On May 1st, both Intel and AMD surprisingly agree on what defines an AI PC, emphasizing the necessity of a trifecta of CPU, GPU, and NPU, especially advocating for the integration of new NPUs. However, NVIDIA, as a graphics card manufacturer, takes a different stance, asserting that only PCs equipped with powerful RTX GPUs can truly be considered AI PCs.
The NVIDIA RTX GPUs, introduced in 2018, have made their way into various industries over the years, offering broad application acceleration in areas such as gaming, content creation, multimedia, productivity, development, and everyday life.
In terms of computing power, NPUs are mainly designed for sustained low-load AI tasks, such as processing effects in Windows video conferences, and currently can only achieve 10-45 TOPS.
The computing power of RTX graphics cards can exceed 100 TOPS, with the highest reaching over 1300 TOPS, making them easily handle any AI workload. For instance, even the RTX 4050 can outperform Apple's M3 for large language models, and the competitive flagship RX 7900 XTX still falls short compared to the RTX 4070 SUPER.
In terms of ecology, NVIDIA is particularly proud. RTX graphics cards now support AI acceleration for over 500 games and applications, providing a comprehensive and powerful development platform.
Among them, RTX AI powers over 370 games, utilizing features like DLSS 3 super resolution, RTX Remix game enhancements, ACE digital humans, all benefiting from RTX AI acceleration. Notably, DLSS can deliver up to a 4x performance boost.
RTX AI applications extend to more than 125, spanning image editing, video editing, 3D design, live streaming, and various unique AI SDKs, offering up to a 10x performance increase.
For instance, Stable Diffusion sees a 7x speed boost in image generation, Autodesk Arnold rendering speeds increase by 6x, DaVinci Resolve video editing speeds improve by 2.5x, Adobe Premiere Pro audio editing speeds enhance by 4.5x, and so on.
