Nvidia, watch out! IBM has a new analog AI chip that could give the H100 some bang for the buck
There’s every chance IBM just unveiled the blueprint for the future of AI development with an analog AI chip that’s said to be up to 14 times more energy efficient than current leading-edge components.
One of the biggest problems with generative AI is how power-hungry the technology currently is — and will one day become. The costs involved in training models and running the infrastructure will only skyrocket as the space matures. For example, ChatGPT costs over $700,000 per day to run Insider.
IBM’s prototype chip, which the company unveiled in Nature, aims to ease the pressure on companies building and operating generative AI platforms such as Midjourney or GPT-4 by reducing energy consumption.
This is because of the way the analog chip is built; These components differ from digital chips in that they can manipulate analog signals and understand gradations between 0 and 1. Digital chips are the most widespread today, but they only work with different binary signals. There are also variations in functionality, signal processing and application areas.
Nvidia’s chips, including the H100 Tensor Core GPU And A100 Tensor Core GPUare, above all, the components that power many of today’s generative AI platforms. However, should IBM replicate the prototype and get it ready for the mass market, it could one day displace Nvidia as the current mainstay.
IBM claims its 14nm analog AI chip, which can encode 35 million phase-change memory devices per component, can model up to 17 million parameters. The company has also said the chip mimics the way a human brain would work, with the microchip performing calculations directly in memory.
It demonstrated the benefits of using such a chip in several experiments, including one in which a system was able to transcribe audio from people speaking with accuracy very close to digital hardware setups.
The IBM prototype was about 14 times more efficient per watt, though simulations have done so previously shown such hardware could be between 40 and 140 times more energy efficient than today’s leading GPUs.