Intel introduces neuromorphic Loihi 2 processor on Intel 4 process

Intel has presented Loihi 2, a new version of its neuromorphic processor that can simulate neurons and synapses and is thus self-learning to some extent. The company uses Intel 4 as a test process to make the chip.

According to Intel, the basic architecture of Loihi 2 is the same as that of the first Loihi. Intel introduced this neuromorphic chip in 2017. Intel makes it at 14nm, where Loihi 2 is produced on Intel 4, or Intel’s upcoming 7nm process. For the time being, this concerns pre-production capacity that Intel uses since Intel 4 is not yet production ready.

Despite the same basic architecture, according to Intel, there are the necessary improvements. The use of the smaller node means that the cores of the chip take up a smaller surface. It is still a maximum of 128 neuron cores per chip, but according to Intel, the effective core capacity has increased through more flexible and efficient memory management and many more neurons can be simulated with those cores.

Support for neuron models has also been expanded. A neuron model involves a short series of microcode instructions that describe the behavior of a single neuron. Loihi 1 was optimized for a single spiking neural network but version two allows for programmable pipeline neuron models. The neuron cores communicate with each other with spike messages and with the new version of the chip this is possible with a 32-bit spike payload for which hardware acceleration is present so that this does not cost more energy.

Intel has doubled the number of processors per chip from three to six. These processors are for traditional processing of C and Python and with the previous Loihi chip, the three processors often formed a bottleneck, Intel admits.

Furthermore, the possibilities to communicate with the outside world have been greatly expanded. Instead of a proprietary interface to communicate with other systems, Loihi 2 supports various Ethernet standards, GPIO and SPI, among others.

Intel is currently making two different Loihi 2 systems available to research institutions through its Neuromorphic Research Cloud. The first is the Oheo Gulch system with single neuromorphic chip and access via an Arria 10 fpga and ethernet. The second is not yet available. This is Kapoho Point, a compact system with a 4×4″ form factor and eight Loihi 2 chips and Ethernet and GPIO interface. Multiple Kapoho Point systems can be connected together.

In neuromorphic computing, manufacturers focus on systems that, just like with human brains, can strengthen connections in the network in order to ‘learn’ and process data more efficiently. Loihi is a network-on-a-chip that can be used to perform certain computational tasks faster and more economically. To simplify writing applications that can run on neuromorphic platforms, Intel makes available Lava, an open source framework that is platform independent.