AMD is evaluating the launch of independent NPU products, locking table machines and professional AI PC markets, to complement the current gap in AI computing accelerators. According to the report of CRN, AMD Vice President and Client Business Group...
AMD is evaluating the launch of independent NPU products, locking table machines and professional AI PC markets, to complement the current gap in AI computing accelerators.
According to the report of "CRN", AMD Vice President and Client Business Group General Manager Rahul Tikoo said that the company is studying the launch of an AI engine in the form of independent cards to help promote the goal of "Automatic computing for everyone".
At present, AI PCs are mostly built-in NPUs for CPUs or SoCs, while desktops and professional workstations still mainly rely on GPUs as AI accelerators, which makes GPUs highly load-running and running high power consumption over a long period of time and requires sharing resources with graphics computing, limiting the effectiveness of some AI applications. Independent NPUs can provide dedicated computing performance at low power consumption, reduce GPU and CPU burden and improve overall efficiency.
AMD may plan to use the previous Xilinx technology to expand it to an independent NPU card. The product will optimize AI recommendations and play a role in high-compute scenarios such as generative AI and LLM recommendations.
With Lianying, Dale and HP began to explore PC equipment that equipped with independent NPUs, the market demand for dedicated AI accelerators has increased. At present, this AMD NPU is still under preliminary evaluation, and will continue to observe market development, and combine its own CPU, GPU and NPU technology to step by step to promote AI hardware.
AMD Is Reportedly Looking to Introduction a Dedicated Discrete NPU, Similar to Gaming GPUs But Targeted Towards AI Performance On PCs; Taking Edge AI to New Levels
