Nvidia GeForce RTX 5090

Nvidia RTX 5090
(Image credit: Nvidia)
Swipe to scroll horizontally
RTX 50-series Blackwell specifications
Header Cell - Column 0 RTX 5090RTX 5080RTX 5070 TiRTX 5070
CUDA cores217601075289606144
Boost clock (GHz)2.412.622.452.51
Base clock (GHz)2.012.32.302.16
Tensor core TOPS335218011406988
Ray tracing core TFLOPS31817113394
Memory32 GB GDDR716 GB GDDR716 GB GDDR712 GB GDDR7
Memory bus width512-bit256-bit256-bit192-bit
Total Graphics Power (watts)575360300250
Required system power (PSU wattage)1000850750650
Power connector1x 600 W PCIe Gen 5 OR 4x PCIe 8-pin adapter1x 450 W PCIe Gen 5 OR 3x PCIe 8-pin adapter1x 300 W PCIe Gen 5 OR 2x PCIe 8-pin adapter1x 300 W PCIe Gen 5 OR 2x PCIe 8-pin adapter
Price$1,999$999$749$549
Release dateJan 30, 2025Jan 30, 2025Feb 2025Feb 2025

The RTX 5090 was announced during Nvidia's CES 2025 keynote presentation in January, along with three other new GPUs to form the vanguard of the new RTX Blackwell series of graphics cards.

As well as significantly more CUDA cores, and a healthy dollop of extra GDDR7 memory, the RTX 5090 is also able to make use of the new DLSS 4 technology, which includes Multi Frame Generation. That's a Blackwell-exclusive feature which allows the card to generate up to three subsequent frames, using the power of AI, for every one frame generated via traditional compute methods.

This is how Nvidia's able to claim the RTX 5090 will perform at twice the level of the RTX 4090. Though, mostly only in the games and apps that support DLSS 4 and Multi Frame Generation.

Latest about Nvidia GeForce RTX 5090