I've tried it with my DAC1 USB, and, based on the first impressions, it certainly seems to sound better to upsample to 96K using SRC in Foobar2000. Upsampling has always had an immediate impact on the sound for me that sounds good at first.
But then, after awhile, I find myself getting fatigued by this sound. I usually end up going back to standard 44K playback after a few days and forget about upsampling for a few months. And then I try it again after a few months, with the same results.
I strongly believe that SRC is manipulating the sound (not just upsampling it). Some people like it (and give rave reviews to SRC compared to other resampling DSP plug-ins, particularly certain versions of it...which I guess is an indication that it is treating the data differently in some versions, which supports the idea that it is doing more than just upsampling, in my opinon).
No, upsampling is not necessary (with the DAc1 USB or any other DAC that supports 44K input). One might argue that since the DAC1 is upsampling itself (to something other than 96K, by the way, so if you upsample on your PC your music data will effectively be upsampled twice) that there is no point in doing it on the PC. Others might argue that an upsampler running on the PC might do a better job than the one in the DAC1, so getting it to 96K for the DAC1 helps.
My ears tell me to leave the data alone (and save my CPU some work, as an added bonus). However, I highly recommend using whatever settings sound best to your ears.