Hi Ludo,
Small differences in box sizes are not likely to make a big difference
in speed. Fourier transforms are done using FFTW, which may work
somewhat better at specific sizes, but RELION is generally not
speed-limited by the Fourier transforms. This is because it works almost
entirely in Fourier space, so that after doing an initial Fourier
transform for each particle, all subsequent (expensive) calculations are
done without further FFTs.
HTH,
S
On 12/15/2014 11:40 AM, Ludovic Renault wrote:
> Hi Kai,
> Thanks. So 1/3 is a good consensus.
> Any idea about processing with different box size that are close? like 168
> being faster than 160 pixels for example??
> Cheers,
> Ludo
>
> On Mon, Dec 15, 2014 at 11:35 AM, Kai Zhang <[log in to unmask]>
> wrote:
>> Hi Ludo,
>>
>> My suggestion is that in any case your boxsize should be at least 1/3
>> larger, especially if your particle are not CTF corrected. One thing for
>> the benefit of larger box is that there is a lot of information around
>> your particles because of CTF modulation in Fourier space. Even if your
>> particles are already CTF corrected, FFT introduces artifact if the box
>> is too small.
>>
>> Kai
>>
>> On Mon, 2014-12-15 at 11:15 +0000, Ludovic Renault wrote:
>>> Hi all,
>>> I was wondering if like in eman2 there were any better box sizes to
>>> get faster processing? Especially for the 2d classification step which
>>> is the bottleneck really.
>>> Just out curiosity, is adding 1/3 of particle diameter to your box is
>>> a good consensus in relion? 250 ang particle diameter with a box of
>>> approx. 330 ang diameter for example?
>>> Thanks,
>>> Ludo
>>
--
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
|