Skip to main content

Nvidia announces DLSS 4 with Multi Frame Generation

New AI graphics models powered by the transformer architecture found in ChatGPT and Gemini run faster and use less VRAM to generate multiple frames at once.

New AI graphics models powered by the transformer architecture found in ChatGPT and Gemini run faster and use less VRAM to generate multiple frames at once.

Nvidia is revealing a big upgrade to its Deep Learning Super Sampling (DLSS) technology today. DLSS 4 will include new neural rendering capabilities that, on systems with the new RTX 50 Series GPUs, can do Multi Frame Generation, generating “up to three additional frames per traditionally rendered frame, working in unison with the complete suite of DLSS technologies to multiply frame rates by up to 8X over traditional brute-force rendering.”

According to Nvidia, that’s a big enough upgrade to make 4K 240fps, fully ray-traced gaming possible. Also, in an upgrade that will work on all GeForce RTX GPUs, DLSS games with Ray Reconstruction, Super Resolution, and DLAA can be updated to new transformer AI models that use the same tech as AI tools like ChatGPT, which it says is a first for the graphics industry.

Nvidia says its new frame generation model is 40 percent faster and uses 30 percent less VRAM than the old one. By replacing the old Convolutional Neural Networks with its new transformer model, the claim is that generated pixels will have less ghosting, more detail, and more stability in a scene for a clearer picture overall.

Nvidia is promising there will be 75 games and apps that support Multi Frame Generation on day zero, with Alan Wake 2, Cyberpunk 2077, Indiana Jones and the Great Circle, and Star Wars Outlaws supporting it when RTX 50 Series GPUs launch at the end of January, while “over 50 games” will have support for the new transformer-based DLSS features that can run on all GeForce RTX hardware.

We will want to see how the new frame generation tech affects latency, but Nvidia also announced Reflex 2 with Frame Warp technology, which it says can cut latency by up to 75 percent:

As a frame is being rendered by the GPU, the CPU calculates the camera position of the next frame in the pipeline, based on the latest mouse or controller input. Frame Warp samples the new camera position from the CPU, and warps the frame just rendered by the GPU to this newer camera position. The warp is conducted as late as possible, just before the rendered frame is sent to the display, ensuring the most recent mouse input is reflected on screen.

Nvidia’s figures show Reflex 2 cutting latency from 27ms with Reflex 1 to just 14ms in testing with The Finals. It says the tech is “coming soon” to both The Finals and Valorant for gamers with 50 Series GPUs first, before it’s available on other RTX hardware.

For games and apps that don’t have updates available immediately but have DLSS support, the updated drivers and Nvidia app have new DLSS Override switches. Users can force Multi Frame Generation on the new GPUs when frame generation is already turned on in the game, rendering based on the latest models when DLSS is turned on in the fame, and using updated Super Resolution rendering when that feature is turned on in the game.

Here’s a slide showing which DLSS 4 features are available on which series of RTX GPUs:

Nvidia DLSS 4 features and which generation of GPUs support them
Image: Nvidia

DLSS 4 arrives more than two years after Nvidia introduced DLSS 3 with Frame Generation to boost performance in a variety of games. Nvidia then introduced DLSS 3.5 in August 2023 with an AI-powered Ray Reconstruction technique to improve the quality of ray tracing and introduce path tracing (full ray tracing) in select titles. In August last year, Nvidia revealed it now has more than 600 games and applications with RTX support.

Follow topics and authors from this story to see more like this in your personalized homepage feed and to receive email updates.