FFXIV Dawntrail performance metrics and visual improvements
FFXIV Dawntrail performance metrics and visual improvements
I know there's more than a few FFXIV players here. It has been announced the Dawntrail stand alone benchmark will drop tomorrow (14 April), 7AM UTC, midnight PDT. I'll jump to the most interesting technical update: support for FSR and DLSS! FSR 1.0 will be the default option. User has a slider for 50% to 100% render scale. There is an option for dynamic resolution. In the above screenshot the dynamic resolution setting, while not selected, shows it could be active when below 60 fps. Be interesting what other options are there as they weren't shown. If DLSS2 is selected, the game takes over resolution scaling and dynamic resolution. I'm not sure about this and static user settings outside of dynamic resolution would have been nice. Dumb upscaling has the potential to look horrible for a game like this, but I expect they do it right. Only the 3D parts of the world should be upscaled (characters, NPCs, monsters, the world itself) and not UI related parts (e.g. name indicators). The benchmark usually exposes the same graphical settings as the game, so it should be possible to test some things out once it is released. I was originally thinking it'll be a simple comparison vs Endwalker benchmark, but this could get a lot more involved as it becomes a bigger quality vs performance tradeoff that wont be static if dynamic resolution is used. A reminder of the updated requirements from the graphical update:
CPU, GPU, and RAM settings have been adjusted for optimal performance. The average minimum specs are around 7980XE 3070 with 3x 16GB RAM, 2R 2133/1440p support, and a max of 77.6 frames per second at 33 FPS. For the 4070 with 2x 8GB RAM, performance peaks at 1080p with 126.4 FPS. The 3070 Laptop reaches 195.7 FPS at 1080p and 81 FPS at 1440p.
Testing on comparable configurations is underway. Benchmark results suggest the Dawntrail "minimum" GPU now aligns with the Dawntrail recommended GPU. CPU demands have increased slightly, though clarity remains uncertain. The A380 model technically meets DT’s minimum requirements, while the 5775C CPU stays below it—likely sufficient.
A potential issue was identified with the 7980XE/3070 setup; running at 1080p produced nearly identical outcomes across resolutions. CPU thread 3 consistently hit 100% utilization, indicating a possible bottleneck or hardware quirk. This observation prompts further investigation into system-specific adjustments.
Dynamic resolution scaling is being explored, particularly for lower-end GPUs, to better gauge graphics performance impacts. A capture device should be prepared for detailed analysis.
EW DT CPU GPU memory settings are set to preserve performance under average conditions. The average minimum ratio stands at 0.71, indicating solid stability. Performance varies across resolutions: 1080p maxes out at 195.7 FPS with a 76% utilization, while 4K reaches 126.4 FPS at 60% load. Ratios suggest efficient usage in typical scenarios. The system handles higher frame rates better, though CPU constraints become noticeable at peak settings. Adjustments like XMP and power plans had minimal effect.