HandBrake · Primary-source explanation; examples are illustrative. Instructions are based on publisher guidance; a UK IPTV account was not tested on this platform.
Inspect the pipeline rather than one label
HandBrake documentation for NVIDIA encoding explicitly lists CPU work which remains when the hardware video encoder is used. These stages include audio encoding, filters, subtitle handling and muxing, with video decoding dependent on the selected path. High CPU use by itself therefore does not establish that the encoder choice was ignored. Identify the encoder actually selected and the other work in the job before drawing that conclusion.
Compare one necessary processing choice
For an original diagnostic plan, use a short permitted sample with the same source and output target. Record the encoder, active filters and relevant log information. Compare only an optional filter whose removal is acceptable for the test, retaining the initial settings. The documented NVIDIA path can fall back to software decoding when filtering requires a CPU round trip, including crop and scale. This is a pipeline explanation, not advice to remove processing the picture genuinely needs.
Keep the observation hardware specific
A faster result after removing a stage suggests a useful next question; it does not benchmark every GPU or prove a particular component was saturated. Hardware families, drivers and source formats differ, so check the current compatibility documentation instead of copying an old support table. Report total elapsed work and the job configuration together. This article gives an illustrative comparison and does not claim a measured speed increase, a tested driver version or universal hardware decoding behaviour across encoders.