Nvidia’s GeForce RTX 40 Series is official after months of leaks and rumours. Sporting Nvidia’s new 4nm Ada Lovelace architecture and Nvidia’s seriously upgraded DLSS 3.0 technology, the new RTX 40 Series offers a solid 2-4x performance boost compared to top-end 30 series models including the RTX 3080 Ti and RTX 4090 Ti.
Here’s everything you need to know about Nvidia’s next-gen graphics card range, from release date and pricing to the latest details on the Ada Lovelace architecture and the new graphics cards themselves.
Nvidia confirmed that the top-end RTX 4090 will be available on 12 October 2022, though there’s no official word on when we can expect either variant of RTX 4080. Rumours suggest a release is scheduled for sometime during the first two weeks of November, but that’s yet to be confirmed.
Nvidia Ada Lovelace architecture
The next-gen Nvidia RTX 40 Series is based on Nvidia’s latest and greatest architecture, codename Ada Lovelace – considered by many to be the world’s first computer programmer.
The company worked with TSMC to implement an impressive 4nm process on the new graphics card collection, down from the rumoured 5nm process rumoured prior to launch, and that delivers both improved performance and better power efficiency.
Headline features include 76 billion transistors, 90 Shader TLOPs via a new streaming multiprocessor, 200 RT TFLOPs with 2x ray-triangle intersection for better ray-tracing performance and 1,400 Tensor TFLOPs for improved deep learning capabilities. There’s also new Shader Execution Reordering, which Nvidia claims is as big a deal as out-of-order execution tech for CPUs.
While that’s a lot of numbers, what it means in real-world terms is a big improvement to ray-traced graphics performance, with a 2-3x improvement in 3D apps like Blender and a 25% boost to RTX gameplay.
That’s coupled with DLSS 3.0, Nvidia’s latest generation of deep learning super sampling tech that utilises the improved deep learning capabilities of Nvidia’s latest graphics card collection. The highlight of the new tech is the ability to predict entire frames, rather than pixels as with DLSS 2.0. The tech processes the current and prior frame to get data including the direction and velocity of pixels and geometry of objects in the scene, feeds that into a neural network and generates intermediary frames.
That’s complex, but what it means is a huge improvement to gameplay performance. Nvidia showcased a new RTX Max mode in Cyberpunk 2077 during the event, with an average of 100fps+ using DLSS 3.0 tech. It’s also a big help to physics-driven games like Flight Simulator that have traditionally been bottlenecked by the CPU.
In fact, Nvidia claims that gamers will see a 2x jump in frame rate in rasterised games and an impressive 4x performance jumps in titles that utilise ray-tracing technology.
Nvidia also revealed three new graphics cards at GTC 2022; the RTX 4080 (12GB and 16GB variants) and RTX 4090.
Nvidia RTX 4080 (12GB and 16GB)
While Nvidia was admittedly light on details, the company revealed that the RTX 4080 offers an impressive 2-4x jump in performance compared to the last-gen RTX 3080 Ti, a graphics card with already-impressive graphical capabilities.
Rumours prior to launch suggested the 16GB model clocks in at 23Gbps, with a boost clock of 2505MHz. It’s said to have a 340W TGP, but that this could be modified up to 516W for added performance.
Rumours also suggest the 12GB model boasts a clock speed of 21Gbps, sporting a more limited 285W TGP with the ability to boost to 366W when necessary.
The top-end graphics card boasts an impressive 24GB of GDDR6X memory, with Nvidia claiming it’s between 2 and 4x faster than the incredible RTX 3090 Ti, with 4x more processing throughput.
As with the RTX 4080, Nvidia was light on details on the RTX 4090 at its announcement, but leakers give us a good idea at what to expect. Current whispers suggest RTX 4090 will feature 126 Streaming Multiprocessors, providing a total of 16,128 CUDA cores, and that its 24GB of GDDR6X memory clocks in at 21 Gbps. That suggests a 384-bit memory bus and upwards of 1TB/s bandwidth.
However, despite the top-end spec, leakers suggest it’ll feature the same 450W TDP as its predecessor, though it can allegedly boost to 660W when necessary. We’ll update this section once we get official numbers from Nvidia.
Nvidia RTX 4090
The RTX 4090 is a serious graphics card for dedicated gamers and 3D artists.
The top-end graphics card boasts an impressive 24GB of GDDR6X memory, with Nvidia claiming it’s between 2 and 4x faster than the incredible RTX 3090 Ti, with 4x more processing throughput.
As with the RTX 4080, Nvidia was light on details on the RTX 4090 at its announcement, but leakers give us a good idea at what to expect. Current whispers suggest RTX 4090 will feature 126 Streaming Multiprocessors, providing a total of 16,128 CUDA cores, and that its 24GB of GDDR6X memory clocks in at 21 Gbps. That suggests a 384-bit memory bus and upwards of 1TB/s bandwidth.
However, despite the top-end spec, leakers suggest it’ll feature the same 450W TDP as its predecessor, though it can allegedly boost to 660W when necessary.