrix would require 1,235,200 cycles for a conventional compute architecture to perform. In a DigAn matrix computer, this requires just 32 cycles. When reducing neural networking operations from 1,235,200 cycles to 32 cycles, the application benefits from remarkable improvements in performance and
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48
