Powered by Ada Lovelace Architecture, DLSS 3, and DLSS 3. Third-Gen RTX up 4x faster than NVIDIA Ampere Architecture GPUs
NVIDIA delivers Quantum Leap Performance and Introduces New Era in Neural Rendering with GeForce RTX40 Series
NVIDIA today revealed the GeForce RTX(r), 40 Series of GPUs. These GPUs are designed to deliver unprecedented performance for gamers and creators. The flagship GPU, the RTX4090 GPU, is the most powerful, offering up to 4x performance.
The RTX 40 Series is the first GPUs based upon the NVIDIA(r), Ada Lovelace architecture. It offers massive generational leaps of performance and efficiency and represents a new era in real-time ray tracing, neural rendering, and uses AI to generate pixels.
Jensen Huang, NVIDIA founder and CEO, stated that the age of RTX ray tracing is at its peak and that the Ada Lovelace architecture will take it to the next level. He spoke at GeForce(r), Beyond: Special Broadcast at GTC.
Ada is a quantum leap forward for gamers, and it opens the door for those who want to create fully simulated worlds. Ada sets a new standard in the industry with up to 4x performance than the previous generation,” he stated.
DLSS3 Generates Complete Frames for Faster Gaming
Huang also announced NVIDIA DLSSS 3, the next generation of the company’s Deep Learning Super Samling neural-graphics technology. This AI-powered technology generates entire frames to allow for significantly faster game play. It allows the GPU to create entire frames without requiring CPU performance.
This technology will soon be available to some of the most popular game engines in the world, including Unity and Unreal Engine. It has been supported by many of the top game developers around the globe, with over 35 games and apps expected to follow.
The RTX 40 Series GPUs also feature a variety of technological innovations such as:
Streaming multiprocessors up to 83 Teraflops of shader performance — 2x more than the previous generation
Third-generation RT Cores up to 191 efficient ray-tracing teraflops — 2.8x more than the previous generation.
Fourth-generation Tensor Cores up to 1.32 Tensor Petaflops — 5x more than the previous generation with FP8 acceleration
Shader Execution Reordering is a technique that increases execution efficiency. It reschedules shading workloads in real time to make the most of the GPU’s resources. As important an innovation as out of-order execution for CPUs was, SER improves raytracing performance by up to 3x and in game frame rates by up 25%.
Ada Optical Flow Accelerator has 2x faster performance, allowing DLSS 3 predict movement in a scene. This allows the neural network to increase frame rates and maintain image quality.
Adding architectural improvements to TSMC 4N process technology can result in a 2x increase in power efficiency.
Dual NVIDIA Encoders, (NVENC), reduce export times by as much as half and provide AV1 support. OBS, Blackmagic Design DaVinci Resolve and Discord are using the NVENC AV1 encode.
For Immersive Games, RayTracing Technology is Now Available
Ray-traced rendering with accurate lighting in real-time has been a hallmark of graphics for decades. As 3D graphics and games strive to render the best representations of the real-world, the geometric complexity of objects and environments has increased.
It takes a lot of computational power to achieve physically accurate graphics. Cyberpunk 2077, a modern raytraced game, uses over 600 calculations to calculate lighting for each pixel — 16x more than the original raytraced game introduced four years ago.
Two new hardware units have been added to the third-generation RT Cores. They are 2x faster at ray-triangle intersection testing, and offer 2x faster ray-triangle tracing. A Micro-Mesh Engine creates micro-meshes instantly to increase the speed of ray tracing alpha-test geometry. The Micro-Mesh Engine offers the benefits of greater geometric complexity without the storage and performance limitations of complex geometries.
Creativity redefined with RTX Remix and New AV1 Encoders
NVIDIA Studio creators will benefit from the RTX 40 Series GPUs as well as DLSS 3. 3D artists are able to render fully ray-traced environments using accurate physics and realistic material, and can view the changes in real-time, without any intermediaries.
Live streaming and video editing also benefit from the new eighth-generation AV1 encoders and improved GPU performance. Three updates to the NVIDIA Broadcast software developer kit are now available for partners. These include Face Expression Estimation and Eye Contact, as well as quality improvements to Virtual Background.
NVIDIA Omniverse ™, which is part of the NVIDIA Studio suite, will soon include NVIDIA RTX Mix, a modding platform that allows you to create amazing RTX remasters for classic games. RTX Remix allows modders easy to capture game assets, automatically enhance them with powerful AI tools and enable RTX quickly with ray tracing or DLSS.
Portal is RTX ON!
NVIDIA Lightspeed Studios used RTX Remix to reimagine Valve’s legendary video game Portal. It is widely regarded as one the greatest video games ever made. The game has a new look thanks to advanced graphics features like full ray tracing, DLSS 3 and DLSS 3. Portal with RTX will become available as a free, official downloadable content in November for the classic platformer. This release comes just in time to celebrate Portal’s 15th Anniversary.
GeForce RTX 4090 & 4080: The Ultimate GPUs
The RTX 4090 gaming GPU is the fastest in the world, with outstanding power, acoustics, and temperature characteristics. The RTX 4090 with DLSS3 is 4x faster than the RTX 3090 Ti. The RTX 4090 with DLSS 3 is up to 2x faster than the RTX 3090 Ti with DLSS 2. It also consumes 450W less power. It boasts 76 billion transistors and 16,384 cores. Additionally, it has 24GB of high speed Micron GDDR6X micron memory. This ensures that the RTX 4090 delivers more than 100 frames per second in 4K resolution gaming. On Wednesday, October 12, the RTX 4090 will become available starting at $1,599.
The company also revealed the RTX4080, which will be available in two configurations. The RTX4080 16GB is equipped with 9,728 CUDA Cores and 16GB high-speed Micron GDDR6X Memory. It has DLSS 3 and is 2x faster than the GeForce RTX3080 Ti in today’s games. It also features a lower-power GeForce RTX 3990 Ti, which is more powerful than the RTX 4080 Ti. The RTX4080 12GB has 7,680 CUDA Cores and 12GB Micron GDDR6X Memory. It is faster than the RTX3090 Ti, the previous-generation flagship GPU.
Both RTX 4080 configurations are available starting in November at $1,199 or $899 respectively.
Where to buy
As custom boards, the GeForce RTX 4090, 4080 and 4090 GPUs will be made available from top add-in card suppliers such as ASUS and Colorful. These include stock-clocked and factory overclocked models as well as stock-clocked models.
NVIDIA also produces the RTX 4090 (16GB) and RTX 4080 (16GB), in limited Founders editions to fans who want an NVIDIA-designed product.
You will find the GeForce RTX 40 Series GPUs on gaming systems made by Acer, Alienware and ASUS as well as Dell, HP and Lenovo. These are some of the most renowned system builders around, including MSI and MSI.