Numenta has showcased the immense potential of Intel Xeon CPUs in outperforming even the most potent CPUs and GPUs in AI workloads, thanks to a pioneering approach. This remarkable feat, achieved through a set of innovative techniques branded as the Numenta Platform for Intelligent Computing (NuPIC), has opened up new frontiers in AI inference on conventional CPUs, as reported by Serve the Home.
What makes this achievement truly astonishing is that it outperforms CPUs and GPUs specifically tailored for AI inference tasks. A striking example of this prowess lies in Numenta’s comparison with Nvidia’s A100 GPU. Numenta subjected a workload for which Nvidia had reported performance figures using its A100 GPU to an augmented 48-core 4th-Gen Sapphire Rapids CPU. In every scenario, the Intel Xeon CPU outperformed Nvidia’s chip in terms of total throughput. In fact, it raced ahead, clocking in as much as 64 times faster than a 3rd-Gen Intel Xeon processor and an impressive ten times faster than the A100 GPU.
