[gmx-users] multiple GPU usage for simulation
praveen kumar
praveenche at gmail.com
Thu Jan 31 11:43:15 CET 2019
Dear Paul
Many thanks for your help.
As per your suggestion. Now I am able to perform a simulation using two
GPUs.
Earlier there are some unnecessary flags while installing. Now I have
modified installation like this:
cmake .. DGMX_THREAD_MPI=ON -DGMX_GPU=ON -DGMX_X11=ON
For running simulation I have used like this :
gmx mdrun -v -deffnm test -nb gpu -pme gpu -npme 1 -ntomp 5 -ntmpi 4 (for
running on both the GPU's)
gmx mdrun -v -deffnm test -nb gpu -pme gpu -npme 1 -ntomp 5 -ntmpi 2
-gpu_id 0 (or 1) (for running on indicidual the GPU)
However, now simulation running on both the GPUs. The sample performance on
a small benchmark test is given below for (98 k atoms)
2 GPU's together will give ~ 42 ns/day
1 GPU (without using other GPU) ~ 33 ns/day
If two simulations run parallelly on individual GPU's give us the ~ 22
ns/day
from "nvidia-smi" I can see that maximum of ~57% of GPU utilization on both
the GPUs. Is this maximum I can achieve with this?
Thanks
Praveen
On Wed, Jan 30, 2019 at 12:08 PM <
gromacs.org_gmx-users-request at maillist.sys.kth.se> wrote:
> Send gromacs.org_gmx-users mailing list submissions to
> gromacs.org_gmx-users at maillist.sys.kth.se
>
> To subscribe or unsubscribe via the World Wide Web, visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users
> or, via email, send a message with subject or body 'help' to
> gromacs.org_gmx-users-request at maillist.sys.kth.se
>
> You can reach the person managing the list at
> gromacs.org_gmx-users-owner at maillist.sys.kth.se
>
> When replying, please edit your Subject line so it is more specific
> than "Re: Contents of gromacs.org_gmx-users digest..."
>
>
> Today's Topics:
>
> 1. multiple GPU usage for simulation (praveen kumar)
> 2. Re: multiple GPU usage for simulation (pbuscemi at q.com)
> 3. modeling evaporation NVP NVE (pbuscemi at q.com)
> 4. how can I get two parallel lysin in the same box?
> (Giuseppe R Del Sorbo)
> 5. Re: how can I get two parallel lysin in the same box?
> (pbuscemi at q.com)
> 6. High potential energy (Ali Ahmed)
> 7. Gromacs 2018.5 with CUDA (???????? ????????)
>
>
> ----------------------------------------------------------------------
>
> Message: 1
> Date: Tue, 29 Jan 2019 22:29:10 +0530
> From: praveen kumar <praveenche at gmail.com>
> To: gromacs.org_gmx-users at maillist.sys.kth.se
> Subject: [gmx-users] multiple GPU usage for simulation
> Message-ID:
> <
> CALAzg1PRoRjDoumVHfRpWS4-0-BdPwQ_muG6J-wwbKq7Q+suVA at mail.gmail.com>
> Content-Type: text/plain; charset="UTF-8"
>
> Dear gromacs users
>
> I am working on molecular simulation using gromacs 2018.4, we have a new
> gpu machine which has two GPU cards.
>
> I have new workstation with 20 cores (Intel I9 processors) with 3.3 GHz
>
> Nvidia Getforce X1080 Ti cards(2 nos)
>
> I have tried running simulation using the command gmx mdrun -v -deffnm
> test
>
> getting this message
>
> Using 1 MPI process
> Using 10 OpenMP threads
>
> 1 GPU auto-selected for this run.
> Mapping of GPU IDs to the 2 GPU tasks in the 1 rank on this nodes:
>
> It seems simulation runs by make use of one GPU instead of two. I have
> checked it using nvidia-smi.
>
> similar error has got already in the given link
> https://mailman-1.sys.kth.se/pipermail/gromacs.org_gmx-
> users/2018-April/119915.html
>
> I have tried all those options given by Mark and other people, I am
> wondering whether this issue cab be solved ? I.e. Make use of two gpu cards
> for one simulation run
>
> I would be really thankful to if anyone can help me in this regard.
>
> Thanks
> Praveen
>
>
> --
> Thanks & Regards
> Dr. Praveen Kumar Sappidi,
> National Post Doctoral Fellow.
> Computational Nanoscience Laboratory,
> Chemical Engineering Department,
> IIT Kanpur, India-208016
>
>
>
> --
> Thanks & Regards
> Praveen Kumar Sappidi,
> National Post Doctoral Fellow.
> Computational Nanoscience Laboratory,
> Chemical Engineering Department,
> IIT Kanpur, India-208016
>
>
> ------------------------------
>
> Message: 2
> Date: Tue, 29 Jan 2019 11:50:12 -0600
> From: <pbuscemi at q.com>
> To: <gmx-users at gromacs.org>
> Subject: Re: [gmx-users] multiple GPU usage for simulation
> Message-ID: <000901d4b7fb$15edfe60$41c9fb20$@q.com>
> Content-Type: text/plain; charset="us-ascii"
>
> I am not expert on this subject but have recently gone through the
> exercise...
>
> Firstly, does nvidia-smi indicate both cards are active ?
>
> Secondly, for the nvt or npt runs have you tried mdrun commands similar
> to
> :
>
> mdrun -deffnm file -nb gpu -gpu_id 01
> or
> mdrun -deffnm file -nb gpu -pme gpu -ntomp 5 -ntmpi 10 -npme 1 -gputasks
> 111110000
> or
> mdrun -deffnm file -nb gpu -pme gpu -ntomp 5 -ntmpi 10 -npme 1 -gpu_id
> 01
>
> this may help select both
>
> hope it helps
> Paul
>
> -----Original Message-----
> From: gromacs.org_gmx-users-bounces at maillist.sys.kth.se
> <gromacs.org_gmx-users-bounces at maillist.sys.kth.se> On Behalf Of praveen
> kumar
> Sent: Tuesday, January 29, 2019 10:59 AM
> To: gromacs.org_gmx-users at maillist.sys.kth.se
> Subject: [gmx-users] multiple GPU usage for simulation
>
> Dear gromacs users
>
> I am working on molecular simulation using gromacs 2018.4, we have a new
> gpu
> machine which has two GPU cards.
>
> I have new workstation with 20 cores (Intel I9 processors) with 3.3 GHz
>
> Nvidia Getforce X1080 Ti cards(2 nos)
>
> I have tried running simulation using the command gmx mdrun -v -deffnm
> test
>
> getting this message
>
> Using 1 MPI process
> Using 10 OpenMP threads
>
> 1 GPU auto-selected for this run.
> Mapping of GPU IDs to the 2 GPU tasks in the 1 rank on this nodes:
>
> It seems simulation runs by make use of one GPU instead of two. I have
> checked it using nvidia-smi.
>
> similar error has got already in the given link
> https://mailman-1.sys.kth.se/pipermail/gromacs.org_gmx-
> users/2018-April/119915.html
>
> I have tried all those options given by Mark and other people, I am
> wondering whether this issue cab be solved ? I.e. Make use of two gpu cards
> for one simulation run
>
> I would be really thankful to if anyone can help me in this regard.
>
> Thanks
> Praveen
>
>
> --
> Thanks & Regards
> Dr. Praveen Kumar Sappidi,
> National Post Doctoral Fellow.
> Computational Nanoscience Laboratory,
> Chemical Engineering Department,
> IIT Kanpur, India-208016
>
>
>
> --
> Thanks & Regards
> Praveen Kumar Sappidi,
> National Post Doctoral Fellow.
> Computational Nanoscience Laboratory,
> Chemical Engineering Department,
> IIT Kanpur, India-208016
> --
> Gromacs Users mailing list
>
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before
> posting!
>
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
> * For (un)subscribe requests visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
> send a
> mail to gmx-users-request at gromacs.org.
>
>
>
> ------------------------------
>
> Message: 3
> Date: Tue, 29 Jan 2019 12:25:34 -0600
> From: <pbuscemi at q.com>
> To: <gmx-users at gromacs.org>
> Subject: [gmx-users] modeling evaporation NVP NVE
> Message-ID: <001301d4b800$070b0320$15210960$@q.com>
> Content-Type: text/plain; charset="us-ascii"
>
>
>
> Dear Users,
>
>
>
> I am trying to understand a model of water evaporation and am asking for a
> few of suggestions or just point to a link.
>
>
>
> So far I have modeled 10000 molecules of spce water and equilibrated to a
> density near 1 gm/cc. I then double the size of the box and then using
>
> p-couple Perinnelo-Rahmen,
>
> Pcouple type = surface-tension or semiisotropic,
>
> compressibility = 4.5e-5 4.5e-5 ,
>
> ref P =1 0
>
> ref temp 300 or 400K and run for 0.5 ns
>
>
>
> What occurs is a few molecules from the cube form ( apparently ) a gas
> phase. What I expected was the water to expand to fill the enlarged box.
> My
> run time may be too short so I'll run longer times later.
>
> But are my expectations incorrect - should not all the water expand ? I
> used an mpd type NPT - should I use NVE and if so how does that differ
> from NPT other than setting tcoupl = no and pcoupl = no. I can - and will
> - try this but it may be an incorrect approach as well
>
>
>
> Paul
>
>
>
>
>
> ------------------------------
>
> Message: 4
> Date: Tue, 29 Jan 2019 20:14:32 +0100
> From: Giuseppe R Del Sorbo <del-sorbo at ill.fr>
> To: gromacs.org_gmx-users at maillist.sys.kth.se
> Subject: [gmx-users] how can I get two parallel lysin in the same box?
> Message-ID: <35108cbf2c24aabc29cd8e05ae8d913f at ill.fr>
> Content-Type: text/plain; charset=US-ASCII; format=flowed
>
>
> Dear all,
>
> I am using gromacs 5.1.2 and I am doing simulation with lysin and
> surfactant.
>
> Now in the same box I want two lys30 (two identical lysin) oriented once
> in parallel and once in orthogonal each other.
>
> I tried using gmx insert-molecules ... but it's a random orientation.
>
> I also tried with -ip file.dat but still I didn't get what I want.
>
> Do you have any suggestions?
>
> Thanks
>
> Best,
>
> Giuseppe
>
>
> ------------------------------
>
> Message: 5
> Date: Tue, 29 Jan 2019 14:14:07 -0600
> From: <pbuscemi at q.com>
> To: <gmx-users at gromacs.org>
> Subject: Re: [gmx-users] how can I get two parallel lysin in the same
> box?
> Message-ID: <000901d4b80f$30a580c0$91f08240$@q.com>
> Content-Type: text/plain; charset="us-ascii"
>
> Giuseppe,
>
> Use Avogadro, import or construct the four ( if I understand your model )
> molecules in the desired orientations. Save as a pdb. Used editconf to
> construct the box.
>
> Paul
>
> -----Original Message-----
> From: gromacs.org_gmx-users-bounces at maillist.sys.kth.se
> <gromacs.org_gmx-users-bounces at maillist.sys.kth.se> On Behalf Of Giuseppe
> R
> Del Sorbo
> Sent: Tuesday, January 29, 2019 1:15 PM
> To: gromacs.org_gmx-users at maillist.sys.kth.se
> Subject: [gmx-users] how can I get two parallel lysin in the same box?
>
>
> Dear all,
>
> I am using gromacs 5.1.2 and I am doing simulation with lysin and
> surfactant.
>
> Now in the same box I want two lys30 (two identical lysin) oriented once in
> parallel and once in orthogonal each other.
>
> I tried using gmx insert-molecules ... but it's a random orientation.
>
> I also tried with -ip file.dat but still I didn't get what I want.
>
> Do you have any suggestions?
>
> Thanks
>
> Best,
>
> Giuseppe
> --
> Gromacs Users mailing list
>
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before
> posting!
>
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
> * For (un)subscribe requests visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
> send a
> mail to gmx-users-request at gromacs.org.
>
>
>
> ------------------------------
>
> Message: 6
> Date: Tue, 29 Jan 2019 17:55:00 -0600
> From: Ali Ahmed <aa5635737 at gmail.com>
> To: gmx-users at gromacs.org
> Subject: [gmx-users] High potential energy
> Message-ID:
> <
> CAFqrJrqgE3soZbbKG88eH-WosLOfLDE25dnoq3TMd_U4gy9kdw at mail.gmail.com>
> Content-Type: text/plain; charset="UTF-8"
>
> Hello GMX users
> I'm simulating a complex system. I started with energy minimization step
> which went fine then followed by a short NPT for equilibration then
> followed by energy minimization for more minimization and its output as
> below
> Steepest Descents converged to Fmax < 10 in 3783 steps
> Potential Energy = -3.0612996e+04
> Maximum force = 9.8396473e+00 on atom 3209
> Norm of force = 1.0086957e+00
>
> However, when I started NVT production, I see the potential of the system
> goes up to a positive value
> Energies (kJ/mol)
> Angle Proper Dih. Ryckaert-Bell. Improper Dih. LJ-14
> 7.61233e+04 8.43296e+02 2.83000e+04 2.89344e+02 6.94750e+04
> Coulomb-14 LJ (SR) Disper. corr. Coulomb (SR) Coul. recip.
> -1.58903e+04 -8.17942e+04 -1.07861e+04 5.24188e+03 5.14340e+03
> Potential Kinetic En. Total Energy Conserved En. Temperature
> 7.69455e+04 1.16251e+05 1.93197e+05 -7.02956e+04 3.00338e+02
> Pres. DC (bar) Pressure (bar) Constr. rmsd
> -3.58345e+02 2.99966e+02 6.24109e-05
> I don't know what happened and why the potential goes up. Should I worry
> about this?
>
> Thanks
> ?Ali
>
>
> ------------------------------
>
> Message: 7
> Date: Wed, 30 Jan 2019 09:37:10 +0300
> From: ???????? ???????? <bogdanov-vladimir at yandex.ru>
> To: gromacs.org_gmx-users at maillist.sys.kth.se
> Subject: [gmx-users] Gromacs 2018.5 with CUDA
> Message-ID: <2469211548830230 at sas1-46c84f197234.qloud-c.yandex.net>
> Content-Type: text/plain; charset="us-ascii"
>
> An HTML attachment was scrubbed...
> URL: <
> http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20190130/10c10fed/attachment.html
> >
>
> ------------------------------
>
> --
> Gromacs Users mailing list
>
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before
> posting!
>
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
> * For (un)subscribe requests visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
> send a mail to gmx-users-request at gromacs.org.
>
> End of gromacs.org_gmx-users Digest, Vol 177, Issue 93
> ******************************************************
>
--
Thanks & Regards
Praveen Kumar Sappidi,
National Post Doctoral Fellow.
Computational Nanoscience Laboratory,
Chemical Engineering Department,
IIT Kanpur, India-208016
More information about the gromacs.org_gmx-users
mailing list