I would highly recommend to invest in a passively cooled system, as to function in as similar a manner to embeded systems as possible. This is quite feasible too, akasa make tiny, passively cooled cases which support CPUs with a TDP (thermal design power) of up to 35W (they *can* handle more, but will be hotter). This will allow you to use up to a T-suffix i7 such as a 4785T which will provide more than ample computational power (though most of its performance is achieved through a variable clockspeed which is concerning for large variations in polyphony, but the base clockspeed is sufficiently high to combat any issues). If anything the T-suffix i7 is too much power if you were merely running it as a module, the i5 variant would be more than sufficient. There also lies the concern of a redundant onboard graphics core (which is pretty powerful) but that will bear utility if you decide to attatch a monitor.
Because pianoteq is so tiny relative to sampled pianos you can use a miniscule msata SSD to run it and little RAM. Its compatibility with linux also makes it useful for these embedded applications given the capacity to utilise a realtime kernel and run headless as well as great control over the buffer with JACK. Having the OS boot into pianoteq standalone in fullscreen without an unnecessary desktop environment running made far easier too.
stuff to look at:
Akasa euler
A soundcard/audio interface which can handle a small buffer and is linux compatible
msata ssd
How to administrate linux and use minimal distros for embedded.