Thanks for the suggestions, David.
This is the error that I get when I pmemd crashes:
Error: an illegal memory access was encountered launching kernel kNLSkinTest
I am running pmemd.cuda in amber18. Since my simulations were fine at 2fs
timestep but not 4 fs timestep, the parmtop was the first thing I checked.
I am indeed using the modified parmtop from parmed.
I will poke around some more and try to change to ff14SB/tip4pew along with
the other tweaks you mentioned. I have already tried setting nscm = 500 and
mcbarint=50, and gamma_ln=2. I'll report back if I find anything that
works. Anecdotally, in the past I have seen ff14SB helps when I got
instabilities (kNLSkinTest errors) with ff19SB.
Thanks,
Neville
*From: *David A Case <david.case.rutgers.edu>
*Subject: **Re: [AMBER] knlskintest error when using HMR with 4fs*
*Date: *June 7, 2021 at 14:47:40 EDT
*To: *AMBER Mailing List <amber.ambermd.org>
*Reply-To: *AMBER Mailing List <amber.ambermd.org>
On Mon, Jun 07, 2021, Neville Bethel wrote:
If I use a 4 fs time step for production, my simulations crash within a few
nanoseconds, if I restart these simulations they will run for a few
nanoseconds before crashing again.
Thanks for the report: a few questions:
1. Is there a reason you put "knlskintest error" in your subject line? Did
you get any kind of error message?
2. What version of Amber are you using? Are you running pmemd.cuda? Are
there non-protein components in your simulation?
3. It's worth double-checking that your prmtop file is indeed the one
that has the repartioned masses. (This may sound trivial, but I've made the
mistake myself of using the original rather than the modified prmtop....)
4. A long shot here, and you might want to wait for others to reply:
consider, as a test, rebuilding your system with ff14SB rather tha ff19SB.
You should not need to re-equilibrate: just see if using the ff14SB prmtop
file (with HMassRepartition) gives different behavior. Since it only takes
a few nanoseconds to fail, this should not be all that much work. There is
anecdotal information that something in ff19SB [CMAP ??] might become
unstable in certain circumstances.
5. Again, since you seem to get failures quickly, try some tweaks to the
input file: (a) set iwrap=0; (b) set ntb=1 and ntp=0. These are just
tests to possible help narrow down the range of problems.
...thx...dac
Hi all,
I recently started using HMR with 4fs timestep at the suggestion of the
amber community.
I used parmed's hmassrepartition command to repartition.
My systems are equilibrated with the amberMDprep protocol, which uses a 1
fs time step with SHAKE.
If I use a 4 fs time step for production, my simulations crash within a few
nanoseconds, if I restart these simulations they will run for a few
nanoseconds before crashing again.
If I use a 2 fs time step with HMR I don't have this issue.
I am using ff19SB with OPC on amber 18. I am not repartitioning water. I've
also included an example input for production below. My system is about
600,000 atoms
Is there something else I should be doing to make my HMR simulations more
stable?
-Neville
&cntrl
imin = 0, nstlim = 500000, dt = 0.004,
ntx = 5, irest = 1, ig = -1,
ntwx = 50000, ntwv = -1, ioutfm = 1, ntxo = 2, ntpr = 50000, ntwr =
50000,
iwrap = 1, nscm = 1000, vlimit = 10.0,
ntc = 2, ntf = 2, ntb = 2, cut = 10.0,
ntt = 3, gamma_ln = 5, temp0 = 298.0, tempi = 298.0,
ntp = 1, barostat = 2, pres0 = 1.0, mcbarint = 100,
ntr = 0,
&end
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Jun 10 2021 - 16:00:03 PDT
