I'm using Workstation for many years, from 5.5 to 8.0.6, primarily for audio software development. In all versions I used, host audio playback/recording devices cannot be specified independently for a VM. Actually, only host playback audio device can be specified for a VM; Workstation always determines recording device automatically.
It is extremely ugly and primitive solution. Starting from 3.0, Windows supported more than a single recording and playback devices, and even first MME interface allows applications to use any available device for recording and any available device for playback, regardless of their physical relations. All Windows audio interfaces enumerate recording and playback devices independently; to find an "opposite" device that has physical relation to a given device, you must perform some special operations, it is not obvious. Moreover, there are "single-direction" devices, like USB microphones/headphones, that have no physically related "opposite" device at all.
So most audio applications capable to record and play back, allow to specify recording and playback devices independently. It is both more easy for programming and more convenient for a user, allowing to use all available audio devices without limitations. Almost any advanced user that deals with audio has more than a single audio adapter in the computer and frequently uses one adapter for recording and another for playback. Only a small amount of very proprietary (and strange) applications are adapter-oriented, requiring that both devices, recording and playback, must belong to a single physical adapter.
It is very inconvenient that Workstation implements the second way, not allowing VMs to use any available recording device together with any available playback one. Maybe 9.x or 10.x are more friendly in such case?
If not, is such support at least planned? It would be much more easy and reliably to specify both devices independently than specifying only a playback one and finding an appropriate recording one automatically, or assuming that it has the same ordinal number.
Does anybody know why VMware pays no attention to this issue?