NVIDIA today presented different proposals and technologies focused on bringing generative AI to users of computers powered by NVIDIA GeForce RTX, both in laptops and desktops, as well as development tools to optimize the use of AI models on NVIDIA technology.
<
Table of Contents
div class="wp-block-image">
NVIDIA has become the quintessential AI company. Slowly but surely, it has been bringing AI technology closer to consumers, after taking over the data center and server market with its professional solutions. Today, NVIDIA announced its new NVIDIA RTX technology to power both AI assistants and digital humans on new GeForce RTX AI laptops.
Project G-Assist is a technology demonstration that runs on RTX graphics, offering contextual help in PC games and apps. This demo has debuted with ARK: Survival Ascended, from Wildcard Studio. It also introduced the first NVIDIA NIM inference microservices (NVIDIA Inference Manager) for the NVIDIA ACE digital human platform.
These technologies are made possible by the NVIDIA RTX AI Toolkit, a suite of tools and SDKs that help developers optimize and deploy large generative AI models on Windows PCs.
In addition, RTX AI PC laptops from ASUS and MSI with GPUs up to the RTX 4070 and efficient SoCs with Windows 11 AI PC features have been announced. These teams will receive a free upgrade to the Copilot+ PC experiences when it becomes available.
The era of AI PCs began in 2018, with the advent of RTX Tensor Core GPUs and DLSS technology. Now, with Project G-Assist and NVIDIA ACE, new possibilities are opening up in AI experiences for more than 100 million RTX AI PC users.
Project G-Assist, an AI assistant based on NVIDIA GeForce technology
AI assistants are on their way to transforming gaming and app experiences, and Project G-Assist is a preview of what’s to come. Project G-Assist uses voice or text as an input method, while adding contextual information from the game screen. With all this information, tailor-made responses are generated, both in the form of voice and text.
NVIDIA has teamed up with Wildcard Studio to develop the technology demo for ARK: Survivla Ascended. Project G-Assist can answer questions about creatures, objects, targets, and more. In addition, Project G-Assist can configure the game for optimal performance by modifying graphics settings, overclocking, or reducing power consumption without compromising desired performance.
The first ACE PC NIM arrives
NVIDIA ACE technology to move digital humans is coming to RTX-enabled AI PCs and NVIDIA NIM (NVIDIA Inference Microservices) workstations that enable developers to reduce deployment time from weeks to minutes.
ACE NIMs offer high-quality inferences, running locally on devices, with the ability to understand language, generate speech or perform facial animations, among other skills.
The debut of NVIIDA ACE NIM on PC will come in the Covert Protocol tech demo, in collaboration with Inworld AI. At the moment, it shows the NVIDIA Audio2Face and NVIDIA Riva functionalities for automatic speech recognition, running locally on the devices.
Windows Copilot Runtime Adds GPU Acceleration to SLM Models Running Locally on PCs
Microsoft and NVIDIA are collaborating to help developers bring new generative AI capabilities to native Windows apps and web apps. Programmers will have access to easy-to-use APIs that allow Small Language Models (SLMs) to use GPU technology to accelerate inference and integrate RAG (retrieval-augmented generation) skills into Windows Copilot Runtima computers.
RAG allows you to fine-tune inferences from local data or specific user-defined data, just like Chat with RTX does now.
4x faster and 3x smaller models with the RTX AI Toolkit
The AI ecosystem already has thousands of open-source models, but most have been trained for general-purpose tasks and designed to run in data centers.
To bring these models to PCs, NVIDIA has introduced the RTX AI Toolkit, a suite of tools and SDKs for tuning models and optimizing and deploying them on RTX AI PCs. This toolkit will be available in June.
Developers can customize a pre-trained model with open-source QLoRa tools. They can then use the NVIDIA TensorRT Model Optimizer to take up to three times less RAM and run at up to 4 times faster than the pre-trained model.
The NVIDIA AI Inference Manager SDK is available in a preview version, simplifying the deployment of ACE on PCs. It preconfigures the PC with AI models and dependencies, orchestrating inferences in a transparent way between the PC and the cloud.
Software partners such as Adobe, Blackmagic Design, and Topaz are integrating components of the RTX AI Toolkit into their apps to accelerate AI performance on the PC.
AI for Content Creation
NVIDIA is integrating RTX AI acceleration into apps for creators, modders, and video enthusiasts. Last year, NVIDIA introduced RTX acceleration using TensorRT for one of Stable Diffusion’s most popular interfaces, Automatic1111. This week, RTX will accelerate ComfyUI, with up to a 60% improvement in performance and 7x better than the M3 Max MacBook Pro.
NVIDIA RTX Remix is the platform for game modding, featuring classic DirectX8 and DirectX 9 game remastering technologies with ray tracing, NVIDIA DLSS 3.5, and material physics improvements.
RTX Remix Runtime is open-source and has been used by 20,000 modders, with over 130 remasters in development on the RTX Remix Showcase Discord. Now, NVIDIA has made the RTX Remix Toolkit open-source as well. And it’s making the RTX Remix Toolkit’s functionalities accessible via a REST API, allowing integration with tools like Blender, Hammer, or ComfyUI.
NVIDIA RTX Video is a video upscaling feature that is compatible with Google Chrome, Microsoft Edge, and Mozilla Firefoz. It is now available as an SDK for all developers. Before long, Blackmagic Design’s DaVinci Resolve and Wodershare Filmora will come with RTX Video to upscale low-resolution content to 4K and to convert SDR to HDR.
In addition, VLC media player will soon add RTX Video HDR to the existing super-resolution functionality.
Comments