Most-likely, the CPU is NOT the bottleneck for this particular process on your particular computer. It’s probably disk access/transfer speed.
When converting (or rendering) you can sometimes double the speed (or almost double the speed) by using two hard drives, reading from one hard drive and writing to the other. That way, the read/write head doesn’t have jump back-and-forth between the reading & writing locations and it doesn’t have to switch between reading & writing.
But even with multiple drives, everything has to share the same data bus.
Audacity is not designed as a “format converter”. It is designed as a “multi-track audio editor”. The relevance of this is that for it’s design goal it works internally with uncompressed 32 bit audio data. Audacity does not directly edit or convert files, it works with “projects”.
When an audio file is brought into Audacity (“Imported”), the audio data is copied as 32 bit float PCM data. This allows efficient high quality editing and processing.
To create an audio file (“Export”), the audio data from the project is copied to the new file.
Importing and Exporting may require encoding/decoding of the audio data.
The amount of data in uncompressed form within an Audacity project can be 20 x greater than a compressed audio file format. Importing a 100 MB MP3 file can expand to over 2GB of uncompressed audio data. This is not a fast or efficient way to transcode from one file format to another, but that is not the primary objective for Audacity.
Currently Audacity can only use one core at a time, but as you say, it seems unlikely that that is the bottleneck.
It’s a really fast disk with fast write/read/IOps, so I really don’t believe that temp folder/disk is the problem here. Also, there’s a lot of RAM, especially since my PC doesn’t use pagefile…
Or did you mean something else?
Well, that sheds some light, however I’d expect a bit more hardware usage.
True. Maybe I should turn pagefile back on (that’s the only “odd” thing about my PC), it would sit on SSD anyway, but I really doubt that is the problem, in fact, I presume it pushes programs to use RAM as much as possible.
I’ve seen a LAME/MP3 encoder around here that is written for multicore purpose, but I don’t see the point, since I saw only 25% usage on 1 core…
FYI - The CPU is never truly “idiling” unless it’s in a sleep mode or low-power mode. When there’s nothing to do, or when it’s waiting for data or user input it’s still running full speed executing NOPs (“no operation” commands.)