Where to buy RTX 5090 and RTX 5080 — these are the retailers you should check

Latency should not be a problem

Latency is one of those things that always comes up when talking about Nvidia’s DLSS tech. While the frame generation feature makes games look smoother, it doesn’t actually change latency since it’s more of an AI trick than a brute-force upgrade powered by the GPU.

When we saw DLSS 4 generating two extra frames on top of what DLSS 3 could do, it did raise a few eyebrows at the event. Would cramming in extra frames lead to noticeable latency? Thankfully, we were able to put those concerns to rest pretty quickly. Running around Night City, it was clear that those additional frames from Multi Frame Gen didn’t add any extra latency compared to what we experienced with DLSS 3.

Now, this potential issue depends a bit on your setup. If you’re playing at lower frame rates, latency between frames can be more noticeable. But let’s be real—if you’re targeting 100+ FPS (and with DLSS, that’s pretty much the norm), most people won’t notice a thing. It’s smooth sailing all the way.

AI is the key to everything

Nvidia is introducing a host of new RTX features, many of which prominently feature the term “neural.” These include Neural Rendering for full ray and path tracing, Neural Shaders that use AI to enhance in-game textures, and Neural Faces, which leverage generative AI trained on real human expressions to push past the uncanny valley.

One feature that immediately stood out to me during the demos was RTX Neural Materials. Traditionally, game developers need to manually embed textures and define how objects interact with lighting and ambiance within the game world. RTX Neural Materials streamlines this process by using AI to compress the underlying data, resulting in up to a fivefold boost in processing speed.

The end result is textures that look dramatically realistic — rather than just looking like shiny pieces of foil on the ground. Alongside this, RTX Mega Geometry is able to vastly increase the amount of detail on any character to really give this a shot of shooting past uncanny valley.

RTX 5080 gaming PCs from $2,299

While we’re still waiting for RTX 5090 and RTX 5080 GPUs to become available, we’re seeing RTX 5080-equipped gaming desktops trickle in at Adorama — and prices are starting at $2,299!

Right now, this CyberPowerPC Gamer Supreme desktop PC with an RTX 5080 is available for $2,299, and it also includes an AMD Ryzen 9 9900X CPU, 32GB of RAM and 2TB SSD. This is currently the lowest on offer, but more gaming desktops are available with an RTX 5080, including another CyberPowerPC Gamer Supreme for $2,769 with an AMD Ryzen 9 9950X, 64GB of RAM and a 4TB SSD.

Now that we’re seeing gaming desktops on offer, we’re sure to see RTX 5090 and RTX 5080 GPUs be made available soon enough. Stay tuned as we keep an eye out for more to be made available.

Cyberpunk 2077 at 4K 240 FPS

Nvidia has made it abundantly clear that 4K 240 fps is the new target for its GPUs — something that companies are reacting to with monitors that reach these eyewatering levels of smoothness. I got to try Asus’ impressive option out and I was blown away!

And through some AI trickery, the GPU maker has done it. Multi-frame generation is the big thing coming to RTX 50-series GPUs — rather than just being able to generate one additional frame, it can now create three without any overt impact being made to the latency.

That means rather than seeing frame rates of around 80 fps in Cyberpunk 2077 with everything turned up to 4K Ultra with path tracing, you can see that spike at a bonkers 265 FPS in my own testing.

DLSS 4 explained

Deep Learning Super Sampling (DLSS) 4 is the next generation of Nvidias AI game enhancement tech, and it’s very different from past generations.

The key difference under the hood is moving away from the convolutional neural network (CNN). This older method works to find patterns in any in-game image to sharpen them up and make them smoother by increasing the number of frames. However, it’s a bit of a flawed system that tries to analyze literally every little bit of the screen — leading to ghosting (seeing the outline of a fast-moving item on screen follow behind it).

Instead, Nvidia has moved to a transformer model — the kind of thing you see the likes of ChatGPT running on. This adds the additional intelligence of DLSS being able to evaluate the importance of each pixel across an entire frame. That means if can think a few more steps ahead with greater efficiency. And in practice, it means vastly smoother edges with little to no ghosting on them, and a huge increase in frame rates (more on that later).

RTX 5090 by the numbers

I understand the first thing on everyone’s mind is evaluating the improvements in rendering performance without relying on AI enhancements. Nvidia has heavily emphasized terms like “DLSS” and “neural,” which has led many of our readers to question just how much of the progress comes down to raw processing power alone.

To answer this question, our first benchmarks focused on this, and you’re looking at a roughly 22-23% average increase in game frames per second (FPS) over the 4090.

What’s the current state of play?

Morning everyone, and welcome to the RTX 50-series stock checker live blog! I’ll be taking you through any and all stock we find of the RTX 5090 and RTX 5080 — set to go on sale January 30.

My anticipation is that preorders will be available in the next couple of days! So keep this live blog bookmarked and we’ll let you know once they do. This will give you the best chance to avoid the scalpers.

If you’re on the fence about whether to buy one of these new cards (priced at up to $2,000 for the Founders Edition or even more from other companies), Senior Computing Editor Ales Wawro has broken down the three reasons to buy and two to skip.