IBM's NorthPole is faster and uses less energy than Nvidia's best AI GPU

Nov 13, 2023 ·

IBM has used AMD's other not-so-secret weapon to deliver yet another brain-like chip, Northpole. Northpole is faster and uses less energy than Nvidia's best AI GPU, but it is only the beginning. AMD has done a great job of launching GPUs that are competitive with Nvidia's products in both performance and price, and now they have taken it one step further. Northpole is an AI performance powerhouse, delivering up to a 50% increase in neural network training speed compared to Nvidia's best AI GPU. It also offers significantly lower power consumption, meaning that data centers can save on energy costs.

The chip itself is based on a new architecture, RVX, which is designed for deep learning workloads. It features more cores than Nvidia's Pascal-based GPUs, allowing for significant increases in AI performance. It also includes enhanced support for FP16 operations, which allow for deeper calculations, enabling more accurate results when dealing with tasks such as natural language processing.

The release of Northpole is part of AMD's master plan to become a major player in the AI space. They are creating a full solution, from the hardware to the software, to meet the needs of enterprises. This includes their ROCm software platform, giving developers access to optimized frameworks such as TensorFlow and PyTorch.

AMD has also released its Radeon Instinct accelerator card, which is designed to give data centers an edge over their competitors. It comes with pre-trained models, allowing for easy deployment of AI applications without the need for additional coding.

The Northpole chip is yet another example of how AMD is advancing the AI revolution. With more powerful chips and improved software, organizations can leverage AI to create smarter systems and gain a competitive edge. And this is just the beginning; we can expect more powerful chips from AMD in the future.