Martin,
> I prefer to use SUN computers (workstations). I would like to know is you
> could recomend me a specific model of a SUN workstation?
This is a tough question because there is no easy answer. It's just
a question of trade-offs.
In our group we've used DEC Alphas (now Compaq's) and found them to
very adequate -- and much faster than Suns. If you prefer Suns for
other reasons, however, you'll probably just want to buy the fastest
machine you can afford (particularly for fMRI analyses).
In our experience it's worthwhile having sufficient RAM such that most
analyses can be done without paging to the disk. This produces a
significant increase in speed because disk access is slow relative to
RAM and the processor speed can be used more fully if the CPU is not
waiting for disk access. 1GB RAM seems to suffice in most cases.
BTW, if you get a big block of RAM, make sure to increase the maxMem
value in spm_spm.m to make use of the extra memory. If you don't want
to go the RAM route then make sure your disk is fast (RAID arrays are
particularly nice).
BTW, multi-processor machines don't necessary help much because Matlab
isn't setup to take advantage of multi-threaded environments yet. So
a single process can only use one processor. Of course, if you have
multiple SPM sessions running then a multi-processor system will help.
If you have the opportunity, I'd recommend a Compaq XP1000 (for a
single user) or an ES40 (for a group of users) with a RAID array of
>100GB disk and 1GB per processor. We just got such a system and it's
increibly fast. An group fmri analysis (14 session) takes about 3
hours from start to finish (slice timing, realignment, normalization,
model estimation).
Best of luck.
- Joe
*************************************************************************
Joseph Devlin, PhD * email: [log in to unmask]
Center for Speech and Language * [log in to unmask]
Experimental Psychology Department *
University of Cambridge * Phone: 01223 766 454
Cambridge, CB2 3EB * Fax: 01223 766 452
England *
*************************************************************************
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|