NVIDIA GeForce RTX 50-Series Gen on Gen Comparison

Unless you’re new to PC gaming, the chances are you either have an existing PC or graphics card that you’re looking to replace. This blog provides a high-level overview of the NVIDIA GeForce RTX 50-series, so you can compare them with your current setup.

GeForce RTX 50-Series Models

There launch line-up includes four models, from the flagship 5090 to the high-end 5080 and 5070 Ti and finally the mid-range 5070.

Model	RTX 5090	RTX 5080	RTX 5070 Ti	RTX 5070
Architecture	Blackwell	Blackwell	Blackwell	Blackwell
Shader Performance (TFLOPS)	106	56	44	29
Ray Tracing Performance (TFLOPS)	318	171	133	94
AI Performance (TOPS)	3,352	1,808	1,406	988
CUDA Cores	21,760	10,752	8,960	6,144
Tensor Cores	680 5th Gen	TBC	TBC	TBC
RT Cores	170 4th Gen	TBC	TBC	TBC
Memory	32GB GDDR7	16GB GDDR7	16GB GDDR7	12GB GDDR7
Memory Controller	512-bit	256-bit	256-bit	192-bit
TDP	575W	360W	300W	250W
Minimum PSU	1,000W	850W	750W	650W
Launch Price	£1,939	£979	£729	£539

The RTX 50-series introduces the Blackwell architecture to consumer GPUs and a number of new or enhanced technologies delivering significant uplifts in performance:

1. DLSS 4 uses transformer-based AI frame generation (as opposed to CNN-based upscaling in previous DLSS versions), to rapidly detect and predict in-game patterns, reducing the computational load and latency, whilst boosting framerates.
2. Reflex 2 builds upon the first iteration of latency reduction by offering 75% faster responsiveness by synchronising CPU and GPU more effectively via a technique called Frame Warp whereby the system records the mouse cursor position and transforms the whole image right before sending it to the screen.
3. Optimisation for neural networks reducing the memory footprint to allow more simultaneous tasks and improved performance.
4. Fifth-generation Tensor cores are optimised for neural rendering and feature a high-speed format - INT4 - which significantly boosts throughput while reducing memory requirements by half.
5. Fourth-generation ray tracing (RT) cores focus on handling mega geometry more efficiently, processing larger and more complex scenes.
6. Improved Shader Multiprocessors (SM) optimised for neural shaders, offering twice the bandwidth and improved throughput for handling complex tasks, especially those involving deep learning and neural shading.
7. Neural Radiance Cache dynamically trains a neural network during gameplay using the GPU, allowing light transport to be cached spatially, enabling near-infinite light bounces in a scene.
8. RTX Neural Faces leverages subsurface scattering—a technique borrowed from film rendering— to simulate light penetrating and diffusing through materials, delivering a subtle glow in skin tones adding depth and realism to characters in motion.
9. GDDR7 memory, a new industry standard that delivers twice the speed of GDDR6X while cutting power consumption in half, and delivering higher bandwidth and improved energy efficiency, addressing two key challenges in high-performance graphics.

All these advancements are supported by 50-series GPUs, so let’s take a look at each of the four graphics cards in turn, comparing them directly with the most popular graphics cards from the last couple of graphics card generations.