I don't have information about specific speech synthesis systems, but contemporary speech synthesis often relies on neural networks for best quality, and those

I don't have information about specific speech synthesis systems, but contemporary speech synthesis often relies on neural networks for best quality, and those can benefit from GPUs. If you truly prefer robotic sounding, that may not be relevant to you. But if it's all about synthesis speed, that's where GPUs may bridge the performance gap.