External DLSS 5 Processing Pushes 007 First Light Down to 18 FPS in 4K Tests

Two independent tests published on September 15 and 16, 2026 found that experimental external DLSS 5-style processing can drive 007 First Light down to 18–25 frames per second in tested RTX 5090 scenes and 20–28 FPS on an RTX 5080 at 4K with path tracing. The game officially supports DLSS 4.5 features, while these DLSS 5 results came from external tools rather than a native game integration.

What the two tests measured

The RTX 5090 test used NeuralScreen 1.4.0 at 3840×2160 with path tracing and built-in DLSS 4.5 Performance. Its reference results were 47 FPS in a hotel lobby, 73–76 FPS in a nightclub and 57–73 FPS in an outdoor forest area. After external processing was enabled, those scenes reached 18, 25 and 18 FPS respectively.

The separate RTX 5080 comparison used a Ryzen 9 7950X3D, 32 GB of DDR5-6000 memory, Windows 10 64-bit and GeForce driver 616.92. Its 4K result was 20–28 FPS with path tracing, DLSS Performance and the loader's default profile. Because the hardware, software, scenes and test procedures differed, the RTX 5090 and RTX 5080 figures are useful as separate observations rather than a direct GPU-to-GPU ranking.

The RTX 5090 experiment also recorded 99% GPU utilization, 475–580 W of board power, 68–78°C and more than 18 GB of allocated video memory. Those measurements belong to that specific configuration.

Why path tracing makes the cost so visible

Path tracing already places a heavy load on the GPU by calculating lighting, reflections and shadows through a more demanding rendering process. The selected RTX 5080 comparison shows a substantial performance difference between path tracing enabled and disabled before any external DLSS 5-style layer is added.

The RTX 5090 results then show what happens when another processing stage is placed on top of that workload. The test's author attributed the penalty to external framebuffer reprocessing and tensor-unit demand, and reported that deeper integration with IO Interactive's Glacier Engine would be needed to avoid that overhead. That explanation remains tied to the test's interpretation; the measured frame rates are the concrete result.

Official DLSS 4.5 support is a separate matter

NVIDIA lists DLSS 4.5 Super Resolution and Dynamic Multi Frame Generation for 007 First Light. The external experiments used NeuralScreen or the DLSS eXtended Loader to apply DLSS 5-style processing after the game's own rendering pipeline. In other words, the measured 18–28 FPS results describe experimental external processing, not an official native DLSS 5 mode in the game.

That distinction matters when reading performance claims. Frame-generation figures can look dramatically higher than the underlying rendered frame rate, so the base FPS and the exact upscaling and frame-generation settings need to be stated together. Community discussion around 007 First Light focused heavily on that difference.

Image quality versus frame rate

The visual trade-off is not as simple as “newer is always better.” In its comparison, the RTX 5080 test judged path tracing combined with external DLSS 5-style processing the best-looking configuration overall. It also preferred path tracing alone to DLSS 5 alone for lighting, shadows, reflections and overall scene coherence.

The practical choice is therefore a familiar PC-gaming compromise: the combined configuration may produce the strongest image in the tested comparison, but the frame-rate cost can make 4K path tracing difficult to play smoothly on the tested hardware.

What settings improved playability

Under the RTX 5080 test's conditions, 1080p with DLSS Quality reached at least 40 FPS with path tracing and the external loader. A 50% render scale at 1440p was another option, but it reduced image quality.

Those settings describe the RTX 5080 experiment rather than a universal prescription for every GPU or scene. For the configurations actually measured, reducing the rendered resolution was the change associated with a more playable result.