I just learned of the Nvidia Tesla GPU just recently. While it's an interesting product, we try to avoid making software for proprietary systems like that and make more generic speed improvements. Someone posted a similar question last year about the Cuda GPU.
[
afni.nimh.nih.gov]
Minimum memory recommendations have changed over the years. In 2003, we used to recommend 512MB, but now 4GB is a good start. What you need to do is always a consideration. We use both Mac and Linux systems. For a Linux system, we have had better luck with Nvidia graphics boards.