Nvidia has revealed three new RTX 40-series graphics cards, the newest generation of its popular line of GeForce GPUs, during the company’s GTC 22 event on Tuesday. The first cards from the line to be revealed are the RTX 4090, the most powerful card in the line, followed by two versions of the RTX 4080 — a 16 GB version and a cheaper 12 GB version.
The first card that Nvidia revealed was its new RTX 4090, the most powerful consumer card the company has produced, and its standout card from the RTX 40-series line. Nvidia is promising twice the performance of last generation’s best card, the RTX 3090 Ti, in games like Cyberpunk 2077 and Microsoft Flight Simulator. The card features 16,384 CUDA cores and 24 GB of memory. Of course, a card this powerful needs a lot of power in return, which is why the RTX 4090 will require three 8-pin power adapters with an optional fourth for overclocking.
The Nvidia GeForce RTX 4090 is set for release on Oct. 12, with prices starting at $1,599.
Nvidia also announced two versions of the RTX 4080, and while you may think that the only difference between the 16 GB and the 12 GB versions is their memory, you would be mistaken. Instead, the two GPUs actually feature fairly different specs. Most notably, the 16 GB version has 9,728 CUDA cores, while the 12 GB model only features 7,680. The latter model also has slower memory than the 16 GB version. In other words, it might be better to consider this an entry-level RTX 4080 that’s closer to a RTX 4070.5.
The Nvidia GeForce RTX 4080 (16 GB) will start at $1,199 and the Nvidia GeForce RTX 4080 (12 GB) will start at $899. Both of the RTX 4080 cards are scheduled to be released in November.
Finally, Nvidia also introduced a host of new features for its RTX 40-series cards, including DLSS 3. This new generation of Nvidia’s real-time image reconstruction process features great improvements in visuals and frame rate over previous versions, but those frame rate upgrades aren’t exactly solid.
The new process that Nvidia’s using employs an interpolation method, meaning that the card predicts the next frame and renders its own version, rather than the one coming directly from the game. This means that while the frame rates will be higher, the game might not feel as smooth since it won’t be reading inputs for each of the interpolated frames.
Along with all the new cards, Nvidia also announced a few games with RTX-enabled ray tracing upgrades, including Portal.