Unless there has been some huge change in Labview 8.6, Labview 8.5 and 8.2.1 seem to use one core per VI. If you have multiple operations in parallel, then Labview will scale across the cores accordingly. I've found that vision code tends to be extremely linear, at least if it's expected to be real time (one VI goes into another, goes into another). In that case, you'd want to have fewer, faster cores rather than several more slower ones (For example, a higher clocked dual core over a slower clocked quad core). Also remember that more cores are better for system stability, you can load 3 out 4 cores on a quad core computer and still be able to navigate and use the computer normally.
Also remember that, depending on your camera configuration, you may be using a lot of CPU load on just bringing in the frames. For an example: My school used a Intel Q6600 (quad core) with a Gig-e camera. Just bringing the frames into Labview (no processing) used an entire core. USB and Gig-e cameras are high, Firewire are generally low, but everything needs processing power unless something else is doing the processing for you.
If you don't need real time, and you're planning on processing multiple frames at a time, you can pipeline your code so that you can use more cores. Pipelining means that you're performing multiple vision operations in parallel (if you had 4 operations to perform, you'd perform operation #1 on image 4, while performing operation #2 on image 3, operation #3 on image 2, and operation #1 on image 1). This consumes more memory and has some loss, but it's one of the only ways to get all the cores working for you.
I'm a big fan of the Q6600, just because it's so cheap and powerful. I don't like the Xeons too much because they always seem overpriced for the power you're getting (I'm a gamer, just my opinion).
