“Actual Rate” in the right-hand section of the Audacity Status Bar aims to shows (when playing) the rate being communicated by Audacity to the sound card. “Actual Rate” when recording aims to show the rate communicated by the sound card to Audacity.
But what the device outputs is part of the playback equation too. And it’s more complex than Doug says, because even without ASIO, the host you choose in Audacity affects playback and recording bit depths and sample rates.
If you choose Windows DirectSound or Windows WASAPI host in Audacity and enable both Exclusive Mode boxes in Windows (where you already chose the bit depth and sample rate) then the device “should” output the bit depth and sample rate you choose in Audacity, subject to the depths and rates that device supports. The bit depth and sample rate you specify in Windows is then ignored (and so is the Windows recording bit depth and sample rate ignored).
If you choose MME host in Audacity or uncheck the “Exclusive Mode” boxes then the Windows playback or recording bit depth and sample rate you specify is observed. You will get depth and rate conversions if you set Audacity in contradiction to the Windows settings.
So what this means is that assuming the DAC does not support 32-bit playback you will always get 24-bit playback. If you wanted to minimise theoretical bit depth conversions you would set Audacity to 24-bit and host to Windows WASAPI with Exclusive Mode on (using Windows DirectSound host in Exclusive Mode, Windows Vista and later is believed to do conversions to and from 32-bit float, like it does if you choose MME). You can’t set WASAPI host in 2.0.5 or 2.0.6 release for physical recording inputs, though.
Given Audacity works internally in 32-bit float the general advice is that if you are doing edits that change the sample amplitudes, it’s more important to set Audacity to 32-bit float.
You should always get 96000 Hz playback in a theoretical sense assuming you leave the Audacity project rate at 96000 Hz. However MME host is limited to 44100 Hz/16-bit. So Windows will I believe be downconverting in that case to 44100 Hz/16-bit, then upconverting to 32-bit/96000 Hz then finally down to 24-bit/96000 Hz. You won’t be getting any ultrasonics above the frequency range of human hearing that were there in the first place.
Use Windows DirectSound or WASAPI host with Exclusive Mode on to ensure that the device receives the sample rate set in Audacity project rate, and to WASAPI with Exclusive Mode on to minimise the bit depth/sample rate conversions.
Gale