How long would a 5 microsecond run last? In general, I would make runs
no longer than the amount of time you are willing to lose if the output
is lost, in case it applies here. I.e. my preference would be runs of <=
2 hours, if I was able to learn of crashes in that sort of time frame.
Bill
On 9/21/18 2:34 AM, Nikolay N. Kuzmich wrote:
> Dear Matias and Amber users,
>
> 2) several days ago I tried to start a 5 microseconds production run
> simulation of a small protein
> in implicit solvent (igb=8) but a runtime error occured:
>
> At line 717 of file mdin_ctrl_dat.F90 (unit = 5, file =
> 'abk_gb_ProdRun5us.in')
> Fortran runtime error: Integer overflow while reading item 4
>
> Error termination. Backtrace:
> nick.user039l:~/amber18/ABK_protein_MD> #0 0x2b67f0db05cd in ???
> #1 0x2b67f0db1115 in ???
> #2 0x2b67f0db187a in ???
> #3 0x2b67f0f14d79 in ???
> #4 0x2b67f0f1523b in ???
> #5 0x2b67f0f18be3 in ???
> #6 0x2b67f0f1a65f in ???
> #7 0x2b67f0f2254f in ???
> #8 0x2b67f0f22808 in ???
> #9 0x4282cf in ???
> #10 0x4cad6d in ???
> #11 0x4b1a59 in ???
> #12 0x40612c in ???
> #13 0x2b67f1cb1f49 in ???
> #14 0x419219 in ???
> at ../sysdeps/x86_64/start.S:120
> #15 0xffffffffffffffff in ???
>
> So the item 4 was actually ntslim; neither 2500000000 nor 3000000000 worked,
> only run with ntslim=2000000000 could be started.
> I have found reports of similar issues in the Amber mail reflector
> but just wanted to make it clear.
>
> P.S. Matias, thank you for the answer on the 1st question
>
> Kind regards,
> Nick
>
> ______________________________________________________________________
>
>
>
> Dear Nick,
>
> 1. A similar issue was already addressed in [
> https://www.researchgate.net/post/How_can_I_reduce_temperature_oscillation_in_NVT_ensemble_simulation_of_FPMD
> ]
>
> According to thermodynamics, the temperature fluctuation in a closed system
> containing N atoms is proportional to 1/sqrt(N), that is essentially what
> you expect from a normal distribution as is the case in a proper canonical
> ensemble (I mean "proper" if you are not using the Berendsen thermostat).
>
> Hence, the larger the system the smaller the temperature fluctuations...
> That is the reason why in implicit simulations you expect larger
> fluctuations comparing to explicit ones...
>
> Just to give you an example: for a system of 1300 particles you would
> expect a fluctuation of 1/sqrt(1300) = 0.02773501 (or 2.7%), so running at
> 300?K the RMS(T) is ~ 300 * 0.02773501 ~ 8?K
>
> 2. I don't understand this question... are you not able to set more nstlim
> steps? Then you can just continue the simulation up to the desire time...
>
> Best,
>
> Matias
>
> ------------------------------------
> PhD.
> Researcher at Biomolecular Simulations Lab.
> Institut Pasteur de Montevideo | Uruguay
> [http://pasteur.uy/en/laboratorios-eng/lsbm]
> [http://www.sirahff.com]
>
> ----- Mensaje original -----
> De: "Nikolay N. Kuzmich" <nnkuzmich.gmail.com>
> Para: amber.ambermd.org
> Enviados: Jueves, 20 de Septiembre 2018 9:29:24
> Asunto: [AMBER] Temperature RMS, ntslim limit
>
> Dear Amber users and developers,
>
> 1. is the temperature RMS=10.0K within the limits for a good quality
> simulation conducted in implicit solvent? This is the mdin file:
>
> Stage 2 equlibration 0-5ns
> &cntrl
> imin=0, irest=1, ntx=5,
> ntc=2, ntf=2, ig=-1, ntt=3,
> tautp=0.5, tempi=300.0, temp0=300.0,
> ntpr=500, ntwx=500, saltcon=0.15, ntb=0,
> igb=8, gamma_ln=1.0, cut=999., rgbmax=999.
> /
>
> I just noticed that RMS(T) of explicit solvent simulation was remarkably
> lower...
>
> 2. Is ntslim=2 000 000 000 really a limit for the input?
>
> Kind regards,
> Nick
> _______________________________________________
> 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 Fri Sep 21 2018 - 03:00:02 PDT
