External Email - Use Caution
Hello Pradyumna,
I have not read thru the mailing list posts so see what other comments there
may be about this, but the posted system
requirements,https://surfer.nmr.mgh.harvard.edu/fswiki/SystemRequirements- do
not mention anything like Xeon class server chips.
Desktop (non-server class) machines with a modern Intel i7 or i9 chip with base
(non-turbo, non-overclocked) CPU speeds of ~3 GHz are common. I’ve run a
complete recon-all in about 7-8 hours on a mid-2016 MacBook Pro with a 2.9Ghz
i7 chip (4 physical, 8 virtual cores) with 16GB of old higher latency
DDR3 memory using all SSD disk storage (with the machine doing nothing else).
I don’t have any data w.r.t the pros/cons of L2 or L3 level cache. I
don’t know the details of how many threads the commands executed by recon-all
can exploit. From my own experience of running various commands, 4 or 8
physical cores (hyperthreaded to yield 8 or 16 cores) should be fine.
I’ve also found that 24G to 32G of RAM is helpful with performance. Since
data/files will be read/written from disk, at certain points performance could
be I/O bound by how fast disk read/writes occur. This can be helped by
having your data stored on a 7200, 10,000 RPM disk drive (or better an SSD
drive) to speed things up that is physically separate from the drive with the
boot partition.
But you could certainly meet or upgrade the specs to outfit a server machine
with Xeon class chips and drives for multiple users.
- R.
On Nov 13, 2019, at 15:00, Pradyumna Bharadwaj p...@email.arizona.edu
wrote:External Email - Use
CautionHello,For running
recon-all or tracula on a local workstation, is it recommended to go for a
higher processor speed/number of cores, or a larger cache size for the
processor (Intel Xeon W series). Or is it likely to be a more complex interplay
between the number of recon-all processes running at any given
time?Thanks,Pradyumna___Freesurfer
mailing
listfreesur...@nmr.mgh.harvard.eduhttps://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
___
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
