NVIDIA officially announced GeForce RTX 40 series video cards, at least two high-end models – GeForce RTX 4090 and RTX 4080, which work on the new Ada Lovelace architecture. The GeForce RTX 4090 will go on sale in the US on October 12th, starting at $1,599, while the GeForce RTX 4080 will arrive later in November, starting at $899.

The RTX 4090 is the top video card for the Lovelace generation. It will come with 24GB of GDDR6X memory. NVIDIA claims that the new graphics accelerator is 2-4 times faster than the RTX 3090 Ti, but will consume the same amount of power as the previous generation card. NVIDIA recommends a power supply of at least 850 W for a PC with a Ryzen 5900X processor.

GeForce RTX 4090 contains 16,384 CUDA cores, has a base clock frequency of 2.23 GHz, which increases to 2.52 GHz, performance in operations using tensor cores reaches 1,321 teraflops (Tensor-TFLOPs), ray tracing calculation speed 191 teraflops ( RT-TFLOPs) and 83 teraflops “in traditional image construction using shaders” (Shader-TFLOPs).

The GeForce RTX 4080 will be available in two versions, one with 12 GB of GDDR6X memory and the other with 16 GB of GDDR6X memory. NVIDIA claims that this video card is 2-4 times faster than the RTX 3080 Ti. The 12GB model will cost starting at $899, include 7,680 CUDA cores, a base clock frequency of 2.31 GHz, which can be increased to 2.61 GHz, and performance in operations using tensor cores reaches 639 teraflops (Tensor-TFLOPs), trace calculation speed beams of 92 teraflops (RT-TFLOPs) and 40 teraflops “in traditional image construction using shaders” (Shader-TFLOPs).

The RTX 4080 model with 16 GB will differ not only in the increased amount of video memory. With a price starting at $1,199. The manufacturer also offers more power with 9,728 CUDA cores, a base clock frequency of 2.21 GHz that increases to 2.51 GHz, and performance is claimed at 780 Tensor-TFLOPs, 113 RT-TFLOPs, and 49 Shader-TFLOPs. The RTX 4080 with 12GB of memory will require a 700W power supply, while the 16GB version will require at least 750W. Both RTX 4080 models will hit the market in November.

The new graphics cards will also be available with NVIDIA’s own new design, which will receive a Founders Edition console.

“For our new GeForce RTX 40 Series Founders Edition graphics cards, we’ve further optimized the Dual Axial Flow Through system, increasing fan sizes and fin volume by 10%, and upgrading to a 23-phase power supply,” explains Andrew Burns, a technical marketer at NVIDIA. “Memory temperatures are reduced, and the new, substantially more powerful Ada GPUs are kept cool in ventilated cases, giving gamers excellent overclocking headroom.”

NVIDIA supports the 16-pin PCIe Gen-5 connector with its own RTX 40-series cards instead of the custom solution it created for RTX 30-series Founders Edition GPUs. Like the previous connector, this is a single cable that will power the 40-series card, or the user can use the included adapter to connect three 8-pin power connectors. NVIDIA says you can even plug in a fourth for “more headroom for overclocking.”

ATX 3.0 power supplies with support for 16-pin PCIe Gen-5 connectors will appear in October from ASUS, Cooler Master, FSP, Gigabyte, iBuyPower, MSI, and ThermalTake. Models from other manufacturers are expected to appear soon.

The Ada Lovelace architecture, named after the English mathematician and writer, is the third generation of RTX graphics chips that power the GeForce RTX 4080 and RTX 3090. It is designed to significantly improve ray tracing and support Deep Learning Super Sampling (DLSS) 3.0. This new architecture will also underpin NVIDIA’s next-generation consumer GPUs.