DLSS 5 Supercharges Performance with Second GPU, Doubles Frame Rates and Reduces Latency

How the Dual-GPU DLSS 5 Mod Works

According to technical demonstrations shared by developer Marcelo Guibout, the ReShade-based add-on alters the standard frame pipeline. Traditional setups handle neural post-processing on the same GPU rendering the game. With the new method, the primary graphics card creates the frame and transmits it to a second card via the PCIe link. The second GPU applies the neural processing pass and outputs the final image to its own connected display.

Tests conducted by Guibout on a Ryzen 7 7800X3D system with 32GB of DDR5 memory utilized two GeForce RTX 5060 Ti 16GB cards operating through PCIe 5.0 x8 connections. A secondary test on a Ryzen 5 5600 system with DDR4 memory indicated that the technique does not require a high-end platform. However, the configuration requires two displays, with one monitor connected to each individual GPU.

Performance Gains Across DLSS Modes

Data shared from tests in The Blood of Dawnwalker at 1080p highlights significant performance differences depending on the active DLSS mode. When running in Ultra Performance mode, the primary rendering card reached 172 frames per second with neural processing disabled. That figure dropped to between 69 and 71 frames per second when the primary card handled the neural pass locally.

Offloading the neural workload to the second RTX 5060 Ti raised performance in Ultra Performance mode to 157 frames per second, marking a gain of more than 125 percent over the single-card neural configuration. In Quality mode, the primary card reached 54 to 55 frames per second with single-card neural rendering, while the dual-card setup increased output to 91 frames per second. Performance in DLAA showed minimal variation because the heavy render workload left less room for the neural pass to become the primary bottleneck.

Trade-Offs: Heat Reduction Versus Added Latency

While offloading neural rendering lowers thermal stress on the primary hardware, it introduces specific performance trade-offs. According to technical evaluations, display latency doubles because completed frames must route through the second GPU before reaching the monitor. This behavioral penalty mirrors the trade-offs found in standard frame generation technologies.

A second performance mod called Neural Upstream, developed by Matias Lombo, attempts a different optimization by running the neural processing step just before DLSS upscaling occurs rather than at the end of the pipeline. According to coverage from Digital Foundry, this alternative approach sacrifices some visual fidelity to achieve higher frame rates at a lower computational cost.

Frequently Asked Questions

What hardware is required to run the dual-GPU DLSS 5 mod?

The technical demonstration used a system equipped with two GeForce RTX 5060 Ti graphics cards, a Ryzen 7 7800X3D processor, and 32GB of DDR5 memory. The setup also requires two separate displays, with one monitor hooked to each GPU.

Does the mod affect system temperatures?

Yes. According to developer Marcelo Guibout, the primary rendering card ran 21 degrees Celsius cooler because it no longer processed the heavy neural workload.

Why does display latency increase with this method?

Latency doubles because rendered frames are redirected from the primary GPU across the PCIe link to a second graphics card for the final neural processing pass before outputting to the display.