In the era where AI is omnipresent, to help ordinary users and developers better understand and leverage AI, keeping pace with the times, NVIDIA has specifically launched a new column called "AI Decode" on its official blog, with a new post every Monday.
NVIDIA AI Decode Blog Link:
https://blogs.nvidia.com/blog/ai-decoded-rtx-pc
Chinese Version Link:
https://blogs.nvidia.cn/2024/03/06/ai-decoded-rtx-pc
NVIDIA is undoubtedly one of the most qualified giants to talk about AI, having dedicated over a decade to research. The RTX series GPU is the cornerstone of AI PC, bringing AI acceleration to millions of users.
NVIDIA believes that we are currently experiencing an "iPhone moment" for AI, where it is entering ordinary households and becoming part of the daily lives and work of the general public.
NVIDIA's "Decode AI" column aims to introduce new AI hardware, software, and tools on a weekly basis, with a focus on RTX GPUs and their applications in gaming, creative work, development, and productivity. The content will be presented in layman's terms, avoiding complex technical jargon to make AI more accessible to the general audience.
The initial topics include:
Among these, NVIDIA's recent release, RTX Video HDR, stands out. It utilizes AI to supersample streaming videos, transforming low-resolution, low-definition SDR videos into HDR videos with resolutions of up to 4K. With just a single click, it supports nearly all videos played on Chrome and Edge browsers.
Chat With RTX, on the other hand, is a locally run personalized AI chatbot. It's completely free, easy to use, and performs better due to its lack of reliance on cloud services, ensuring privacy is maintained.
In the first article, NVIDIA starts off by breaking down what AI really is – systems that autonomously perform tasks, ingest and analyze massive amounts of data, and then identify patterns within that data.
Generative AI represents a new stage of evolution, capable of rapidly creating new content and iterating based on various inputs such as text, images, audio, animations, 3D models, or other types of data, then generating new content in either the same or different forms.
An AI PC, on the other hand, is a computer equipped with AI-accelerated hardware. This can be in the form of CPU processors, GPU graphics cards, or NPU neural processing units. It can operate locally without relying on cloud connectivity, resulting in faster response times, lower latency, and better privacy and security protection.
Among these, GPU acceleration stands out prominently, especially with RTX GPUs featuring dedicated accelerators like Tensor Cores, which significantly boost AI performance.
The fundamental measure of AI performance is computing power, commonly quantified in units known as TOPS, which represent how many trillion operations can be executed per second.
Today's GeForce RTX 40 series GPUs offer computing power ranging from 200 to 1300 TOPS, spanning both laptops and desktops. In contrast, AI PCs without dedicated GPU cards typically only achieve 10 to 45 TOPS, which is not even in the same ballpark.
Thanks to its powerful computing capabilities and robust ecosystem, NVIDIA RTX GPUs are providing AI acceleration in various application scenarios.
For example, DLSS, targeting gamers, can generate pixels to improve frame rates and image quality. ACE microservices can create lifelike game characters, while RTX Remix can upgrade classic games to 4K high-definition games based on physical rendering features. Currently, there are over 500 RTX-accelerated ray tracing and DLSS games available.
For creators, Adobe Premiere Pro offers enhanced voice functions, and DaVinci Resolve features magic masking, with over 125 AI applications accelerated by RTX.
For developers, AI Workbench enables the rapid creation of AIGC models and LLM large language models. TensorRT allows for swift model optimization, while TensorRT-LLM for Windows accelerates text-based applications, among others.
NVIDIA will gradually unveil these AI applications accelerated by RTX GPUs in the future.
