Nvidia Says Its GPUs Can Run Circles Around Intel And AMD NPUs For AI Tasks

Intel and AMD have been promoting neural processing units (NPUs) embedded in their CPUs to handle AI tasks. However, Nvidia is now challenging this narrative, asserting the superiority of GPUs for advanced AI tasks on PCs.

In a recent press conference dubbed “Nvidia RTX for Windows AI,” Nvidia announced its entry into the AI PC market. The company announced how it has been using RTX GPUs for AI applications on PCs for a long time, highlighting features like Nvidia Super Resolution and Deep Learning Super Sampling (DLSS). By introducing ChatRTX, an AI chatbot, Nvidia is stepping up its efforts to create AI-powered PC experiences.

Nvidia claims that its RTX GPUs perform noticeably better than CPU-based NPUs in comparing NPUs and GPUs for AI jobs. According to data presented by the company, gaming PCs with RTX GPUs can deliver up to 1,300 TOPS, whilst CPUs equipped with NPUs can only offer 10-45 TOPS (tera operations per second).

It is worth noting that Nvidia estimates the current user base for AI PCs to be “0 million,” whereas RTX GPU-equipped PCs have 100 million users. Additionally, Nvidia reveals how its GPUs work with more than 500 AI-based apps in various sectors, allowing customers to use AI capabilities without having to wait for CPU updates. While NPUs might be sufficient for simple jobs, Nvidia argues that GPUs’ greater computational capacity is necessary for sophisticated AI development.

Nvidia’s foray into the AI hardware market presents difficulties for rivals that produce CPUs and GPUs. AMD must simultaneously promote the AI capabilities of its CPU and GPU solutions in the face of Nvidia’s advocacy of GPU superiority for AI tasks.