From CES in Las Vegas comes new announcements from NVIDIA about the RTX 50 series of GPUs and a new ARM mini PC called DIGITS.
The new NVIDIA Geforce RTX 50xx Blackwell graphics cards surprise us with their rather moderate starting price for the entry-level range, about 654 € in Europe and $600 in the USA. In addition, their power consumption is quite low compared to what was rumored. They feature DLSS 4 with Multi Frame Generation technology that further improves their performance.
An ARM mini-PC with NVIDIA processor called DIGITS has also been presented, a machine with a price of about $3000 that is intended for AI work.
Press Release
NVIDIA Geforce RTX 50
The new NVIDIA Geforce RTX 50 series debuts DLSS 4 multiframe technology to boost frame rates by using AI to generate up to three frames per rendered frame. It works in combination with the DLSS suite of technologies to increase performance by up to 8x over traditional rendering, while maintaining responsiveness through NVIDIA Reflex technology.
DLSS 4 also introduces the first real-time application of the transform model architecture in the graphics industry. Transformer-based DLSS super-resolution and ray reconstruction models use 2x more parameters and 4x more computational power to provide increased stability, reduced ghosting, increased detail, and improved anti-aliasing in gaming scenes. DLSS 4 will support GeForce RTX 50 Series GPUs in more than 75 games and applications on launch day.
NVIDIA Geforce RTX 50 Pricing and Consumables
- RTX 5090 = $1,999 – 575 W
- RTX 5080 = $999 – 360 W
- RTX 5070 Ti = $749 – 300 W
- RTX 5070 = $549 – 250 W
NVIDIA Reflex 2 introduces Frame Warp, an innovative technique to reduce gaming latency by updating a rendered frame based on the last mouse input just before it is sent to the display. Reflex 2 can reduce latency by up to 75%. This gives gamers a competitive advantage in multiplayer games and makes single-player titles more responsive.
Blackwell introduces AI in shaders
Twenty-five years ago, NVIDIA introduced GeForce 3 and programmable shaders, which laid the groundwork for two decades of graphics innovation, from pixel shading to computational shading to real-time ray tracing. Along with the GeForce RTX 50 series GPUs, NVIDIA is introducing RTX neural shaders, which incorporate small AI networks into programmable shaders, unlocking cinematic-quality materials, lighting, and more in real-time gaming.
Rendering game characters is one of the most difficult tasks in real-time graphics, as people are prone to notice the smallest errors or artifacts in digital humans. RTX Neural Faces takes as input a simple raster face and 3D pose data, and uses generative AI to render a high-quality, temporally stable digital face in real time.
RTX Neural Faces is complemented by new RTX technologies for hair and skin ray tracing. Together with the new RTX Mega Geometry, which enables up to 100 times more ray-traced triangles in a scene, these advances are poised to deliver a huge leap in the realism of game characters and environments.
The power of neural rendering, DLSS 4 and the new DLSS transformer model is showcased on GeForce RTX 50 Series GPUs with Zorah, an innovative technology demonstration from NVIDIA.
NVIDIA Puts Grace Blackwell on Desktop with DIGITS
NVIDIA’s DIGITS project with the new GB10 superchip is touted as the world’s smallest AI supercomputer capable of running 200,000-parameter models.A personal AI supercomputer that gives researchers, data scientists and students around the world access to the power of the NVIDIA Grace Blackwell platform.
Project DIGITS incorporates the new NVIDIA GB10 Grace Blackwell superchip, which delivers a petaflop of AI computing performance for prototyping, tuning and running large AI models.
With Project DIGITS, users can develop and run inference on models using their own desktop system, and then seamlessly deploy the models in accelerated cloud or data center infrastructures.
GB10 a petaflop of energy-efficient AI performance
The GB10 Superchip is a system-on-a-chip (SoC) based on the NVIDIA Grace Blackwell architecture and delivers up to 1 petaflop of FP4-accurate AI performance.
GB10 incorporates an NVIDIA Blackwell GPU with next-generation CUDA® cores and fifth-generation Tensor Cores, connected via NVLink®-C2C chip-to-chip interconnect to a high-performance NVIDIA Grace™ CPU, which includes 20 low-power cores built on the Arm architecture. MediaTek, a market leader in Arm-based SoC design, has collaborated on the GB10 design, contributing to its best-in-class power efficiency, performance and connectivity.
The GB10 Superchip enables Project DIGITS to deliver powerful performance using only a standard electrical outlet. Each Project DIGITS features 128 GB of unified, coherent memory and up to 4 TB of NVMe storage. With the supercomputer, developers can run large language models of up to 200 billion parameters to power AI innovation. In addition, using the NVIDIA ConnectX® network, two Project DIGITS AI supercomputers can be connected to run models of up to 405 billion parameters.
Grace Blackwell AI supercomputing at your fingertips
With the Grace Blackwell architecture, enterprises and researchers can prototype, tune, and test models on local Project DIGITS systems running Linux-based NVIDIA DGX OS, and then seamlessly deploy them on NVIDIA DGX Cloud™, accelerated cloud instances, or data center infrastructure.
This allows developers to create AI prototypes on Project DIGITS and then scale them on cloud or data center infrastructure, using the same Grace Blackwell architecture and NVIDIA AI Enterprise software platform.
Project DIGITS users can access an extensive library of NVIDIA AI software for experimentation and prototyping, including software development kits, orchestration tools, frameworks and models available in the NVIDIA NGC catalog and on the NVIDIA developer portal. Developers can tune models with the NVIDIA NeMo™ framework, accelerate data science with NVIDIA RAPIDS™ libraries, and run common frameworks such as PyTorch, Python, and Jupyter notebooks.