DLSS 5 reduced the performance of the RTX 5090 from 114 to 60 FPS in 4K in Forza Horizon 6

GAMES

05 September 2026

Views: 4908

Experts conducted detailed hardware benchmarking of the NVIDIA DLSS 5 neural network rendering technology in a racing simulator. Forza Horizon 6 at Extreme graphics presets in native 4K resolution with ray tracing enabled. The flagship processor-based configuration was used as a testbed. AMD Ryzen 7 9800X3D with a frequency of 5225 MHz, 32 GB of DDR5 RAM with a frequency of 6000 MHz, and a top-of-the-line NVIDIA GeForce RTX 5090 graphics accelerator. The experts focused primarily on a direct comparison of the video card telemetry before and after activating the cinematic neural network profile.

Image e4xwihe4xwihe4xw ggpu

The benchmark results demonstrated a catastrophic drop in performance for the flagship graphics chip when running heavy post-processing algorithms. Without DLSS 5, the graphics card delivered an average rendering speed of 114 FPS with a frame rate of only 8.6 ms, a core clock of 2820 MHz, and a power consumption of 487.1 W. Enabling DLSS 5 Cinematic mode resulted in an immediate drop in frame rate to 60 FPS, representing a net performance loss of 47%. At the same time, the frame generation time more than doubled to 17.4 ms, rare 1% Low events dropped from 97 FPS to 56 FPS, and 0.1% Low events dropped from 72 FPS to 56 FPS.

image 90896y7sebtfwftb8wfrynw97fty89wy ggpu

The colossal computational load from the neural network matrices led to a sharp increase in the operating conditions of the graphics adapter hardware. The graphics card’s power consumption increased by 80.3 W and reached an extreme 567.4 W, which led to a rise in the die temperature from 61°C to 67°C and a forced reset of the processor clock rate from 2820 MHz to 2640 MHz. The video buffer remained loaded at 16 GB with a video memory frequency of 14001 MHz, and the GPU utilization increased to a maximum of 99%. Testing clearly demonstrated that, at this stage, the architecture of DLSS 5 algorithms is capable of neutralizing the potential of even the most powerful graphics accelerators on the market.

Source

DISCUSSION CENTER
@GameGPU_com

Join the discussion, share your benchmark results, and debate hardware performance in X.


Discuss on X