There are two types of upsampling in the DS: PCM is upsampled to 352.8k or 384k and 352.8k PCM, 384k PCM and DSD are upsampled to 56.448MHz. No work is saved by pre-upsampling, the upsamplers are always running. FPGA’s just don’t work like the processors people are used to. Everything is running all of the time and you just choose the answers you want… And a general purpose processor isn’t more powerful (at upsampling) than the FPGA.
But ignoring the theoretical, upsampling to 352.8k externally and then using the DS for further upsampling is a fine thing to do if you like the sound you get or upsampling to DSD externally is a fine thing to do if you like the sound. All that really matters is if you like what you are hearing.
Upsampling PCM to any rate lower than 352.8k externally probably isn’t optimal for sound, but there’s nothing wrong with trying it and seeing what you think.