\u201cNorthPole outperforms all prevalent architectures, even those that use more-advanced technology processes.\u201d<\/p><\/blockquote>\n
One of the major impediments to improving AI processing is called the \u201cvon Neumann bottleneck.\u201d Using currently available architecture, AI chips tend to have faster processing capabilities than the memory they require to run processes. As a result, latency is introduced whenever information is sent between the processing unit and random access memory.<\/p>\n
This is especially true at \u201cthe edge,\u201d where chips and data are stored together. Removing this bottleneck has long been considered by many experts to be the key to running powerful neural networks locally on devices. <\/p>\n
According to IBM Research, the new prototype chip built in the company\u2019s Alamaden, California laboratory bypasses the von Neumann bottleneck by, essentially, integrating the memory component onto the processing chip itself. <\/p>\n
As the chip\u2019s lead developer, Dharmendra Modha, puts it, NorthPole is \u201can entire network on a chip\u201d that \u201cforges a completely different path from the von Neumann architecture.\u201d<\/p>\n
![](\"https:\/\/s3.cointelegraph.com\/uploads\/2023-10\/7535f1e5-9730-4bf1-a61e-65b4ddc430c3.jpg\")
<\/p>\n
The benchmark used to demonstrate the chip\u2019s effectiveness, ResNet50, is a 50-layer neural network primarily used to test computer vision tasks such as image classification. <\/p>\n
The NorthPole hardware\u2019s reported results on this benchmark indicate that it could perform exceptionally well at associated tasks such as autonomous surgery, operation of self-driving cars and other vehicles, and numerous robotics-related endeavors. <\/p>\n
IBM Research is already years into research on the next chip using the NorthPole architecture. According to the company blog, \u201cthis is just the start of the work for Modha on NorthPole.\u201d <\/p>\n