Brain-inspired, from the silicon up

Our processors work the way neurons do, staying quiet until a signal changes and then responding in an instant, so a device can listen and decide for itself with no connection to the cloud.

The Catalyst N4-Edge chip

Quiet until something happens

A microphone on a hillside hears wind and silence for most of the day, a conventional processor works through every one of those moments regardless, whereas our chip only does work when the sound changes, so the moment that matters gets its full attention.

Conventional processor
Catalyst N4-Edge
Radio
One alert, sent once

Why spiking networks

90.28 percent

Accuracy on the full Spiking Heidelberg Digits test set, run on our own hardware, against 90.9 percent published for Intel Loihi 2.

18 million

Events routed with none dropped, because a neuron is only updated when an event reaches it, so silence costs us nothing.

Bit exact

The hardware matches our reference simulator at every timestep, so a network that works in simulation works on the chip.

0.94 seconds

A network of 1.5 million connections loads from memory, so a new model is data as opposed to a new chip.

events per 10 ms 0100200300 290 in the busiest slice under 10 a slice from 0.53 s 0 0.2 0.4 0.6 seconds
The same utterance counted in 10 millisecond slices, 8,216 events in 0.71 seconds. The work follows the sound, so once the speaker stops there is almost nothing left for the chip to do.

How we compare

The first column is from our own hardware. The other two are published figures, from Intel and from the research literature.

85%90%95%100%
Catalyst N4, on our FPGAour FPGA, full test set 90.28%
Intel Loihi 2, publishedpublished 90.90%
Research best, publishedpublished 96.41%
Spiking Heidelberg Digits, accuracy on the full 2,264 sample test set. The scale starts at 85% so the three figures can be told apart.
Catalyst N4, on our FPGA Intel Loihi 2, published Research best, published
Spiking Heidelberg Digits90.28 percent, full test set on chip90.9 percent96.41 percent, software
How the figure was obtainedRun on hardware, all 2,264 samplesIntel published figurePublished software result
Event lossNone across 18 million eventsNot publishedNot applicable
PowerNot measured, no board exists yetNot comparedNot compared

Signed off at 130 nm

The neuron update block as routed at 130 nm
The neuron update block as routed at 130 nm on the IHP sg13g2 open process, 22,477 cells in 0.235 mm².
0.235 mm²Die area of the block
0DRC violations
MatchedLVS against the schematic
33 MHzClosing frequency

The neuron update block, one part of a core, has completed physical signoff on an open 130 nm process. Nothing above that level has passed its checks yet, and we say so because the difference matters.

Inside N4-Edge*

The N4-Edge floorplan

Neuromorphic Compute Tiles

Brain-inspired cores that react the instant something changes.

On-Chip Model Memory

The model lives right beside the cores that run it.

RISC-V Management Subsystem

Its own processor brings the chip up and keeps it running.

Host Interface

A single link to the device it lives in.

Direct Sensor Interfaces

Sensors wire straight to the chip, with no host in between.

Always-On Power Domain

Stays awake so the rest of the chip does not have to.

The N4-Edge floorplan

Neuromorphic Compute Tiles

Brain-inspired cores that react the instant something changes.

On-Chip Model Memory

The model lives right beside the cores that run it.

RISC-V Management Subsystem

Its own processor brings the chip up and keeps it running.

Host Interface

A single link to the device it lives in.

Direct Sensor Interfaces

Sensors wire straight to the chip, with no host in between.

Always-On Power Domain

Stays awake so the rest of the chip does not have to.

From signal to decision

01

Sensor

A microphone, an accelerometer or a gas sensor produces a continuous signal, and most of it is quiet.

02

Events

The signal is turned into events, so only the moments where something changes carry forward.

03

Spiking network

The trained network sits on the cores, and only the neurons an event reaches are updated.

04

Decision

A result comes out of the chip, and the radio only wakes when there is something worth sending.

700ch 350 0
Spike raster of one spoken digit, 8,216 events across 700 channels over 0.71 seconds
0.00.20.40.6 seconds
One spoken digit from the Spiking Heidelberg Digits test set, 8,216 events on 700 channels over 0.71 seconds. This is the input the chip reads, and the quiet between the events is work it never does.

Inside the design

Blue blocks are built and validated on FPGA hardware. Outlined blocks are designed and not yet built.

Validated on an AMD Xilinx VU47P FPGA in July 2026, 48 cores at 62.5 MHz. N4-Edge is the smaller variant of the same core.

Sensor interfaces

Host interface

Strict request and response

Model memory

Networks load from DDR in 0.94 s

Always-on domain

Neuromorphic fabric

48 cores on FPGA at 62.5 MHz, bit exact at every timestep

Neuron models

LIF, CUBA and adLIF, fixed point

Spike delivery

Weighted, event driven

RISC-V management core

CVA6, firmware drives the chip

Events out

Only when there is something worth sending

Telemetry

Every event counted, none dropped

Debug

On-chip learning

Specifications

Verified on hardware

  • 48 cores at 62.5 MHz on an AMD Xilinx VU47P
  • Full SHD test set run on chip, 90.28 percent, 2,044 of 2,264
  • Bit exact against the reference model at every timestep
  • Zero dropped events across more than 18 million routed
  • Byte identical results across five successive builds

How far it scales

  • 1,024 recurrent adLIF neurons with 20 readouts deployed
  • 1.5 million connections in the deployed network
  • 393,216 neuron fabric matched ideal maths on all 2,264 samples
  • A model loads from memory in 0.94 seconds
  • A RISC-V management core runs the chip from its own firmware

On the way to silicon

  • Neuron update block through physical signoff at 130 nm
  • 0.235 mm² of die, DRC clean, LVS matched, closing at 33 MHz
  • N4-Edge target is 28 nm at 100 to 300 mW
  • Power is a target, we have measured none of it yet
  • First fabricated silicon planned for 2027

Specification in full

Dimension Catalyst N4-Edge, on our FPGA Target for 28 nm silicon
Compute48 neuromorphic cores at 62.5 MHz, three neuron models in fixed point, LIF, CUBA and adLIF32 to 64 cores at 200 to 400 MHz
Neurons24,576 in the validated fabric, 393,216 in the largest build, which matched ideal maths on all 2,264 samples0.5 to 2 million, set by the SRAM on the die
Connections1.5 million resident, 32,768 per core64 to 256 million on the die
Neuron state17 bytes per neuronsame, and it sets the die split
Model storeDDR4, a 1.5 million connection model loads in 0.936 smodel memory on the die, switchable in the field
Models resident32 networks held at once and switched, byte identical per contextsame
Management coreCVA6 RISC-V, firmware drives the chipsame
Host interfacePCIe and MMIO on AWS F2, AXI on the Kria K26SPI, QSPI and AXI, runs with no host
Sensor inputevents injected by the hostAER in, event camera and acoustic front end
Determinismbit exact against the reference model at every timestep, byte identical across five buildscarried forward by construction
Event integrityzero drops across more than 18 million routed eventstelemetry and counters kept as a product feature
Latency69 µs per timestep on an earlier measured configurationunder a millisecond from event to decision
Powerno device measurement exists, and no board exists yet100 to 300 mW running, under 10 mW in the wake domain
Process and packagenot applicable, the design runs on an FPGA28 nm, package not yet chosen
Operating rangenot definedindustrial range is the intent, nothing is qualified
Physical signoffneuron update block at 130 nm on the IHP sg13g2 open PDK, 22,477 cells, 0.235 mm², DRC 0, LVS matched, 33.3 MHznot started
CNN and DSP blocksnone, this is a spiking designnone planned
Toolchaintrain in PyTorch, quantise to int16, check against the bit exact emulator, then place and programsame, plus adaptation on the device

Four generations to N4

N1

Our first spiking processor, validated on FPGA hardware.

N2

Added programmable neuron models and learning on the chip.

N3

Split the design into tiles so it could scale.

N4

Built for the edge, and the design we are taking towards silicon. Everything on this page is about N4.

From trained model to running chip

Train in PyTorch

Spiking networks are trained with surrogate gradients, in a framework engineers already know.

Check in simulation

Our reference simulator matches the hardware exactly, so what works in simulation works on the device.

Run on hardware

The same network runs on FPGA hardware today, and N4-Edge is built to run it unchanged.

Put it in a product

N4-Edge silicon is planned for 2027, so today the same network runs on our FPGA and not yet in a product.

Sources

  • Accuracy, event counts and timing come from our own runs on an AMD Xilinx VU47P FPGA in July 2026, on the full Spiking Heidelberg Digits test set.
  • Power at 100 to 300 mW is a target for 28 nm. No power figure has been measured, and no board exists yet.
  • Intel Loihi 2 at 90.9 percent and the research best at 96.41 percent are published figures, they are not our measurements.
  • The N4-Edge image is an architectural visualisation of the design, it is not a photograph of a manufactured chip.
  • The 130 nm area, DRC, LVS and frequency come from the signoff reports of an open 130 nm flow, and they cover the neuron update block only.

Put intelligence where the signal is

Tell us what you are building, and we will help you put intelligence right where it is needed.

Get in touch