Hi Giulia,
4Gb of RAM does not sound like a lot. That might be the limiting factor,
though the program would crash rather than give non-sense. Have you tried
to combine the results through the network instead of through the disc
(--dont_combine_weights_via_disc)?
Best,
Sjors
> Hi Sjors and everyone.
>
> I am having issues with 3D refinement in relion1.3. It runs fine until
> the last step, producing nice looking 3D half maps.
> After the final iteration I get the two _unfil.mrc maps, and they are
> not right. Sometimes both of them, sometimes only one, they show weird
> streaks, and obviously my final result does not make sense.
>
> This is remarkably similar to what Mike Cianfrocco had reported
> previously in this thread, although he solved the issue by increasing
> available disk space (temporary needs seem quite high for the last
> refinement step). I have >1T of space left (and <2G of data) so that is
> not my problem, also memory seems not to be the issue (I have 4G
> available and the log file tells me >1.05G are needed for expectation
> step).
>
> I started having this problem only when I started using a new cluster
> for testing. The same dataset runs just fine on our local 48 core
> machines. The main difference between the two systems is that our
> machines (where refinement works) are AMD and the cluster (where I have
> the problem), is Intel.
> I should mention the cluster allows 2D and 3D classification to run to
> the end with no problems.
>
> Any help with this would be greatly appreciated!
>
> Thanks!
> Giulia
>
> --
> Giulia Zanetti
> ISMB, Birkbeck College
> Malet St. London
> WC1E 7HX
> 02076316898
>
--
Sjors Scheres
MRC Laboratory of Molecular Biology
Francis Crick Avenue, Cambridge Biomedical Campus
Cambridge CB2 0QH, U.K.
tel: +44 (0)1223 267061
http://www2.mrc-lmb.cam.ac.uk/groups/scheres
|