Compressor encoding time

I tested a similar video in an M1 Max, and here is what I found:

using "Faster", the CPU use was rather low, some 55% for a "Transcoder Service" plus a few VTDecoderXPCService instances, some 6% each, for a total of less than 1 core.

using "Slower", the CPU use of VTEncoderXPCService was between 300% and 400%.

This pretty much confirms my suspicion that "slower" is software based and "faster" is hardware based.

Luis Sequeira1 wrote:

This pretty much confirms my suspicion that "slower" is software based and "faster" is hardware based.

Have you tested if the "faster" HW (GPU?) based output is the same quality as the "slower" SW based output especially with difficult input (poor lighting, lots of noise and movement etc)?

I tested this ages ago and AFAIR SW output was better quality but I am not anymore sure about that and which macOS and FCP versions I used then.

At least ffmpeg is said to have clearly better quality with SW encoding. I have not yet tried that and there might be a difference between GPU and Intel CPU's VAAPI hardware acceleration.

Matti Haveri wrote:

Luis Sequeira1 wrote:

This pretty much confirms my suspicion that "slower" is software based and "faster" is hardware based.

Have you tested if the "faster" HW (GPU?) based output is the same quality as the "slower" SW based output especially with difficult input (poor lighting, lots of noise and movement etc)?

I tested this ages ago and AFAIR SW output was better quality but I am not anymore sure about that and which macOS and FCP versions I used then.

At least ffmpeg is said to have clearly better quality with SW encoding. I have not yet tried that and there might be a difference between GPU and Intel CPU's VAAPI hardware acceleration.

No, I have not tested with difficult to encode material. I may do so but not today and probably not tomorrow, as I have some things that need to be finished. I will try when I have more time and see how it goes. I am also curious how different, if at all, the encoded results are in an M1 Max compared to an Intel mac.

Better is really quite unnecessary for that type of material. Interesting tests though. Surprising.

Indeed, surprising... I imagine that the faster one may use the dedicated hardware encoder, and the other may bog down in software, but it's just guessing on my part.

Can you notice any difference between this multipass encoded version and Compressor's single pass?

His media may not be representative as a test. Better to try it with something more complex.

Compressor may estimate 2 hours, but I highly doubt, if you actually run it, that it takes that long. Those estimates are not all that accurate.