Nvidia GeForce RTX 5090 GPU Benchmarks Reveal Up to 46 Percent Performance Improvement: Report

The Nvidia GeForce RTX 50 Series Blackwell GPUs were announced at the 2025 Consumer Electronics Show (CES) earlier this month, but are not yet available in the market. The RTX 5090 and RTX 5080 models will be available on January 30. Ahead of their availability, reviewers have reportedly revealed synthetic benchmarks for the top-tier GeForce RTX 5090 GPU, suggesting performance improvements of up to 46%, depending on the benchmark platform used to run the tests.

Nvidia GeForce RTX 5090 GPU Benchmarks

VideoCardz compiled the benchmark results in a report. Citing conversations with several reviewers, the publication claims that gamers can expect about a 20% increase in performance from the new Nvidia GeForce RTX 5090 GPU compared to its predecessor, the RTX 4090. While these numbers are fairly accurate, the report stresses that performance can vary significantly based on game, resolution, and settings.

FireStrike is a DirectX 11 benchmark for gaming PCs that reportedly highlights a 36% performance improvement. It was evaluated in performance, extreme and super scenarios. At the same time, the DirectX 12 TimeSpy benchmark showed a 33% improvement over the RTX 4090 GPU in terms of performance and extreme settings.

Other reviewers also conducted benchmark tests such as SpeedWay, Port Royal, and Steel Nomad, and concluded that the overall performance improvement could be as high as 46%. It should be noted that these are overall graphics scores, not overall performance figures for the chip. There may be differences in the numbers obtained from synthetic benchmarks and real-life gaming scenarios.

Nvidia GeForce RTX 5090 GPU specs

Nvidia’s GeForce RTX 5090 GPU comes with 32GB GDDR7 VRAM (24GB GDDR6 VRAM in the RTX 4090 GPU), has a peak clock speed of 2.4GHz, and has 21,760 CUDA cores. It is built on Nvidia’s Blackwell architecture, with fifth-generation Tensor cores and fourth-generation ray tracing cores.

Overall, the GPU has 92 billion transistors and is capable of performing 3,352 trillion operations per second (TOPS). Nvidia claims this is the company’s most powerful consumer PC GPU yet.