These inference benchmarks are not accurate for real-time results for full analysis. They fail to state which method, ect. were used, ect..
I would suggest for the review team to download any Ai inference REAL creative software, many have FREE DEMOS, and compare the encode format, resolution used, video runtime, Ai model used, bitrate setting used, ect. for a FULL TRANSPARENT result for these GPUs and Apple silicon.
The compact size is what caught my attention here. Getting RTX 5070 performance into a 25 cm, 2-slot card is pretty useful for smaller builds, and the smartphone control for tuning is a nice extra. "Etech Devices" is also worth checking for PC hardware, while "KFA2 and NVIDIA" are obvious good names to compare for graphics cards.