Dear Nvidia Jetson Experts,
Following my previous question about the CPU consumption in case of multiple H264/H265 recording and streaming over the network (via v4l2loopback devices), I would like to have another question/confirmation regarding the number of encoders of the Xavier NX with following test-cases:
a. Only recording, no streaming (i.e. 1x 4K encoder + 3x 1080p encoders or 4x 4K encoders (!?!?!?) for recording to MP4 files) → works well (i.e. recorded framerates reach expected value which is 30) for the 4x v4l2loopback cloned devices even there’s a 4K encoding,
b. Recording & streaming with only 1080p resolution (i.e. 4x 1080p encoders for recording to MP4 files and 4x 1080p encoders for streaming over the network) → works well (i.e. recorded framerates reach expected value which is 30),
c. Mix of 4K and 1080p recording while only 1080p streaming (i.e. 1x 4K encoder + 3x1080p encoders for recording to MP4 files and 4x 1080p encoders for streaming over the network) → does NOT work well, especially the frame-rate of the 4K recording is often less than expected (~20fps to 24fps) and sometime it causes the framerate of other recording decreased.
It looks like in the c) case, 4K encoding configuration had to be used (for the recording of 4K) therefore the total number of the encoders (8) exceeded the number of 4K encoders (4x 4K30).
Do you think above observation is coherent with the spec of Jetson Xavier NX in terms of encoding configuration which is : 2x 4K60 (H.265) | 4x 4K30 (H.265) | 10x 1080p60 (H.265) | 22x 1080p30 (H.265) ?
Thanks and best regards,
Khang