Digital instruments based on large sample libraries should not require vastly more CPU than an ordinary sampler. However, they are extremely heavy on memory and disk use. I would therefore say that large/fast memory and a fast hard-drive (e.g. SSD) is more important.
For example, I don't know about how Vienna works, but the Play sampler (EastWest/QL) keeps the first fraction of a second for every note/articulation/velocity in memory, and when that note is played it then has that fraction of a second to read the rest of the sample from the disk. (SSDs are particularly great because not only are they fast in general, but they don't have a long "seek time" when reading from a new location.)
In contrast, instruments such as the Roland V-Piano don't use samples but use physical simulation to produce the sound - this takes more CPU but they don't need to constantly read from the disk.
Nerd Time: the strongest benefit of hyperthreading for digital audio is when you have lots of multiple completely independent effects that are CPU/memory heavy (i.e. not limited by the disk). Certain calculations use multiple areas of the CPU in sequence, so if your synth uses each calculation's result in the next calculation, it has to wait for each calculation to go through all the stages before starting the next one. Hyperthreading is a way to fill in the gaps, so it will interleave the calculations for multiple synths through a single CPU, a bit like a production line.