How to prepare equilibratation for REMD?

User discussions
1

I have a problem in understanding the inputs for equilibrate as i read i have to equilibrate the initial structure at every temperature before starting the REMD simulation. so if I have 10 temperatures I should have 10 files with.mdp with ref-temp in the setting .mdp file contains different temperatures each file of them… I made a for loop for this but couldn’t run the second command will it be like that ?

count=1
for temp in 300.00 306.37 312.84 319.42 326.11 332.92 339.85 346.89 354.06
do

echo "; RUN CONTROL PARAMETERS
###the settings parameters ###

gmx grompp -f nvt$count.mdp -c nvt.gro -t nvt.cpt -p topol.top -o nvt$count.tpr
count=$((count+1))
done
mpirun -np 36 gmx_mpi mdrun -v -deffnm nvt$count -multidir nvt1 nvt2 nvt3 nvt4 nvt5 nvt6 nvt7 nvt8 nvt9 nvt10 $

but i got File input/output error:
nvt13.tpr

2

is there any help please ? as the documentation and tutorial is not obvious in using multiply temperatures

3

The problem is coming from the fact that somehow your counter is reaching 13 when it shouldn’t, hence you are telling mdrun to process a file (nvt.tpr) that does not exist. Equilibration is performed like any other simulation and does not necessarily require the use of -multidir.

4

apperciate your reply … Thanks a lot
so the code should be like that

count=1
for temp in 300.00 306.37 312.84 319.42 326.11 …
echo "; RUN CONTROL PARAMETERS
###the settings parameters ###
do
mx grompp -f nvt$count.mdp -c nvt.gro -t nvt.cpt -p topol.top -o nvt$count.tpr;

gmx mdrun -v -deffnm nvt$count -nb gpu

count=$((count+1))

as i got

Changing nstlist from 10 to 80, rlist from 0.9 to 1.041
On host lambda-quad 2 GPUs selected for this run.
Mapping of GPU IDs to the 2 GPU tasks in the 2 ranks on this node:
PP:0,PP:1
PP tasks will do (non-perturbed) short-ranged and most bonded interactions on the GPU
PP task will update and constrain coordinates on the CPU
Using 2 MPI threads
Using 8 OpenMP threads per tMPI thread

WARNING: This run will generate roughly 2022 Mb of data

NOTE: DLB will not turn on during the first phase of PME tuning
starting mdrun ‘UNNAMED in water’
15000000 steps, 30000.0 ps.
step 0:

Is there any way to make it faster as I have 2 gpu and 10 temperatures and that code creates 6 files for the first temperature with extensions edr , xtc , log . mdp, tpr , trr and it will take 3 months to finish the first temperature only !

5

Your system is probably too small to benefit from trying to split the run over two GPUs (only really big systems will parallelize well) so try using only one GPU for the run with an appropriate -gpu_id to mdrun.

1 Like
6

excuse me do you mean like that

gmx mdrun -v -deffnm nvt$count -gpu_id 0

as i got

starting mdrun ‘UNNAMED in water’
15000000 steps, 30000.0 ps.
step 100, will finish Wed Jan 12 07:58:52 2022

i think it still same time :-(

my nvidia-smi like that

NVIDIA-SMI 470.74       Driver Version: 470.74       CUDA Version: 11.4     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  NVIDIA TITAN RTX    On   | 00000000:1A:00.0 Off |                  N/A |
| 41%   40C    P2    65W / 280W |    236MiB / 24220MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
|   1  NVIDIA TITAN RTX    On   | 00000000:68:00.0 Off |                  N/A |
| 41%   43C    P8    19W / 280W |    183MiB / 24217MiB |      0%      Default |
|                               |                      |                  N/A
7

Without any details on what you’re doing (number of atoms, .mdp settings, etc) there’s not much anyone can suggest to speed things up.

1 Like
8 
echo "; RUN CONTROL PARAMETERS
integrator               = md
tinit                    = 0       ; Starting time
dt                       = 0.002   ; 2 femtosecond time step for integration
nsteps                   = 15000000        ; Make it 200 ns
; OUTPUT CONTROL OPTIONS
nstxout                  = 500 ; Writing full precision coordinates every nanosecond
nstvout                  = 500 ; Writing velocities every nanosecond
nstfout                  = 0     ; Not writing forces
nstlog                   = 500  ; Writing to the log file every step
nstenergy                = 500  ; Writing out energy information every step
nstxtcout                = 500  ; Writing coordinates every 5 ps
; Pressure coupling is off
pcoupl                  = no        ; no pressure coupling in NVT
; Periodic boundary conditions
pbc                     = xyz       ; 3-D PBC
; Dispersion correction
DispCorr                = EnerPres  ; account for cut-off vdW scheme
; Velocity generation
gen_vel                 = no        ; Velocity generation is off
; OPTIONS FOR BONDS
continuation             = yes         ; Restarting after Nvt
constraints              = all-bonds
constraint-algorithm     = Lincs
lincs-order              = 4
lincs-iter               = 1

; Temperature coupling is on
tcoupl                  = V-rescale             ; modified Berendsen thermostate
tc-grps                 = Protein Non-Protein   ; two coupling groups - more accurate
echo "ref_t                   = $temp     $temp           ; reference temperature


gmx grompp -f nvt$count.mdp -c nvt.gro -t nvt.cpt -p topol.top -o nvt$count.tpr;

gmx mdrun -v -deffnm nvt$count -gpu_id 0

count=$((count+1))

9 residues of peptide with 143 atom in explicit solvent

10

i could make the gpu work by

gmx mdrun -v -deffnm nvt$count -ntmpi 2 -ntomp 18 -gpu_id 01

but it will finish 20 temperatures after 40 days … is that right ? i need to make it faster … how can i do that