At CES 2024, NVIDIA unveiled an array of hardware and software aimed at unlocking the full potential of generative AI on Windows 11 PCs.
Running generative AI locally on a PC is critical for privacy, latency and cost sensitive applications. At CES, NVIDIA is bringing new innovations across the full technology stack to enable the generative AI era on PC. RTX GPUs are capable of running the broadest range of applications, with the highest performance. Tensor Cores in these GPUs dramatically speed up AI performance across the most demanding applications for work and play.
NVIDIA introduced the
GeForce RTX 40 SUPER Series family of GPUs. The GeForce RTX 4080 SUPER generates AI video over 1.5X faster and images over 1.7X faster than the GeForce RTX 3080 Ti. Tensor Cores on SUPER GPUs deliver up to 836 trillion AI operations per second (TOPS) — bringing transformative AI capabilities to gaming, creating and everyday productivity.
This latest iteration of NVIDIA Ada Lovelace architecture-based GPUs delivers up to 52 shader TFLOPS, 121 RT TFLOPS and 836 AI TOPS to supercharge gaming and creating — and provide the power to develop new entertainment worlds and experiences. The GeForce RTX 4070 SUPER starts from $599.
New laptops start shipping later this month from every top Original Equipment Manufacturer (OEM) – including Acer, ASUS, Dell, HP, Lenovo, MSI, Razer, Samsung and more. These systems bring a full set of generative AI capabilities out-of-the-box.
With GeForce RTX SUPER GPUs, users can unlock the full potential of AI on Windows PCs.
A 4K Monster: The GeForce RTX 4080 SUPER
The GeForce RTX 4080 SUPER powers fully ray-traced games in 4K resolution. At 1.4x faster than the GeForce RTX 3080 Ti without DLSS Frame Generation, the RTX 4080 SUPER delivers blistering performance with traditional rasterization. With 836 AI TOPS, DLSS Frame Generation delivers an extra performance boost, making the RTX 4080 SUPER twice as fast as the RTX 3080 Ti. The RTX 4080 SUPER features more cores and faster memory for a performance edge. It will be available starting Jan. 31 from $999.
Precision Gaming: The GeForce RTX 4070 Ti SUPER
The RTX 4070 Ti SUPER is the ideal GPU for maxing out games at super-high frame rates at 1440p, and up to 4K. Compared to the RTX 4070 Ti, it has more cores, an increased frame buffer to 16GB, and a 256-bit memory bus, providing a significant memory bandwidth increase to 672 GB/sec. It is 1.6x faster than a RTX 3070 Ti and 2.5x with DLSS 3. The GeForce RTX 4070 Ti SUPER will be available starting Jan. 24 at $799.
Perfectly Balanced: The GeForce RTX 4070 SUPER
The RTX 4070 SUPER arrives with 20% more cores than the RTX 4070, making it faster than an RTX 3090 at a fraction of the power. With DLSS 3, its lead stretches to 1.5x faster. It will be available starting Jan. 17 at $599.
Where to Buy
For the GeForce RTX 4080 SUPER and 4070 SUPER, an NVIDIA Founders Edition Design will be available direct from NVIDIA.com and select retailers. Custom boards, including stock-clocked and factory-overclocked models for all GeForce RTX 40 SUPER Series GPUs, will be available from top add-in card providers such as ASUS, Colorful, Gainward, GALAX, GIGABYTE, INNO3D, KFA2, MSI, Palit, PNY and ZOTAC.
NVIDIA RTX desktops and mobile workstations, powered by the NVIDIA Ada Lovelace architecture, also deliver the performance necessary to meet the challenges of enterprise workflows.
Mobile workstations with RTX GPUs can run
NVIDIA AI Enterprise software, such as TensorRT and
NVIDIA RAPIDS for simplified, secure generative AI and data science development. A three-year license for NVIDIA AI Enterprise is included with every RTX A800 40 GB GPU, providing a workstation development platform for AI and data science.
AI Workbench, a unified toolkit that allows developers to quickly create, test and customize pretrained generative AI models and large language models (LLM), will be released in beta later this month. It provides developers with the flexibility to collaborate on and migrate projects to any GPU-enabled environment. It also offers streamlined access to popular repositories like GitHub.
Once AI models are built for PC use cases, they can then be optimized to take full advantage of Tensor Cores on RTX GPUs through
NVIDIA TensorRT, a library for high-performance AI inference. NVIDIA recently extended TensorRT to text-based applications by releasing
TensorRT-LLM for an open-source library for accelerating large language models. The latest update to TensorRT-LLM is now available, which adds Phi-2 to the growing list of pre-optimized models for PC, which run up to five times faster compared to other inference backends.
With these new tools and libraries, PC developers are primed to deliver even more generative AI applications on top of the over 500 AI-powered PC games and applications currently accelerated by RTX GPUs.
At CES, NVIDIA and its developer partners are releasing several new generative AI-powered applications and services including
NVIDIA RTX Remix, a platform for creating RTX remasters of classic games. It’s releasing into open beta later this month with generative AI texture tools that transform textures from classic games into modern 4K physically based rendering (PBR) materials.
NVIDIA ACE microservices are also releasing to include generative AI speech and animation models to enable developers to add intelligent, dynamic digital avatars to games. With
Chat with RTX, an NVIDIA tech demo, AI enthusiasts can easily connect PC LLMs to their own data using a popular technique known as
retrieval augmented generation (RAG). It’s accelerated by TensorRT-LLM, enabling users to interact with their notes, documents and other content. It’s also available as an open-source reference project so developers can easily implement the same capabilities into their own applications.
Find out more about these and other announcements from
NVIDIA at CES 2024.
blogs.windows.com
