IBM's Cognitive Computing Chips: Further Reading

Earlier today IBM announced an experimental computer chip in which the computational elements and RAM are wired together much closer together than standard CPUs available today. IBM has made two prototypes of the new chip, which it calls a “neurosynaptic core.” Both are built on a standard semiconductor platform with 256 “neurons,” the chip’s computational components. RAM units on the chip act as synapses; one of the chips has 262,144 synapses, while the other has 65,536.

Nature magazine has a run down of what is new about theses chips, what they propose to achieve here.

To understand what makes this approach different you might want to read more about about the current CPU archecture model: Von Neumann, or stored-program architecture (wikipedia). The current model has an inherit bottleneck (wikipedia).

Also here is IBM's official research blog post about the announcement and they plan to release further details at the IEEE Custom Integrated Circuits Conference on September 20 in San Jose, California.

 

Featured Product

Darsi Pro - AI Compute Box based on NVIDIA® Jetson Orin™ NX

Darsi Pro - AI Compute Box based on NVIDIA® Jetson Orin™ NX

Darsi Pro is e-con Systems' rugged Edge AI compute Box powered by NVIDIA® Jetson Orin™ NX, designed for Mobility, Robotics, ITS, and physical AI applications. It delivers high-performance AI compute, supports e-con's broad camera ecosystem, and enables synchronized multi-sensor fusion for real-time perception. The Darsi Pro GMSL variant supports up to eight synchronized GMSL cameras and is compatible with NVIDIA® JetPack 6 and higher. Built with fanless design and industrial-grade enclosure, Darsi Pro is engineered for long-duty operation in harsh environments. It integrates with CloVis Central™- e-con's cloud-based device management platform for remote health monitoring, and secure OTA updates—simplifying large-scale deployment.