It’s the beginning of CES 2025 and NVIDIA kicked it off with the announcement of the brand new generation of GeForce GPUs – the GeForce RTX 50-series. This highly-anticipated family of GPUs come with a whole lot of new features – and of course, more AI. While Jensen Huang, the CEO of NVIDIA, didn’t go into detail on what the new GeForce RTX 50-series will bring, we’re going to do that here instead.
New GeForce RTX 50-series family and prices
For this generation of the GeForce RTX 50-series GPUs, right off the bat – there will be 4 variants already. The GPU models and prices are as shown below, in USD currency.
We all know many people are curious about how the prices of each generation of NVIDIA GeForce GPUs change, so here’s a chart summarizing everything.
Raw specs
Let’s start with the raw specs of each of the GPUs first, as listed down below. The GeForce RTX 50-series are using the latest GDDR7 memory and that will further increase the memory bandwidth to 1.81TB/s. The amount of VRAM of the RTX 5090 has also increased (finally!) as AI applications demand a lot of VRAM, but other variants will remain the same as before.
DLSS 4
DLSS is the umbrella term that encompasses the resolution upscaler, the frame generation technology, ray-reconstruction, and also their own proprietary anti-aliasing method. This new generation of DLSS 4 brings a whole new set of features – but we’re going to highlight the mult-frame generation. Thanks to the new algorithms that they have on the Blackwell architecture, it can generation up to 3 frames ahead of time. Yes, ahead. Those are future frames.
Many current games can also use the new DLSS 4’s multi-frame generation. Just head into the NVIDIA App and you can select this option for that particular game.
The day 0 launch will have around 75 games and apps supporting this new DLSS 4 multi-frame generation, so we can already start enjoying it – if we have the supported hardware, of course. NVIDIA also publicly made this handy chart available – which I appreciate.
The GeForce RTX 40-series GPUs will have enhanced single-frame generation. By leveraging the new technologies used in the new multi-frame generation, it can generate a frame faster and uses less VRAM.
As for the upscaler technology, it is now using transformers to provide much higher quality, as shown by NVIDIA.
DLSS 4 is also coming to Unreal Engine 5 and Blender. This is actually a very good step forward as it brings forth a much smoother working experience while dealing with highly-detailed environments or objects in the viewport. I can see myself making use of this technology in my work too.
For more information about DLSS 4, check it out here.
AI, of course
All of the DLSS 4 features are already based on AI. Since the new GeForce RTX 50-series is running on the new Blackwell architecture that puts a heavy emphasis on AI, we have a lot of other features too. One of which is the RTX Neural Shaders. Game engines need some sort of texture to overlay to the wireframe to create that object and make it look believable – and high-resolution textures are going to tax on the VRAM and its bandwidth. So, RTX Neural Shaders can allow up to 7x compression of the textures to free up the GPU VRAM, but also have the ability to create cinematic-quality textures and have realistic lighting with ray-tracing.
There is also RTX Neural Faces, and this is a new approach to improve face quality using generative AI. It can take a rasterized face and use generative AI to generate a real-time natural face.
Awaiting for GPUs for review
After CES 2025, we’ll be back at our testbench diligently testing all the new RTX 50-series GPUs. We will then find out the true capabilities of these new GPUs and what DLSS 4 really brings to the table, and how it affects the older generations since NVIDIA is bringing some of those features to the predecessors too.
So please do stay tuned.
