AMBER Archive (2007)

Subject: Re: AMBER: Limitation in number of atoms/residues

From: Sascha Rehm (SaschaRehm_at_gmx.de)
Date: Fri Jun 22 2007 - 08:46:05 CDT


Hi Robert,
Until now i'm using amber 9 and sander. But after your email I think
about using pmemd ;-)
Since your system worked fine I tried some other visualization tools.
It turned out that VMD is able to display it correctly, so the solvation
should also be correct and I can use my solvated system for the md. As
you mentioned I checked the size of my solvate box - which was ok.
Since I can use up to 16 or 32 CPUs the calculations should finish
during my lifetime..
Thanks for your quick and helpfull answer.
Best regards, Sascha

Robert Duke schrieb:
> Hi Sascha -
> Are you using sander or pmemd? Are you using amber 9? For pmemd, the
> atom limit should be 999,999, which is currently a limitation imposed
> by the field length of the atom count in the inpcrd file, if memory
> serves. There should not be a residue limit per se, except at the low
> end (ie., you need a certain number of residues if you do a parallel run
> - you are not going to have problems with this). For pmemd, there is
> also a limit on how big the periodic box can be, based on the maximum
> fft grid size in any dimension. Currently this (grid_hi, an integer
> parameter) is set to 256 in mdin_ewald_dat.fpp. It can be bumped up to
> 512 without problems, especially if you do parallel runs or have lots of
> memory. I do not know of errors in the truncated octahedron code in
> tleap in amber 9 - the developers for that code can address those
> issues. However, I have made a solvated orthogonal box full of
> protein/tip3p water with 900,000+ atoms using amber 9 xleap, visualized
> in xleap (no problems, kind of slow in xleap though; I was using a RHEL
> 4 box, dual intel 3 ghz processors, 2 GB RAM). I know there can be
> problems with visualizing truncated octahedra - another thing for
> someone else to comment on. Bottom line - with pmemd and xleap, amber
> 9, using orthogonal unit cell, the current real limit in system size is
> 999,999 atoms. You will definitely want to run pmemd, as sander will be
> really slow on a system of this size, and you will probably want to use
> at least 8 processors (I did the setup runs for this system on 4
> processors with pmemd). Run times for systems in the range of 20K-1M
> are roughly a linear function of system size (double the atoms, you
> double the run time for a given number of steps).
> Regards - Bob Duke
>
> ----- Original Message ----- From: "Sascha Rehm" <SaschaRehm_at_gmx.de>
> To: <amber_at_scripps.edu>
> Sent: Friday, June 22, 2007 6:47 AM
> Subject: AMBER: Limitation in number of atoms/residues
>
>
>> Hi all,
>>
>> is there still a restriction in the maximum number of atom or residues?
>> I found some answers to the same question in the year 2002 (max
>> atoms:99999, max res:9999), but I didn't find anything newer.
>>
>> If I solvate a big system with 8 proteins in tleap with water (TIP3PBOX
>> and solvateoct), one edge of the octaeder ist cut off and there also is
>> a part where it seems to have twice as much water as it should have.
>>
>> If I reduce my system from 8 proteins (after solvation about 105000
>> residues and 330000 atoms) to 6 proteins (say 96000 residues) everything
>> works fine.
>>
>> thanks for your answers
>> -----------------------------------------------------------------------
>> The AMBER Mail Reflector
>> To post, send mail to amber_at_scripps.edu
>> To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu
>>
>
>
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber_at_scripps.edu
> To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu
>

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu