On Mon, Jun 30, 2025, Delwakkada Liyanage, Senal Dinuka wrote:
>
>I redid the simulations as you suggested and it is clear that pressure
>blows up at around 40 steps when simulated on cluster. But volume does not
>change much.
>
> NSTEP = 41 TIME(PS) = 1875.082 TEMP(K) = 314.96 PRESS = 36.4
> Etot = -267522.6040 EKtot = 191254.7188 EPtot = -458777.3228
> BOND = 32978.1193 ANGLE = 162052.8644 DIHED = 76624.6032
> 1-4 NB = 20096.0869 1-4 EEL = -409707.7522 VDWAALS = -130197.6310
> EELEC = -210623.6135 EHBOND = 0.0000 RESTRAINT = 0.0000
> EKCMT = 9410.3072 VIRIAL = 7628.6755 VOLUME = 2268757.3105
> SURFTEN = 231.1757
> Density = 1.0567
> ------------------------------------------------------------------------------
>
>
> NSTEP = 42 TIME(PS) = 1875.084 TEMP(K) = 318.36 PRESS = 443.1
> Etot = -262435.3975 EKtot = 193321.3750 EPtot = -455756.7725
> BOND = 32994.7570 ANGLE = 161954.2425 DIHED = 76630.8841
> 1-4 NB = 22753.8626 1-4 EEL = -409717.6309 VDWAALS = -129664.2671
> EELEC = -210708.6208 EHBOND = 0.0000 RESTRAINT = 0.0000
> EKCMT = 9421.0953 VIRIAL = -12285.7022 VOLUME = 2268764.4685
> SURFTEN = -393.6055
> Density = 1.0567
> ------------------------------------------------------------------------------
>
>
> NSTEP = 43 TIME(PS) = 1875.086 TEMP(K) = 318.65 PRESS = 3402.8
> Etot = -237725.8300 EKtot = 193499.4844 EPtot = -431225.3143
> BOND = 33023.5934 ANGLE = 162269.0515 DIHED = 76678.2024
> 1-4 NB = 20083.2682 1-4 EEL = -409733.6342 VDWAALS = -102785.4946
> EELEC = -210760.3009 EHBOND = 0.0000 RESTRAINT = 0.0000
> EKCMT = 9453.0660 VIRIAL = -157241.6848 VOLUME = 2268853.9432
>
>Attempting the same simulation on my local machine I do not see the huge pressure jump.
>
> NSTEP = 40 TIME(PS) = 1875.080 TEMP(K) = 308.92 PRESS = 0.2
> Etot = -269452.5440 EKtot = 187589.0156 EPtot = -457041.5597
> BOND = 34720.0372 ANGLE = 161704.6559 DIHED = 76616.6017
> 1-4 NB = 20047.4900 1-4 EEL = -409671.2143 VDWAALS = -129811.7300
> EELEC = -210647.4002 EHBOND = 0.0000 RESTRAINT = 0.0000
> EKCMT = 8586.7883 VIRIAL = 8575.4783 VOLUME = 2268446.9863
> SURFTEN = 114.3548
> Density = 1.0568
> ------------------------------------------------------------------------------
>
>
> NSTEP = 41 TIME(PS) = 1875.082 TEMP(K) = 309.88 PRESS = 13.2
> Etot = -269487.2710 EKtot = 188169.7031 EPtot = -457656.9741
> BOND = 34583.4698 ANGLE = 161337.9824 DIHED = 76590.8285
> 1-4 NB = 20016.7529 1-4 EEL = -409676.9620 VDWAALS = -129779.9642
> EELEC = -210729.0815 EHBOND = 0.0000 RESTRAINT = 0.0000
> EKCMT = 8584.7721 VIRIAL = 7936.5309 VOLUME = 2268446.8307
> SURFTEN = 74.6725
> Density = 1.0568
> ------------------------------------------------------------------------------
>
>
> NSTEP = 42 TIME(PS) = 1875.084 TEMP(K) = 310.51 PRESS = 15.8
> Etot = -269480.6821 EKtot = 188555.9688 EPtot = -458036.6508
> BOND = 34468.7725 ANGLE = 161145.8104 DIHED = 76588.7699
> 1-4 NB = 19996.1131 1-4 EEL = -409696.7789 VDWAALS = -129728.7452
> EELEC = -210810.5927 EHBOND = 0.0000 RESTRAINT = 0.0000
> EKCMT = 8571.2364 VIRIAL = 7796.3824 VOLUME = 2268449.3064
> SURFTEN = 5.0685
> Density = 1.0568
> ------------------------------------------------------------------------------
>
>
> NSTEP = 43 TIME(PS) = 1875.086 TEMP(K) = 310.23 PRESS = 10.4
> Etot = -269477.8496 EKtot = 188384.4062 EPtot = -457862.2558
> BOND = 34492.8095 ANGLE = 161361.7377 DIHED = 76610.8168
> 1-4 NB = 19987.3064 1-4 EEL = -409725.3165 VDWAALS = -129724.4427
> EELEC = -210865.1670 EHBOND = 0.0000 RESTRAINT = 0.0000
> EKCMT = 8566.9542 VIRIAL = 8056.7340 VOLUME = 2268452.3052
> SURFTEN = -71.8809
> Density = 1.0568
> ------------------------------------------------------------------------------
>
Thanks for the additonal information. The clearest evidence of the problem
shows up in the VIRIAL, which then afffects the PRESSURE.
>
>Does this mean A100 gpu are very sensitive to system fluctuations and need
>more cpu equilibration steps?
I don't think more equilibration is likely to help here: it looks like there
is something (bad) going on with A100 cards. We have seen some examples of
this in the past. I'm hoping someone on the list with access to A100 cards
will have time to look at this.
The more immediate possible workaround is to use the Monte Carlo barostat
(barostat=2). Is there some reason that you think this won't work for your
type of system?
...thanks again...sorry to not be more helpful....dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Jul 01 2025 - 21:00:02 PDT
