The NVIDIA GeForce RTX 4070 Super is an upper mid-range graphics card designed for gamers and content creators. With this powerful GeForce GPU, a smooth gaming experience in WQHD resolution with stunning ray tracing effects is guaranteed.
The AD104 graphics processor is used as the graphics chip of the RTX 4070 Super. The AD104-350, like all GPUs of the RTX 40 generation, is based on NVIDIA's Ada Lovelace architecture and the 4nm manufacturing process. This offers 3rd generation RT cores as well as 4th generation tensor units, streaming multiprocessors for the most realistic ray tracing graphics, and the latest AI features.
NVIDIA's DLSS 3 (Deep Learning Super Sampling) uses the tensor cores of the RTX 4070 Super to maximize performance with uncompromising image quality thanks to AI computing. With NVIDIA's frame generation technology based on DLSS 3, you can experience story games as smoothly as never before. Fully immerse yourself in immersive game worlds with fascinating graphics.
NVIDIA Reflex gives you the decisive competitive advantage thanks to low input latencies and minimizes the reaction times of your hardware with the help of the GeForce RTX 4070 Super.
With 7,680 shader ALUs, 12 GB of graphics memory, and a memory interface of 192 bits with fast 21 Gbps bandwidth, the NVIDIA GeForce RTX 4070 Super is made for gaming in WQHD resolution and with maximum graphics settings.
With the NVIDIA GeForce RTX 4070 Super, Adobe Creative Cloud tools are faster than ever before. Whether you're a digital artist, professional photographer, editor, or motion graphics specialist - with NVIDIA, you benefit from the best features and the strongest performance of the Creative Cloud.
NVIDIA's NVENC encoder is an independent part of the GeForce RTX 4070 Super GPU. This is used to encode videos and relieve the CPU. This allows your system to free up resources for games while streaming, so you can focus on what matters: a truly stunning transmission. The new RTX 4000 graphics cards feature two NVENC encoders with AV1. This allows you to stream 1440p content at the same performance as H.264 but in 1080p. Additionally, the encoder allows for ShadowPlay recordings in 8K60 and HDR.
We offer gaming graphics cards with NVIDIA GeForce RTX 4070 Super GPU from the following manufacturers:
The Ada GPU architecture offers some optimizations over Ampere. The number of CUDA cores per SM (Streaming Multiprocessors) has increased from 84 to 128. An ADA Lovelace SM can perform either 128 FP32 or 64 FP32 and 64 INT32 calculations per workload. Like Ampere, each SM has a ray tracing calculation core.
With 56 SMs, the RTX 4070 Super has 7,168 CUDA cores (each 3,584 FP32 and FP32/INT32 ALUs) as well as 56 ray tracing cores. The 56 SMs of the RTX 4070 Super are divided into 4 GPCs (Graphics Processing Clusters) and each SM has 4 tensor cores. However, the tensor cores in version 4 on the Ada GPUs are much more powerful than on Ampere.
Ada GPUs are equipped with a new technique called Shader Execution Reordering (SER), which can increase ray tracing performance up to three times compared to the Ampere generation. Command workloads are sorted more efficiently. In-game FPS can increase by up to 25%.
Ray tracing (RT) enables realistic lighting by simulating the physical properties and behavior of light. RT is the calculation of the color of individual pixels as the light travels from the user's eye through a 3D landscape. The light traverses the 3D scenery and is reflected or creates reflections on objects. At the same time, shadows, transparent, or semi-transparent objects are created. This results in refractions or distortions.
Ray tracing combines all these effects to represent the final color of a pixel on the monitor. So far, ray tracing has primarily been used in movies to create special effects. Rendering farms are often used for these scenes, and calculating complex 3D animations and special effects for movies can take several weeks. Due to the complexity of ray tracing, PC applications such as computer games have typically been filled with 3D objects based on a raster model. In rasterization, 3D objects are modeled from a network of triangles or polygons. In the rendering pipeline, these triangles of the 3D model are individually converted into pixels on a 2D surface. The pixels can then be further processed ("shaded").
These matrix-based computing cores are equipped with special data paths for optimized floating-point calculation throughput and also offer increased energy efficiency. Above all, training and inference, as components of deep learning, benefit from the tensor cores. Training, i.e., learning new skills based on existing data, is carried out with the tensor computing cores with up to five times higher TFLOP performance than with Ampere.
Want to know which NVIDIA graphics card is right for you? The following list includes all current models of the Ampere generation, sorted by performance and starting with the strongest graphics card.
NVIDIA's GeForce RTX 4070 Super is the perfect graphics card for ultra-smooth WQHD gaming with high graphics settings. Do you want to fully immerse yourself in immersive game worlds? Then you need a compatible WQHD gaming monitor with G-SYNC!
