Hi Kirtana,
Sander is more tolerant of bad structures than the GPU code since it uses
a full double precision model. The SPFP precision model we use in the GPU
code overflows if forces are unreasonably high. So my suggestion would be
to do the entire heating step here with sander (or pmemd CPU) and then
after that point once the structure has improved then switch to the GPU
code for the relaxation and production.
All the best
Ross
On 6/23/14, 9:42 PM, "kirtana S" <skirtana4.gmail.com> wrote:
>Thank you all for the reply. Yes I checked my structure there are some
>atoms which are very close to each other and thus gives very high energy
>when I start my heating step. Minimization does not solve my problem. As
>said by Dr. Ross I also verified the initial steps and in the third step
>itself the structure is distorted.
>When I was using sander earlier, in the first 500 steps of heating the
>energy adjusts itself and the structure is relaxed with a sudden drop in
>energy . Is it correct. When I use this structure i.e the restrt file in
>pmemd.cuda this works fine.
>
>NSTEP = 0 TIME(PS) = 0.000 TEMP(K) = 0.00 PRESS =
>0.0
> Etot = -6521385.5703 EKtot = 0.0000 EPtot =
> -6521385.5703
> BOND = 46854.0515 ANGLE = 11499.8085 DIHED =
> 5254.3722
> 1-4 NB = -2098.3437 1-4 EEL = 29129.7754 VDWAALS =
> -20512.1273
> EELEC = -6591513.1069 EHBOND = 0.0000 RESTRAINT =
>0.0000
> Ewald error estimate: 0.2201E-06
>
>--------------------------------------------------------------------------
>----
>
>
> NSTEP = 500 TIME(PS) = 0.500 TEMP(K) = 84.97 PRESS =
>0.0
> Etot = 58286.0541 EKtot = 5470.0580 EPtot =
>52815.9961
> BOND = 37269.2949 ANGLE = 11632.2199 DIHED =
> 4587.7362
> 1-4 NB = -2070.4722 1-4 EEL = 28519.3325 VDWAALS =
> -22369.2241
> EELEC = -12345.0692 EHBOND = 0.0000 RESTRAINT =
> 7592.1781
> EAMBER (non-restraint) = 45223.8180
> Ewald error estimate: 0.1185E-03
>
>--------------------------------------------------------------------------
>----
>Thanks
>Kirtana
>
>
>
>On Mon, Jun 23, 2014 at 4:11 PM, Ross Walker <ross.rosswalker.co.uk>
>wrote:
>
>> Hi Kirtana,
>>
>> The *'s mean the value is too big to print in the space provided. In
>>this
>> case your VDW energy is huge suggesting one of three things:
>>
>> 1) Your have a bad structure, atoms too close for example - what was the
>> final energy from the sander run?
>>
>> 2) Your restraint energy is also huge - perhaps you are restraining to
>>an
>> innapropriate reference structure?
>>
>> 3) There is a bug in the code.
>>
>> Firstly I am assuming you are running the minimization with the CPU
>> version of the code and then switching to the GPU version for the
>> dynamics? - If not then start by doing this. The SPFP precision model is
>> very intolerant of strained structures and thus I recommend that one
>> always minimizes with the CPU code.
>>
>> For the dynamics set ntpr=1, nstlim=100, ntwx=1 - then run it and you
>>will
>> be able to see what goes wrong at the beginning and visualize it. Note
>> tautp in your mdin file has no effect with ntt=3.
>>
>> Finally check what version of the GPU code you are running (printed
>>early
>> on in mdout) - make sure it is fully up to date. It should be 14.0 for
>> AMBER 14 and 12.3.1 for AMBER 12.
>>
>> All the best
>> Ross
>>
>>
>> On 6/23/14, 8:40 AM, "kirtana S" <skirtana4.gmail.com> wrote:
>>
>> >Dear Amber users,
>> >
>> >If I notice asteriosk or stars marks in my output file and restart
>>file
>> > what does this imply. I had this job run in sander without any issues
>>and
>> >noticed some error in pmemd cuda. I tried to rectify this by using
>>iwrap 1
>> >and still notice similar result.
>> >I used mpi run and also single processor run.
>> >
>> >Can you tell me if the issue is with my input files or my starting
>> >structure. Minimization run was preformed on the solvent with
>>restraint on
>> >solute before this step.
>> >My input file
>> >Minimization:
>> >&cntrl
>> > imin=1, ntx=1,
>> > dielc=1.0,ntmin=2,
>> > ntrx=1, ntxo=1,
>> > ntb=1,ioutfm=1,
>> > irest=0, cut=10.0,
>> > ntpr=1,
>> > maxcyc=10000,
>> > ntt=0, ntr=1,
>> >&end
>> >Hold the PRO fixed
>> >40.0
>> >RES 1 300
>> >END
>> >END
>> >
>> >Second step which give me the error
>> >
>> >&cntrl
>> > imin=0,ntx=1,
>> > cut=10.0,tempi=0.0,
>> > ntpr=100,ntwx=100,ntwe=100,
>> > nstlim=1000000,temp0=100.0,
>> > dt=0.001,
>> > ntc=2, ntf=2,
>> > irest=0,tautp=4.0,
>> > ntb=1,iwrap=1,
>> > ntt=3,gamma_ln=5.0,ntr=1,
>> > &end
>> >Hold surface fixed
>> >40.0
>> >RES 1 300
>> >END
>> >END
>> >
>> > NSTEP = 100 TIME(PS) = 0.100 TEMP(K) =********* PRESS =
>> >0.0
>> > Etot = ************** EKtot = ************** EPtot =
>> >**************
>> > BOND = -0.0000 ANGLE = 1580104.7120 DIHED =
>> >12738.0737
>> > 1-4 NB = 0.0000 1-4 EEL = -19.0698 VDWAALS =
>> >**************
>> > EELEC = -308.1300 EHBOND = 0.0000 RESTRAINT =
>> >**************
>> > EAMBER (non-restraint) = **************
>> >
>> >mden file
>> >
>> >L0 Nsteps time(ps) Etot EKinetic
>> >L1 Temp T_solute T_solv Pres_scal_solu
>> >L2 Pres_scal_solv BoxX BoxY BoxZ
>> >L3 volume pres_X pres_Y pres_Z
>> >L4 Pressure EKCoM_x EKCoM_y EKCoM_z
>> >L5 EKComTot VIRIAL_x VIRIAL_y VIRIAL_z
>> >L6 VIRIAL_tot E_pot E_vdw E_el
>> >L7 E_hbon E_bon E_angle E_dih
>> >L8 E_14vdw E_14el E_const E_pol
>> >L9 AV_permMoment AV_indMoment AV_totMoment Density
>> > dV/dlambda
>> >L0 100 0.1000000000E+00 0.2542518727E+16 0.2542508165E+16
>> >L1 0.3949644593E+14 0.3949644593E+14 0.2525519783E+22
>>0.1000000000E+01
>> >L2 0.1000000000E+01 0.1015340000E+03 0.1656810000E+03
>>0.3714600000E+02
>> >L3 0.0000000000E+00 0.0000000000E+00 0.0000000000E+00
>>0.0000000000E+00
>> >L4 0.0000000000E+00 0.0000000000E+00 0.0000000000E+00
>>0.0000000000E+00
>> >L5 0.0000000000E+00 0.0000000000E+00 0.0000000000E+00
>>0.0000000000E+00
>> >L6 0.0000000000E+00 0.1056199933E+11 0.1970472220E+10
>>-.3081299803E+03
>> >L7 0.0000000000E+00 -.1024454832E-07 0.1580104712E+07
>>0.1273807373E+05
>> >L8 0.0000000000E+00 -.1906978498E+02 0.8589934592E+10
>>0.0000000000E+00
>> >L9 0.0000000000E+00 0.0000000000E+00 0.0000000000E+00
>>0.0000000000E+00
>> > 0.0000000000E+00
>> >
>> >Can you help me with this.
>> >
>> >Thanks
>> >Kirtana
>> >_______________________________________________
>> >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
>>
>_______________________________________________
>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
Received on Tue Jun 24 2014 - 10:30:02 PDT
