As others have said, although not to clearly, is Mach in either form is not the limiting factor. The limiting factor is the hardware being used to generate the steps.
A computer printer port is only good to 100kHz, but reliability at anything above the lower settings (33 & 40Khz?) can be an issue.
What you need is some form of external controller, such as a SmoothStepper which is good to 4Mhz. Another possible options that I can think of are a KFlop, or a CS-Lab whatever (CSMIO?), but there are others offering various features.