Hi Stephan,
I am using CUDA 12.2.2 and the latest driver version 535.154.05. I
also tried the previous driver version but that also did not help.
Regards
Abhilash
On Wed, Feb 14, 2024 at 12:00 PM Stephan Schott <
s.schott-verdugo.fz-juelich.de> wrote:
> Hi,
>
> Which CUDA version are you using? Try compiling with CUDA 12, if you are
> using an older version.
>
> Best regards,
>
> Stephan Schott Verdugo
> Biochemist
>
>
> Computational Biophysical Chemistry
> Institut für Bio- und Geowissenschaften / Bioinformatik (IBG-4)
> Forschungszentrum Jülich GmbH
> Wilhelm-Johnen-Straße, 52425 Jülich
> Germany
>
>
> ---------------------------------------------------------------------------------------------
>
> ---------------------------------------------------------------------------------------------
> Forschungszentrum Jülich GmbH
> 52425 Jülich
> Sitz der Gesellschaft: Jülich
> Eingetragen im Handelsregister des Amtsgerichts Düren Nr. HR B 3498
> Vorsitzender des Aufsichtsrats: MinDir Stefan Müller
> Geschäftsführung: Prof. Dr. Astrid Lambrecht (Vorsitzende),
> Karsten Beneke (stellv. Vorsitzender), Dr. Ir. Pieter Jansens
>
> ---------------------------------------------------------------------------------------------
>
> ---------------------------------------------------------------------------------------------
>
>
> El mié, 14 feb 2024 a la(s) 5:38 p.m., Abhilash J via AMBER (
> amber.ambermd.org) escribió:
>
>> Hi David,
>>
>> I am already using barostat=2, the issue arises in spite of using
>> barostat=2. I tested the system on A5000, A100, 3090, 2080ti, A40, V100,
>> RTX8000, even the GTX 1080ti, M40 and K80. Only 4500 ada gives this
>> behaviour. Keeping barostat=1 (Berendsen) does not solve the issue.
>> My input file is attached below.
>>
>>
>> ==============================================
>> Production at NPT
>> &cntrl
>> imin=0, ! No minimization
>> irest=1, ! This IS a restart of an old MD simulation
>> ntx=5, ! So our inpcrd file has velocities
>>
>> ! Temperature control
>> ntt=3, ! Langevin dynamics
>> ig=-1, ! Random number seed
>> gamma_ln=1.0, ! Friction coefficient (ps^-1)
>> temp0=300.00, ! Target temperature
>>
>> ! Potential energy control
>> cut=10.0, ! nonbonded cutoff, in Angstroms
>>
>> ! MD settings
>> nstlim=5000000, ! 10 ns total
>> dt=0.002, ! time step (ps)
>>
>> ! SHAKE
>> ntc=2, ! Constrain bonds containing hydrogen
>> ntf=2, ! Do not calculate forces of bonds containing hydrogen
>>
>> ! Control how often information is printed
>> ntpr=10000, ! Print energies every 10000 steps
>> ntwx=10000, ! Print coordinates every 10000 steps to the trajectory
>> ntwr=500000, ! Print a restart file every 500K steps (can be less
>> frequent)
>> ! ntwv=-1, ! Uncomment to also print velocities to trajectory
>> ! ntwf=-1, ! Uncomment to also print forces to trajectory
>> ntxo=2, ! Write NetCDF format
>> ioutfm=1, ! Write NetCDF format (always do this!)
>> nscm=50000,
>>
>> ! Wrap coordinates when printing them to the same unit cell
>> ! iwrap=1,
>>
>> ! Constant pressure control.
>> barostat=2, ! MC barostat... change to 1 for Berendsen
>> ntp=1, ! 1=isotropic, 2=anisotropic, 3=semi-isotropic w/
>> surften
>> pres0=1.0, ! Target external pressure, in bar
>>
>> ! Restraint options
>> nmropt=1,
>>
>> /
>> &wt type='REST', istep1=0,istep2=1500,value1=1.0, value2=1.0, /
>> &wt type='REST', istep1=1501,istep2=2500000,value1=1.0, value2=1.0, /
>> &wt type='END' /
>> LISTOUT=POUT
>> DISANG=restraints
>> &end
>> END
>> END
>>
>>
>> ==============================================
>>
>> On Sun, Feb 11, 2024 at 11:16 AM David A Case <dacase1.gmail.com> wrote:
>>
>> > On Tue, Feb 06, 2024, Abhilash J via AMBER wrote:
>> > >
>> > > We recently bought a machine with RTX 4500 ada GPUs. I tried
>> running a
>> > >MD of a system on it. The system was stable and gave expected results
>> for
>> > >minimization, heating and NVT ensemble parts of the simulation. As
>> soon as
>> > >the NPT part of the simulation started, the volume started increasing
>> and
>> > >density decreased. I checked all 4 GPUs, all are giving this weird
>> result.
>> > >I ran the same system (same parameters and input files) on a different
>> > >machine with a different GPU and it was stable throughout the MD
>> process
>> > >including NPT ensemble. We use a centralized NFS mounted AMBER and CUDA
>> > >installation so AMBER 22 and drivers are identical for both
>> workstations
>> > >(software environment is the same). I also tried changing the nVidia
>> > driver
>> > >version (updated it to the latest) but it did not help.
>> >
>> > It's possible that there is some RTX 4500 specific bug. Try some
>> parallel
>> > simulations (RTX 4500 vs A5000) setting barostat=2. This skips some
>> very
>> > complex code involved in computing virials and the internal pressure.
>> >
>> > ....dac
>> >
>> >
>> _______________________________________________
>> AMBER mailing list
>> AMBER.ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber
>>
>>
>>
>> ------------------------------------------------------------------------------------------------
>>
>> ------------------------------------------------------------------------------------------------
>> Forschungszentrum Jülich GmbH
>> 52425 Jülich
>> Sitz der Gesellschaft: Jülich
>> Eingetragen im Handelsregister des Amtsgerichts Düren Nr. HR B 3498
>> Vorsitzender des Aufsichtsrats: MinDir Stefan Müller
>> Geschäftsführung: Prof. Dr. Astrid Lambrecht (Vorsitzende),
>> Karsten Beneke (stellv. Vorsitzender), Dr. Ir. Pieter Jansens
>>
>> ------------------------------------------------------------------------------------------------
>>
>> ------------------------------------------------------------------------------------------------
>>
>
>
>
> ------------------------------------------------------------------------------------------------
>
> ------------------------------------------------------------------------------------------------
> Forschungszentrum Jülich GmbH
> 52425 Jülich
> Sitz der Gesellschaft: Jülich
> Eingetragen im Handelsregister des Amtsgerichts Düren Nr. HR B 3498
> Vorsitzender des Aufsichtsrats: MinDir Stefan Müller
> Geschäftsführung: Prof. Dr. Astrid Lambrecht (Vorsitzende),
> Karsten Beneke (stellv. Vorsitzender), Dr. Ir. Pieter Jansens
>
> ------------------------------------------------------------------------------------------------
>
> ------------------------------------------------------------------------------------------------
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Feb 14 2024 - 11:00:02 PST
