Hi Patricio Barletta,
Thanks for your reply.
Sorry that I did not include the mdout. Before the error, AMBER printed early steps of run which look stable - no asterisks or NaN. The values, even if they're very early and limited in numbers, are stable. Density, temp, energy, volume hover around the average of the last equilibration phase.
I then reached out the authors of the paper that published the dummy model as this is the main determinant between my current system and the previous ones. One of their input works! And it aligns with your advise as well: reducing the timestep from 2fs to 1fs. Doing this, I don’t see error using pmemd.cuda - still running after 50ns and I hope it stays that way.
Visual test: the dummy cage still preserved. Density, volume, and temp looks stable and fluctuate around the expected values. So, as of now, I can have a bit more confident to interpret what the simulation shows as it starts to look promising.
While I am only an operator and not the expert of MD, my guess would be that the dummy atom requires lower timestep (?). But I still can't correlate why with 2fs, pmemd.MPI can proceed but not pmemd.cuda. Any explanation to this will be greatly appreciated as I keen to keep this in mind for the future reference.
While here, I want to thank the first and last author of the dummy atom model and this paper: https://doi.org/10.1021/acs.jpclett.7b02358 for their help in troubleshooting my problem. The team of the last author discussed with me a lot of to solve this problem.
Best regards,
Yusran
Yusran A. Muthahari
Postdoc Protein Engineering and Evolution Unit
Okinawa Institute of Science and Technology (OIST)
Japan
________________________________
From: Patricio Barletta via AMBER <amber.ambermd.org>
Sent: Friday, 02 October 2026 22:28:27
To: amber.ambermd.org <amber.ambermd.org>
Subject: Re: [AMBER] kNLSkinTest specific error upon switching from Berendsen to MC using pmemd.cuda but not pmemd.mpi
I can't see your mdouts, but I'm guessing that even if the CPU run does not crash, they should still show unusually high energies.
The kernel that causes the error is the one that determines whether a particle has moved enough to warrant a neighbor list rebuild. When you get a clash, particles bounce very quickly, translate a lot between timesteps and NaNs show up in that kernel.
I’d recommend:
*
Set a fixed, positive ig (ie: 26611)
*
If the error is deterministic, as it seems to be, cut your MD short before it crashes, and once you are close to the crash, set ntpr=1 and ntwx=1. Then look at the energies that led to the crash. Crashes come from VDWAALS energies, but something else forced the atoms to get close enough to clash.
*
Once you have the frames prior to the crash, view the trajectory in VMD, and use cpptraj to find the atoms that are clashing (check the manual, cpptraj has a command called 'check')
*
Try dt=0.001 as a diagnostic: whether this delays or removes the failure helps assess integration instability.
You can also switch to NVT if you think the barostat is somehow related.
________________________________
From: amber-request.ambermd.org <amber-request.ambermd.org>
Sent: Wednesday, September 30, 2026 3:00 PM
To: amber.ambermd.org <amber.ambermd.org>
Subject: AMBER Digest, Vol 5034, Issue 1
Send AMBER mailing list submissions to
amber.ambermd.org
To subscribe or unsubscribe via the World Wide Web, visit
http://lists.ambermd.org/mailman/listinfo/amber<
http://lists.ambermd.org/mailman/listinfo/amber>
or, via email, send a message with subject or body 'help' to
amber-request.ambermd.org
You can reach the person managing the list at
amber-owner.ambermd.org
When replying, please edit your Subject line so it is more specific
than "Re: Contents of AMBER digest..."
AMBER Mailing List Digest
Today's Topics:
1. Re: ACES26 warning during minimization (Baker, Joseph)
(Patricio Barletta)
2. Re: ACES26 warning during minimization (Baker, Joseph)
(Baker, Joseph)
3. kNLSkinTest specific error upon switching from Berendsen to
MC using pmemd.cuda but not pmemd.mpi (Yusran Abdillah Muthahari)
----------------------------------------------------------------------
Message: 1
Date: Tue, 29 Sep 2026 19:09:07 +0000
From: Patricio Barletta <pb777.iqb.rutgers.edu>
To: "amber.ambermd.org" <amber.ambermd.org>
Subject: Re: [AMBER] ACES26 warning during minimization (Baker,
Joseph)
Message-ID:
<SA1PR14MB46895D93E276D00A714545BAA18C2.SA1PR14MB4689.namprd14.prod.outlook.com>
Content-Type: text/plain; charset="iso-8859-1"
Hi Joseph,
I'm responsible for that spurious warning, and have already gotten complaints about it : )
You can safely ignore it and it will go away in the future (unless you are doing AFE simulations)
Best,
________________________________
From: amber-request.ambermd.org <amber-request.ambermd.org>
Sent: Tuesday, September 29, 2026 3:00 PM
To: amber.ambermd.org <amber.ambermd.org>
Subject: AMBER Digest, Vol 5033, Issue 1
Send AMBER mailing list submissions to
amber.ambermd.org
To subscribe or unsubscribe via the World Wide Web, visit
http://lists.ambermd.org/mailman/listinfo/amber<
http://lists.ambermd.org/mailman/listinfo/amber><
http://lists.ambermd.org/mailman/listinfo/amber>
or, via email, send a message with subject or body 'help' to
amber-request.ambermd.org
You can reach the person managing the list at
amber-owner.ambermd.org
When replying, please edit your Subject line so it is more specific
than "Re: Contents of AMBER digest..."
AMBER Mailing List Digest
Today's Topics:
1. ACES26 warning during minimization (Baker, Joseph)
----------------------------------------------------------------------
Message: 1
Date: Tue, 29 Sep 2026 00:08:38 -0400
From: "Baker, Joseph" <bakerj.tcnj.edu>
To: AMBER Mailing List <amber.ambermd.org>
Subject: [AMBER] ACES26 warning during minimization
Message-ID:
<CAA5VDpDKg7zr1KGqDts_eXA=3SkSEgnj2JVeNCr_GeBy4y-=sg.mail.gmail.com>
Content-Type: text/plain; charset="utf-8"
Hello,
I am simulating a system and noticing this message in my minimization
output file:
[image: image.png]
I have never seen the ACE26 message before. We are not doing any TI
calculations, and it seems to be randomly giving this message for this
particular system we are simulating (using a very standard min.in file that
we have used for many other systems). Should I be worried about this? This
is with Amber26.
Also, one of the systems this happens for is quite large (over 700
angstroms in the z direction for a rectangular box), and there we are
getting other error messages about:
nfft1 must be in the range of 6 to 512!
I saw in some previous messages that this might be fixed by changing the
gridhi parameter in one of the source files to a higher value and
recompiling. Has anybody had success doing this (for both all-atom and
SIRAH simulations)? The previous threads about those source code
modifications trailed off!
Thanks for any insight you can provide!
Kind regards,
Joe
------
Joseph Baker, PhD
Associate Provost for Research
Barbara Meyers Pelson Chair in Faculty-Student Engagement
Professor, Department of Chemistry
The College of New Jersey
Ewing, NJ 08628
Chair-elect (2026), Trenton Local Section of the ACS
Chemistry Division Representative (2024-2027), The Council on Undergraduate
Research
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image.png
Type: image/png
Size: 26322 bytes
Desc: not available
URL: <http://lists.ambermd.org/mailman/private/amber/attachments/20260929/e9d2fd3c/attachment-0001.png<http://lists.ambermd.org/mailman/private/amber/attachments/20260929/e9d2fd3c/attachment-0001.png<http://lists.ambermd.org/mailman/private/amber/attachments/20260929/e9d2fd3c/attachment-0001.png>>>
------------------------------
Subject: Digest Footer
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber<
http://lists.ambermd.org/mailman/listinfo/amber><
http://lists.ambermd.org/mailman/listinfo/amber>
------------------------------
End of AMBER Digest, Vol 5033, Issue 1
**************************************
------------------------------
Message: 2
Date: Tue, 29 Sep 2026 17:45:04 -0400
From: "Baker, Joseph" <bakerj.tcnj.edu>
To: Patricio Barletta <pb777.iqb.rutgers.edu>, AMBER Mailing List
<amber.ambermd.org>
Subject: Re: [AMBER] ACES26 warning during minimization (Baker,
Joseph)
Message-ID:
<CAA5VDpBs91xvG8S8a-6phXyOxNe_xSeDEK+Yo=o-FHPnuMoeFg.mail.gmail.com>
Content-Type: text/plain; charset="UTF-8"
Ha! Thanks Patricio! Glad that it wasn't something subtle going on we were
overlooking!
Joe
------
Joseph Baker, PhD
Associate Provost for Research
Barbara Meyers Pelson Chair in Faculty-Student Engagement
Professor, Department of Chemistry
The College of New Jersey
Ewing, NJ 08628
Chair-elect (2026), Trenton Local Section of the ACS
Chemistry Division Representative (2024-2027), The Council on Undergraduate
Research
On Tue, Sep 29, 2026, 3:12?PM Patricio Barletta via AMBER <amber.ambermd.org>
wrote:
> Hi Joseph,
>
> I'm responsible for that spurious warning, and have already gotten
> complaints about it : )
> You can safely ignore it and it will go away in the future (unless you are
> doing AFE simulations)
>
> Best,
>
> ________________________________
> From: amber-request.ambermd.org <amber-request.ambermd.org>
> Sent: Tuesday, September 29, 2026 3:00 PM
> To: amber.ambermd.org <amber.ambermd.org>
> Subject: AMBER Digest, Vol 5033, Issue 1
>
> Send AMBER mailing list submissions to
> amber.ambermd.org
>
> To subscribe or unsubscribe via the World Wide Web, visit
>
> http://lists.ambermd.org/mailman/listinfo/amber<http://lists.ambermd.org/mailman/listinfo/amber>
> <http://lists.ambermd.org/mailman/listinfo/amber<http://lists.ambermd.org/mailman/listinfo/amber>>
> or, via email, send a message with subject or body 'help' to
> amber-request.ambermd.org
>
> You can reach the person managing the list at
> amber-owner.ambermd.org
>
> When replying, please edit your Subject line so it is more specific
> than "Re: Contents of AMBER digest..."
>
>
> AMBER Mailing List Digest
>
> Today's Topics:
>
> 1. ACES26 warning during minimization (Baker, Joseph)
>
>
> ----------------------------------------------------------------------
>
> Message: 1
> Date: Tue, 29 Sep 2026 00:08:38 -0400
> From: "Baker, Joseph" <bakerj.tcnj.edu>
> To: AMBER Mailing List <amber.ambermd.org>
> Subject: [AMBER] ACES26 warning during minimization
> Message-ID:
> <CAA5VDpDKg7zr1KGqDts_eXA=3SkSEgnj2JVeNCr_GeBy4y-=
> sg.mail.gmail.com>
> Content-Type: text/plain; charset="utf-8"
>
> Hello,
>
> I am simulating a system and noticing this message in my minimization
> output file:
>
> [image: image.png]
> I have never seen the ACE26 message before. We are not doing any TI
> calculations, and it seems to be randomly giving this message for this
> particular system we are simulating (using a very standard min.in file
> that
> we have used for many other systems). Should I be worried about this? This
> is with Amber26.
>
> Also, one of the systems this happens for is quite large (over 700
> angstroms in the z direction for a rectangular box), and there we are
> getting other error messages about:
>
> nfft1 must be in the range of 6 to 512!
>
> I saw in some previous messages that this might be fixed by changing the
> gridhi parameter in one of the source files to a higher value and
> recompiling. Has anybody had success doing this (for both all-atom and
> SIRAH simulations)? The previous threads about those source code
> modifications trailed off!
>
> Thanks for any insight you can provide!
>
> Kind regards,
> Joe
>
> ------
> Joseph Baker, PhD
> Associate Provost for Research
> Barbara Meyers Pelson Chair in Faculty-Student Engagement
> Professor, Department of Chemistry
> The College of New Jersey
> Ewing, NJ 08628
> Chair-elect (2026), Trenton Local Section of the ACS
> Chemistry Division Representative (2024-2027), The Council on Undergraduate
> Research
> -------------- next part --------------
> A non-text attachment was scrubbed...
> Name: image.png
> Type: image/png
> Size: 26322 bytes
> Desc: not available
> URL: <
> http://lists.ambermd.org/mailman/private/amber/attachments/20260929/e9d2fd3c/attachment-0001.png<http://lists.ambermd.org/mailman/private/amber/attachments/20260929/e9d2fd3c/attachment-0001.png>
> <
> http://lists.ambermd.org/mailman/private/amber/attachments/20260929/e9d2fd3c/attachment-0001.png<http://lists.ambermd.org/mailman/private/amber/attachments/20260929/e9d2fd3c/attachment-0001.png>
> >>
>
> ------------------------------
>
> Subject: Digest Footer
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
>
> http://lists.ambermd.org/mailman/listinfo/amber<http://lists.ambermd.org/mailman/listinfo/amber>
> <http://lists.ambermd.org/mailman/listinfo/amber<http://lists.ambermd.org/mailman/listinfo/amber>>
>
>
> ------------------------------
>
> End of AMBER Digest, Vol 5033, Issue 1
> **************************************
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber<http://lists.ambermd.org/mailman/listinfo/amber>
>
------------------------------
Message: 3
Date: Wed, 30 Sep 2026 10:28:52 +0000
From: Yusran Abdillah Muthahari <YusranAbdillah.Muthahari.oist.jp>
To: "amber.ambermd.org" <amber.ambermd.org>
Subject: [AMBER] kNLSkinTest specific error upon switching from
Berendsen to MC using pmemd.cuda but not pmemd.mpi
Message-ID:
<TY4P286MB71507D7F0D127105DF51C6BC938B2.TY4P286MB7150.JPNP286.PROD.OUTLOOK.COM>
Content-Type: text/plain; charset="Windows-1252"
Dear Amber MD community members,
I wanted to ask for advice as I face a specific problem upon switching from the last NPT equilibration (no restraint, Berendsen barostat) to the production run (no restraint, MC barostat). Strangely enough, this problem occurs only when using pmemd.cuda but not pmemd.mpi. I tested this for a short 500ps of run. When the problem occurred, no error was shown in the mdout; only the following note was displayed: ?Error: an illegal memory access was encountered launching kernel kNLSkinTest?.
My current system includes protein, tip4pew, ligand (parameterized for use with GAFF), and a dummy Mg2+ atom with LJ1264. Months ago, the same input file (*.in) was used for another system with a different protein and a different ligand parameterized for GAFF and Tip3p; yet it proceeded smoothly without such an error. I am using AMBER24.
Wanting to proceed with a 1-microsecond simulation, I am very eager to push for pmemd.cuda. Can someone give me advice? Any advice will be appreciated.
I am sharing the last equilibration and the production run input here, in case it helps.
Production Run 100ns - without any restraint
&cntrl
imin=0, ntx=5, irest=1, nstlim=50000000, dt=0.002, ntf=2, ntc=2, temp0=298.0,
ntpr=1000, ntwx=1000, ntwr=1000, cut=10.0, ntb=2, barostat=2, pres0=1.0,
taup=2.0, ntp=1, ntt=3, gamma_ln=2.0, ig=-1, ntr=0, iwrap=1, ioutfm=1,
/
Last NPT equilibration - without any restraint
&cntrl
imin=0, ntx=5, irest=1, nstlim=250000, dt=0.002, ntf=2, ntc=2, temp0=298.0,
ntpr=500, ntwx=500, ntwr=500, cut=10.0, ntb=2,
ntp=1, taup=2.0, ntt=3, gamma_ln=2.0, ig=-1, ntr=0, iwrap=1, ioutfm=1,
/
Also, this is how I set up my system:
1.
minimization (high to low restraint; three steps) -> pmemd.mpi
2.
heating (10->100->200->298; NVT restraint) -> pmemd.mpi
3.
equilibration 1st phase (NPT restraint 2kcal/mol, berendsen) -> pmemd.mpi
4.
equilibration 2nd-4th phase (NPT restraint gradually decrease to 0kcal/mol, berendsen) -> pmemd.cuda
5.
production (NPT, no restraint, MC) -> problem with pmemd.cuda but not pmemd.mpi
I also tried to do a bridging step between 4 and 5 by performing a short production run with pmemd.mpi for 100ps, which is successful; before then, continue with pmemd.cuda. But again, the error appears again upon using pmemd.cuda. And the error seems to be consistent.
Thank you very much for your kind help in advance.
Best regards,
Yusran Muthahari
-OIST-
------------------------------
Subject: Digest Footer
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber<
http://lists.ambermd.org/mailman/listinfo/amber>
------------------------------
End of AMBER Digest, Vol 5034, Issue 1
**************************************
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
CAUTION: This email originated from outside the organization. Do not click links or open attachments unless you recognize the sender and know the content is safe.
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Oct 02 2026 - 08:00:02 PDT
