The world's largest neurocomputer simulates a monkey's brain

Chinese engineers have unveiled a new neurocomputer that mimics the brain of a macaque.

The system is called Darwin Monkey and reportedly supports more than 2 billion impulse neurons and more than 100 billion synapses, which close to the number of neurons in the brain of monkeys. Under typical operating conditions, the system consumes about 2 thousand watts of electricity. 

It is noted, that this is the world’s first neuromorphic brain-like computer based on specialized neuromorphic chips. It was developed in the main State Laboratory of Neural Machine Intelligence at Zhejiang University.

The computer is equipped with 960 Darwin 3 neuromorphic computing chips. The system consists of з 15 neuromorphic brain-like servers in the blade style. These are third-generation brain-like neuroprocessors developed by led by Zhejiang University in cooperation with Zhejiang Lab. 

According to information People’s Daily, the array of chips in the system supports more than 2 billion spiking neurons and over 100 billion synapses, making Darwin Monkey the world’s first brain-like computer, that combines advanced mental abilities with vision, hearing, speech, and learning functions.

How to reports Global Times, every chip supports more than 2.35 million spiking neurons and hundreds of millions of synapses, as well as a specialized set of instructions for computing that mimic the brain and an online neuromorphic learning engine. The report also notes, that the researchers have deployed several intelligent applications on the computer. The system can run brain model DeepSeek to perform tasks such as logical thinking, content generation, and math problem solving.

Chips Darwin 3, which makes up the computer, has a special set of instructions inspired by the brain and neuromorphic mechanisms of online learning. Darwin Monkey is the result of breakthroughs in a number of technologies, including improved connectivity and integration of the neural system, as well as the development of a new generation of operating systems inspired by the brain.

In 2024, Intel created Hala Point, the first neuromorphic computing system with 1.15 billion neurons. It was initially deployed at the US Department of Energy-funded Sandia National Laboratories.

Using its powerful neural and synaptic resources, the system can pre-model the brains of animals with different neuronal sizes, including those of elegans, danio-ryo fish, mice, and macaques, opening up new opportunities for research in brain science.

Source: Interesting Engineering