IBM Touts Analog-Digital Hybrid Chip for AI Inferencing

Render for IBM's 64-tyle hybrid chip
(Image credit: IBM Research)

IBM, which has been at the forefront of quantum computing and a number of other research fields, recently showcased what it feels the solution to AI processing (and its costs) could be. And if IBM's vision translates into something, the future isn't centered around GPUs: instead, it takes place within mixed-signal, analog chips that could bring about massive improvements in energy efficiency while offering competitive performance against the market's current go-tos.

According to a research paper published in Nature Electronics last week, IBM believes the future of AI inferencing could pass through a chip combining phase-change memory (PCM) alongside digital circuits. According to the paper, matrix-vector multiplication (one of the main workloads for AI inferencing) could be performed directly on chip-stored weights.

Analog chips that break apart the power efficiency barriers would certainly be a welcome move, but as with any new technology, analog AI inferencing chips will have to fight to survive against the already-entrenched technologies, software stack, and techniques deployed today. Network effects and market share are real, and Nvidia's grip on the HPC market through both its hardware and CUDA software stacks is... vice-like, to say the least.

TOPICS
Francisco Pires
Freelance News Writer

Francisco Pires is a freelance news writer for Tom's Hardware with a soft side for quantum computing.