bloodroses75 :
Workstation cards are definitely slower than their normal counterparts outside the programs they are specifically designed fr (CAD, etc). One of the reasons is that they have much higher precision in their math; I.E. a normal video card may figure the value of PI out to say 20 decimal places whereas the workstation would go out to 200. That in itself is quite a bit of overhead.
You're thinking of double-precision operations, but those are separate from regular operations which games etc. will use, so there's no actual speed difference there since they should be about the same. Like I say though, the workstation chips are picked for a constant workload with zero errors; this is done usually by staying a bit behind to pick only the chips that are most reliable under testing. If you compare a gaming card to a workstation card of the same architecture and clocked at the same speed, then the performance difference is negligible for regular applications, but professional ones will see a boost due to unique features and optimisations, plus a bunch of extra RAM. A consumer GPU may also suffer more minor errors under heavy, prolonged use; things that wouldn't matter in a game, but could affect a computational workload. Remember workstation cards are as much as for heavy OpenCL number-crunching now as they are for general graphics use.