Yes. Conclusion: it is _very_ hard to predict overall computational costs
from short tests. However, for a given project one usually develops
reasonable estimates for one's system from several completed runs.
S
> And I guess another variable is that the time taken varies strongly with
> the iteration number (since the OP asked for 1 iteration). I assume this
> is connected with the angular sampling in auto refinement. Hence, need to
> be wary of extrapolating from short test runs.
>
> m
>
> -----Original Message-----
> From: Sjors Scheres [mailto:[log in to unmask]]
> Sent: 20 November 2015 17:35
> To: Winn, Martyn (STFC,DL,SC); [log in to unmask]
> Subject: Re: [ccpem] RELION 3D classification and 3D autorefine
>
> It will even depend on the signal-to-noise ratios in the images and the
> shape of your protein. So it is VERY hard to predict.
> HTH,
> Sjors
>
> On 11/20/2015 03:15 PM, Martyn Winn wrote:
>> Hi,
>>
>> I started collating some timings at
>> http://www.ccpem.ac.uk/user_help/relion_benchmarks.php for auto-refine.
>>
>> Big caveat: will depend a lot on compute hardware, number of
>> nodes/threads, size of structure, version of Relion, chosen precision,
>> etc etc. So I am a little nervous to advertise this, as it is only one
>> possible answer ... I agree the Elife paper is good.
>>
>> HTH
>> Martyn
>>
>> From: Collaborative Computational Project in Electron cryo-Microscopy
>> [mailto:[log in to unmask]] On Behalf Of Ludovic Renault
>> Sent: 20 November 2015 14:54
>> To: [log in to unmask]
>> Subject: Re: [ccpem] RELION 3D classification and 3D autorefine
>>
>> Hi Joshua,
>> There's a nice Elife paper that looked into that. It's cloud computing
>> but will still give you an idea.
>> http://elifesciences.org/content/4/e06664
>> I would suggest to do a few tests as each cluster is different.
>> Ludo
>>
>> On Fri, Nov 20, 2015 at 2:41 PM, Joshua Lobo
>> <[log in to unmask]<mailto:[log in to unmask]>> wrote:
>> Hi everyone
>>
>>
>> I have a question regarding the average time of a run of 3D
>> classification and auto-refine . Is there any formula or rule of thumb
>> to make a rough estimate of how much time the job will take to say run 1
>> iteration ? Also how many processors and how much RAM is norrmally used
>> for the 3D classification and autorefine . I'm looking to get the most
>> out of my cluster so these parameters will and timing info will really
>> help.
>>
>>
>> Sincerely
>>
>> Joshua Lobo
>>
>
> --
> 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
>
>
--
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
|