Intel’s neuromorphic Loihi chip is made accessible through an FPGA board. We will refer to the devices involved in a Loihi model using the following terms.

The Loihi board, which contains one or more Loihi chips.
A Loihi chip, which contains several cores.
A computational unit on a chip. Each chip has several neuron cores, which simulate compartments, synapses, etc. and several Lakemont cores, which are general purpose CPUs for handling input/output and other general tasks.
The FPGA board that the Loihi board is connected to. The host runs a Linux-based operating system to allow programs to interact with the board using drivers provided by Intel.
The PC physically connected to the FPGA board. Typically the superhost and host communicate over ethernet, but it is also possible to communicate over serial USB.
A superhost provided to members of the Intel Neuromorphic Research Community. Whenever we refer to the superhost, you can use the INRC.
Local machine
The computer you are currently using. We usually assume that your local machine is not the superhost, though you can work directly on the superhost.

Nengo Loihi runs on the superhost and will automatically handle the communication with the host and board. Unless you are setting up a new host and board, you will only need to interact with your local machine and the superhost.


If you are setting up a new host or board, see the Board and host page.


Nengo Loihi is a Python package for running Nengo models on Loihi boards. It contains a Loihi emulator backend for rapid model development and easier debugging, and a Loihi hardware backend for running models on a Loihi board.

Nengo Loihi requires the Nengo Python package to define large-scale neural models. Please refer to the Nengo documentation for example models and instructions for building your own models.

Nengo and Nengo Loihi’s emulator backend are pure Python packages that use NumPy to simulate neural models quickly. On your local machine, you only need to install Nengo Loihi and its dependencies, which include Nengo and NumPy. See Installation for details.

Nengo Loihi’s hardware backend uses Intel’s NxSDK API to interact with the host and configure the board. On the superhost, you need to install Nengo Loihi and its dependencies, as well as NxSDK. See Installation for details.

Running models

While you can use most models constructed in Nengo with Nengo Loihi, some models will see degraded performance due to the discretization process used to convert float values to integers for processing on the Loihi chip.

We can recover some of this performance by choosing parameters better suited to the range of values used by the chip. Before you create any Nengo objects, call:


This will change the default parameters for the core Nengo objects, resulting in better performance.

After creating the model, running it on Nengo Loihi is done by replacing:




By default, Nengo Loihi will use the hardware backend if it is available. You can choose to use the emulator even when the hardware backend is installed by doing:

nengo_loihi.Simulator(model, target='sim')

See API reference for additional options and other functions and classes available in Nengo Loihi.