ComputeX 2024: GeForce RTX AI PC, NVIDIA brings AI assistants to life

June 2, 2024, on the eve of the opening of COMPUTEX 2024, NVIDIA announced the launch of a brand-new NVIDIA RTX technology to support AI assistants and digital humans running on new GeForce RTX AI laptops. At the same time, NVIDIA also released a series of updates to AI technology, as well as some new laptops, GPUs, and other products. Let's take a look at them together.

AI Assistant, G-Assist

Today, NVIDIA officially launched the G-Assist project, an AI assistant technology demonstration that is supported by RTX and provides context-aware assistance for PC gaming and applications. The technical demonstration of the G-Assist project will make its debut in Studio Wildcard's "ARK: Survival Ascended." NVIDIA also released the first PC-based NVIDIA NIM inference microservice, specifically built for the NVIDIA ACE digital human platform.

Advertisement

These technologies are supported by the NVIDIA RTX AI Toolkit—a new set of tools and SDKs that help developers optimize and deploy large generative AI models on Windows PCs. They join NVIDIA's full-stack RTX AI innovations, accelerating more than 500 applications and games, as well as laptop designs from 200 OEM partners.

NVIDIA Vice President of Consumer AI Business, Jason Paul, said: "NVIDIA launched the RTX GPU with Tensor Core and DLSS technology in 2018, ushering in the era of AI PCs. Today, with the G-Assist project and NVIDIA ACE, NVIDIA will empower over 100 million RTX AI PC users to unlock a new generation of AI experiences."

PC games offer vast game worlds to explore and complex mechanics for players to master, which can be extremely challenging and time-consuming even for the most seasoned players. The G-Assist project aims to use generative AI to make game knowledge readily available to players.

The G-Assist project can receive players' voice or text input, contextual information from the game window, and process data through AI vision models. These models enhance the context awareness and understanding of specific games and applications of the large language model (LLM) connected to the game knowledge database, then generate customized responses that are delivered to users in text or voice form.

NVIDIA and Studio Wildcard have collaborated to showcase this technology in a demonstration demo for "ARK: Survival Ascended." The G-Assist project can help answer questions about creatures, items, background knowledge, missions, and level bosses. Due to its game environment perception and contextual understanding capabilities, the G-Assist project can provide personalized responses based on the player's progress in the game.

The G-Assist project can also configure players' gaming systems for optimal gaming performance and energy efficiency. It can provide an in-depth understanding of performance metrics, optimize graphics settings based on the player's hardware, perform safe overclocking, and even intelligently reduce power consumption while maintaining performance goals.The first ACE PC NIM debuts

NVIDIA's ACE technology, designed to empower digital humans, is now being applied to RTX AI PCs and workstations through NVIDIA NIM. NVIDIA NIM's inference microservices enable developers to reduce deployment time from weeks to minutes. ACE NIMs can run high-quality inference locally on devices for natural language understanding, speech synthesis, facial animation, and more.

During COMPUTEX, the gaming debut of NVIDIA ACE NIM on PCs will be showcased in the Covert Protocol technology demo, developed in collaboration with Inworld AI, demonstrating NVIDIA Audio2Face and NVIDIA Riva automatic speech recognition technologies running locally on PCs.

Adding GPU acceleration to local PC SLM

Microsoft and NVIDIA are collaborating to help developers apply new generative AI capabilities to Windows native and web applications. The partnership will provide application developers with convenient APIs to access GPU-accelerated small language models (SLMs) and support retrieval-augmented generation (RAG) features driven by Windows Copilot Runtime running on devices.

Small language models offer vast possibilities for Windows developers, including content summarization, content generation, and task automation. RAG capabilities allow AI models to access domain-specific information not fully represented in the base model to enhance SLMs. Developers can utilize application-specific data sources through the retrieval-augmented generation (RAG) API and tailor the behavior and functionality of SLMs according to application needs.

NVIDIA RTX GPUs and AI accelerators from other hardware manufacturers will accelerate these AI capabilities, providing end-users with fast, responsive AI experiences across the entire Windows ecosystem. It is reported that the API will be released as a developer preview later this year.

RTX AI Toolkit Launched, Model Speed Increased by 4x, Size Reduced by 3x

The AI ecosystem has built hundreds of thousands of open-source models for application developers to use, but most models are pre-trained for general purposes and built to run in data centers.To assist developers in building AI models for specific applications that run on PCs, NVIDIA has introduced the RTX AI Toolkit, a suite of tools and SDKs for customizing, optimizing, and deploying models on RTX AI PCs. The RTX AI Toolkit is set to be released in June for broader developer access.

Developers can utilize the open-source QLoRa tool to customize pre-trained models and employ NVIDIA TensorRT™ Model Optimizer for model quantization, reducing RAM consumption by up to three times. NVIDIA TensorRT Cloud will optimize models to achieve peak performance on RTX GPU products, with performance improvements of up to four times compared to pre-trained models.

The NVIDIA AI Inference Manager (AIM) application development toolkit (SDK) is now available in a preview version, perfectly orchestrating AI inference between PCs and the cloud, simplifying the complexity of AI integration for PC application developers. It also preconfigures necessary AI models, engines, and dependency packages for PCs in a unified NIM format and supports all mainstream inference backends, including TensorRT, DirectML, Llama.cpp, and PyTorch-CUDA across different processors (GPU, NPU, and CPU). Software partners such as Adobe, Blackmagic Design, and Topaz have integrated components of the RTX AI Toolkit into their popular creative applications to enhance AI performance on RTX PCs.

Deepa Subramaniam, Vice President of Marketing for Adobe Creative Cloud Products, said: "Adobe and NVIDIA will continue to collaborate to provide breakthrough customer experiences for all creative workflows, including video, image, design, and 3D. TensorRT 10.0 on RTX PCs offers creators, designers, and developers unprecedented performance and AI-driven features, opening up new creative possibilities for content creation in industry-leading creative tools like Photoshop."

Components of the RTX AI Toolkit, such as TensorRT-LLM, have been integrated into popular developer frameworks and applications for generative AI, including Automatic1111, ComfyUI, Jan.AI, Langchain, LlamaIndex, Oobabooga, and Sanctum.AI.

AI for Content Creation

Last year, NVIDIA introduced RTX acceleration with TensorRT for Automatic1111, one of the most popular Stable Diffusion user interfaces. Starting this week, RTX will also accelerate the widely popular ComfyUI, with performance improvements of up to 60% compared to the current version, and up to seven times the performance compared to the MacBook Pro M3 Max.

NVIDIA RTX Remix is a MOD platform that utilizes panoramic ray tracing technology, NVIDIA DLSS 3.5, and physically accurate materials to remaster classic DirectX 8 and DirectX 9 games. RTX Remix includes a Runtime renderer and the RTX Remix toolkit application, simplifying the modification process of game assets and materials.

Since its launch earlier in 2024, more than 20,000 modders have used this toolkit to remaster classic games, developing over 130 RTX remastered games on the RTX Remix Showcase Discord.In June this year, NVIDIA will open-source the RTX Remix toolkit, allowing modders to simplify the replacement of assets and the relighting of scenes, increase the file formats supported by the RTX Remix asset ingestor, and enhance the AI texture tools of RTX Remix with new models.

Additionally, NVIDIA has made the functionality of the RTX Remix toolkit accessible through a REST API, enabling modders to link RTX Remix in real-time with digital content creation tools such as Blender, mod creation tools like Hammer, and generative AI applications like ComfyUI. NVIDIA also provides an SDK for the RTX Remix Runtime, allowing modders to deploy the RTX Remix renderer to applications and games beyond the classic versions of DirectX 8 and DirectX 9.

RTX Video, Super Clear Video

The popular AI super-resolution feature NVIDIA RTX Video is now open to all developers as an SDK and can be used in Google Chrome, Microsoft Edge, and Mozilla Firefox browsers. It assists developers in natively integrating AI for sampling, sharpening, reducing compression artifacts, and high dynamic range (HDR) conversion.

Video editing software such as Blackmagic Design's DaVinci Resolve and Wondershare Filmora will soon support RTX Video, enabling video editing users to upgrade low-quality video files to 4K resolution and convert SDR source files to HDR video. Furthermore, the free media player VLC media is set to add support for RTX Video HDR on top of its existing super-resolution capabilities.

FF-Ready Enthusiast GeForce Card Launched

Compact Small Form Factor PC Layout

At ComputeX, NVIDIA also officially launched the SSF-Ready Enthusiast GeForce Card, a member of the previously advocated SSF small form factor PC new ecosystem.The SSF-Ready Enthusiast GeForce Card has a compact size of 304mm × 151mm × 50mm, designed for assembling compact small-form-factor PCs. Currently, core AIC manufacturers such as Galaxy, Gigabyte, and Zotac have launched or plan to launch SSF-Ready Enthusiast GeForce Cards spanning various tiers from the RTX 4070 to the RTX 4080 SUPER.

Additionally, in terms of compatible case products, manufacturers including Thermaltake, SilverStone, NZXT, InWin, and Lian Li have also introduced SS-Ready case products.

On the eve of the opening on June 2nd, during the traditional pre-heating keynote speech, NVIDIA's focus is undoubtedly to showcase AI in its full glory. The new AI ecosystem built around the RTX GPU is also moving towards a more mature future. As ComputeX 2024 unfolds, RTX AI is expected to be a major highlight of the exhibition. "Micro Computer" will continue to follow the progress of this computer show and bring you the latest IT hardware technology news in a timely manner, so stay tuned.

Comment