Dear Victor.
First, just because you did not run shake ON, it does not mean that
amber does not have it ON as default ! You should look in the manual,
but more importantly, look at the output file, where the value of ALL
variables used is listed, including those that you did not set yourself.
Second, there is a strong reason why implicit solvent simulations are
sometimes slower than explicit ones. Explicit solvent simulations only
have a short cutoff for the direct non-bonded interactions, while the
rest is done if fourier space. Implicit solvent simulations need to
compute all non-bonded pairs. Hence, for small systems, implicit is
faster than explicit, but for a system like yours, explicit actually wins.
Now, this is in nanoseconds/day. What you need to realize is that in
THAT time scale, the atoms move 'less' in explicit due to the solvent
friction, so you might very well be seeing the same level of movement in
implicit as in explicit, at shorter time scales.
This is of course not very quantitative, and care is required before
comparisons are done.
Cheers
Adrian
On 1/18/12 5:11 PM, Victor Ma wrote:
> Well, I was just following the AMBER Tutorial therefore did not turn on
> SHAKE for implicit calculation.
>
> I wouldn't say that I have "an unreasonable" small solvent box. The cutoff
> for the water box is 10A.
> solvateoct complex TIP3PBOX 10.0
> I have ~8000 water molecules for ~110 residues which is reasonable. Maybe
> it's just that I have a small system.
>
> Victor
>
> On Wed, Jan 18, 2012 at 4:01 PM, Aron Broom<broomsday.gmail.com> wrote:
>
>> I have never done implicit solvent calculations, but I don't understand why
>> you can't still apply the SHAKE algorithm to your protein (or whatever no
>> water thing you are simulating) and thus have a 2fs timestep? Or if the
>> thing you are simulating has no hydrogens, just use a 2fs timestep anyway.
>>
>> Other than that the only thing I can imagine is the size of your explicit
>> solvent box (or whatever shape you had). The implicit solvent calculation
>> time will depend on the solvent model and the size of your system (without
>> water obviously), whereas the explicit solvent calculation will depend
>> largely on how many waters you have. Therefore, if your explicit solvent
>> system didn't actually have that many waters (you had an unreasonably small
>> solvent box) then it would be extremely fast. Anyway, point being that
>> implicit will only be faster than explicit when you are getting to ignore a
>> large number of waters. Since I've never run them I can't tell you where
>> they cross-over.
>>
>> ~Aron
>>
>> On Wed, Jan 18, 2012 at 4:48 PM, Victor Ma<victordsmagift.gmail.com>
>> wrote:
>>
>>> hi all
>>>
>>> I am doing two calculations using the same structure with explicit and
>>> implicit solvent for the same time length (5ns). Surprisingly, the
>> implicit
>>> calculation is actually twice longer than explicit simply because SHAKE
>> is
>>> on in explicit calculation thus time step is 2fs while in implicit
>>> calculation it is 1fs. The simulation itself seems ok. But I'm worried
>> that
>>> I might miss something here. People generally say implicit calculations
>> are
>>> a lot faster.
>>>
>>> Here are my inputs for the production run:
>>> IMPLICIT:
>>> &cntrl
>>> imin = 0, ntb = 0,
>>> igb = 5, ntpr = 5000, ntwx = 2500,
>>> ntt = 3, gamma_ln = 1.0,
>>> tempi = 300.0, temp0 = 300.0,
>>> nstlim = 5000000, dt = 0.001,
>>> cut = 12.0,
>>> ntr=1,
>>> ig=-1,
>>> /
>>> complex (>10)
>>> 8.0
>>> RES 1
>>> RES 15 16
>>> RES 67 68
>>> RES 78 79
>>> RES 98 99
>>> RES 109
>>> END
>>> END
>>>
>>> EXPLICIT:
>>> &cntrl
>>> imin=0,irest=1,ntx=5,
>>> nstlim=2500000,dt=0.002,
>>> ntc=2,ntf=2,
>>> cut=10.0, ntb=2, ntp=1, taup=1.0,
>>> ntpr=5000, ntwx=2500,
>>> ntt=3, gamma_ln=2.0,
>>> ntr=1,
>>> temp0=300.0, tempi=300.0, ig=-1
>>> /
>>> complex (>10)
>>> 8.0
>>> RES 1
>>> RES 15 16
>>> RES 67 68
>>> RES 78 79
>>> RES 98 99
>>> RES 109
>>> END
>>> END
>>>
>>>
>>> Thanks for your input!
>>>
>>> Victor
>>> _______________________________________________
>>> AMBER mailing list
>>> AMBER.ambermd.org
>>> http://lists.ambermd.org/mailman/listinfo/amber
>>>
>>
>>
>>
>> --
>> Aron Broom M.Sc
>> PhD Student
>> Department of Chemistry
>> University of Waterloo
>> _______________________________________________
>> AMBER mailing list
>> AMBER.ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber
>>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
--
Dr. Adrian E. Roitberg
Professor
Quantum Theory Project, Department of Chemistry
University of Florida
roitberg.ufl.edu
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Jan 18 2012 - 14:30:04 PST
