At a recent RTX AI media briefing hosted by NVIDIA, a plethora of AI-related updates were announced, showcasing the wide-ranging applications of AI in the PC domain: NVIDIA ACE, ChatRTX, Stable Diffusion, DLSS 3.5, NVIDIA app, and a collection of AIGC artworks created by numerous AI artists based on the RTX 40 series AI PC.
NVIDIA accelerated the shift towards AI computing by introducing RTX technology and the first consumer-grade GPU designed specifically for AI, the GeForce RTX, in 2018. An AI PC is a computer equipped with dedicated AI acceleration hardware, and on RTX GPUs, these specialized AI accelerators are known as Tensor Cores. When running demanding workloads and entertainment applications, Tensor Cores can significantly boost AI performance, such as popular generative AI models like Stable Diffusion 1.5 and SDXL. The NVIDIA TensorRT software development kit can enhance deep learning inference performance.
Quantifying AI Performance
One way to measure AI performance is in teraops. Teraops stands for "trillion operations per second" (TOPS), which serves as a single metric for users to gauge a PC's AI performance. The computing power offered by the GeForce RTX 40 series GPUs ranges from 200 AITOPS to 1200 AITOPS. In comparison, lightweight AI PCs without dedicated GPUs currently only offer computing power ranging from 10 to 45 TOPS.
NVIDIA has also labeled the AI performance of different GPUs on its official website for users to compare easily.
The new UL Procyon AI image generation benchmark test now supports TensorRT acceleration. Internal tests show that this benchmark test can accurately reflect current real-world performance. By combining Tensor Cores with TensorRT, local RTX AI PCs can achieve outstanding generative AI performance.
The RTX AI PC has now become an ecosystem with over 100 million users and 500 AI applications.
AI in Gaming
NVIDIA DLSS is a groundbreaking graphics technology that leverages AI to boost frame rates and improve image quality on GeForce RTX GPUs. DLSS 3.5 sets a new standard in gaming, enhancing ray tracing performance with AI. Highly anticipated titles like "Black Myth: Wukong" and "Eternal Cylinder" are set to support DLSS 3.5 for ray-traced reconstruction and full scene ray tracing, delivering significant improvements in both visuals and performance for GeForce gamers, providing an ultimate gaming experience.
NVIDIA RTX Remix allows modders to easily capture in-game assets, enhance textures automatically using generative AI tools, and quickly create stunning RTX remastered games with the help of ray tracing and DLSS technology.
NVIDIA Freestyle will be available through the all-new NVIDIA app Beta version, allowing users to personalize the visual aesthetic effects of over 1200 games through real-time post-processing filters, with features like RTX HDR, RTX Dynamic Vibrance, and more.
NVIDIA Broadcast application can turn any room into a home studio, providing AI-enhanced voice and video tools for live streaming, including noise and echo cancellation, virtual background and AI green screen, face tracking, video noise reduction, and gaze tracking.
NVIDIA ACE brings lifelike NPCs to games
Non-player characters (NPCs) typically play a crucial role in advancing the storyline of video games, but since they are usually designed for specific purposes with pre-set dialogues and actions, they may appear mechanical and uninteresting during interactions, especially in vast game worlds with thousands of NPCs.
NVIDIA has timely introduced NVIDIA ACE, a powerful tool for game developers and digital creators to create lifelike NPCs. Developers can integrate advanced generative AI models into digital virtual characters within games and applications through the ACE microservice. With the ACE microservice, precise speech-to-text conversion can be achieved during interactions, while also providing realistic facial expressions for NPCs. NVIDIA ACE integrates cognitive, perceptual, and behavioral systems to present RTX-rendered characters in meticulously crafted environments, creating an immersive narrative experience.
ChatRTX - Hear more, see more, chat more
ChatRTX utilizes retrieval-augmented generation technology, powered by NVIDIA TensorRT-LLM software and NVIDIA RTX acceleration, to bring chatbot functionality to RTX Windows PCs and workstations. Running locally, ChatRTX allows users to query their notes and documents with LLM support, generating relevant responses quickly based on the queries.
In the recent update of ChatRTX, not only has support for more LLM large models been added, but interactive support for voice and image has also been included.
AI Applications in Creation
SD Professional Workflow: The RTX 4090 D GPU demonstrates true productivity in complex control workflows and image quality, allowing users to generate high-quality studio-level portrait photos from ordinary selfie shots through portrait generation models. Currently, RTX GPUs can provide Stable Diffusion users with a productivity-level experience.
SD generates 8 images per second: As the fastest Stable Diffusion image generation solution on the market, users can enjoy a speed of up to 8 images per second on the RTX 4090 D GPU. By utilizing the fastest Stable Diffusion acceleration method TensorRT, coupled with further performance optimizations by developers, the RTX 4090 D GPU can currently achieve a maximum image generation speed of over 100 images per second.
Voice cloning: Accelerated by RTX GPUs, voice cloning can replicate voices using just a 5-second recording, enabling personalized voice generation without the need for training. Based on the open-source GPT-SoVITS project, voice cloning generates personalized speech segments from the pronunciation features of a 5-second recording sample through text, achieving high-quality voice cloning. With RTX AI acceleration, this process can be completed in just five minutes. Fine-tuning training with more samples (1 minute) can achieve close to a 1:1 replication of voice cloning.
Real-time Architectural Rendering: In the field of architectural design, ZhiAI has achieved second-level AI real-time painting through diffusion models and distillation technology, accelerated by RTX GPUs. By combining ZhiAI's in-house large-scale models with over 500,000 downloads from the domestic architecture industry, architects can now enjoy the epoch-making speed and convenience of AI real-time rendering while still ensuring extremely high output quality.
The Turning Point of AI PCs, RTX Leading the Way
There is no doubt that the development of AI PCs is changing our daily lives and work efficiency, with the entire industry embracing AI becoming a reality. In the field of AI acceleration, NVIDIA stands out with its RTX GPUs, holding a supreme position. Based on RTX GPUs, AI application computing acceleration not only excels in its own right but also boasts a wide range of application ecosystems. It can be said without hesitation: RTX AI PCs will bring an AI-driven productivity revolution to various industries!
