Hi Kristen,
The memory requirement isn't AFNI dependent as much
as it is data dependent. At a minimum, you should have
RAM to hold 3 times the largest dataset that you might
input to 3dDeconvolve, say.
Some people have crossed the 1 GB dataset size, and need
3-4 GB of RAM to process it, depending on the options used.
If you might have multiple users, that would multiply your
needs quickly.
On a modern system, I would not really consider getting
something with less than 4 GB. Memory has gotten pretty
cheap, relative to what it was.
What sorts of computing requirements do you have (number
of users? dataset sizes? volume/surface data? EPI?, etc.)?
--------------------
I have not heard of anyone running AFNI on a Tesla GPU.
The hard part might be the libraries, I don't know. What
OS will be used?
- rick