From: John Stone (
Date: Thu Jan 31 2019 - 09:21:04 CST

  I would also add that you should try running the NAMD nightly build,
because the binaries we produce here should work fine on the new
Turing RTX GPUs. There should be no reason to compile your own binaries
at this time.


On Thu, Jan 31, 2019 at 09:04:40AM -0600, Joo Ribeiro wrote:
> Hi Jrhau,
> Thank you for reporting the error. The error that you are seeing is
> related to the system size and the number of CPU cores+GPU that you
> selected to run your simulation. I would guess that the system you are
> running is not that big to justify many cores plus a GPU. Reduce the
> number of CPU cores and the simulation should run smoothly.
> Now, on the performance of the multicore version of VMD, did you run
> exactly the same configuration file?
> Please allow me to add some notes about the configuration files produced
> by QwikMD. QwikMD has a lot of ludic behavior in the selection of the MD
> parameters in the config files, namely, high frequency of energy outputs
> and trajectory saving frequency (dcd freq), which is useful when you are
> starting running simulations but has high penalties in the NAMD
> performance. Also, running your simulations with the option "Live view"
> mode (Interactive Molecular Dynamics activated) also decreases your
> performance substantially, as NAMD-VMD communication occurs every so
> often.
> In summary, if you are trying to squeeze the most ns/day from your
> machine, please increase periods of saving frames (dcd freq) and output
> events (outputpressure, outputenergies and etc.) and run your
> simulations in the background (Live view off == IMDon off).
> I hope this helps and I am also copying the NAMD developers on this thread
> so they can comment further on the NAMD issues and how to improve NAMD
> performance.
> Best
> Joao
> On Thu, Jan 31, 2019 at 5:44 AM jrhau lung <[1]> wrote:
> Dear VMD friends:
> In order to run simulation on new Geforce RTX20 vedio card, a NAMD
> was compiled from nightly build Git source to generate a
> [2]Linux-x86_64-multicore-CUDA suported NAMD accroding to the
> recommend process in the relase note. The compile was successful without
> any error and should be successful as the [3]Linux-x86_64-multicore
> verion NAMD was also compiled before generating the CUDA supported
> version and multicore version works fine with QwikMD in MD simulation.
> Unfortunately, running MD simulation using the self-built CUDA NAMD, the
> simulation aborted shortly after launch with the follwing messages.
> Any suggestions and hints would be highly appreciated.
> []^
> Info) Using multithreaded IMD implementation.
> ------------- Processor 64 Exiting: Called CmiAbort ------------
> Reason: FATAL ERROR: ComputeBondedCUDA::copyTupleData, invalid number of
> exclusions
> FATAL ERROR: See [4]
> Charm++ fatal error:
> FATAL ERROR: ComputeBondedCUDA::copyTupleData, invalid number of
> exclusions
> FATAL ERROR: See [5]
> Info) IMD connection ended unexpectedly; connection terminated.
> Another issue would like to have your comments is the simulation
> speed uisng self-built [6]Linux-x86_64-multicore NAMD is
> significantly slower than of the 2.13 multicore version. What woulld be
> the potential causes for this? Is this related to the comfiling tools or
> libs? Thanks
> sincerely,
> Jrhau
> --
> ********************...
> João Vieira Ribeiro
> Theoretical and Computational Biophysics Group
> Beckman Institute, University of Illinois
> [7]
> [8]
> +1 (217) 3005851
> References
> Visible links
> 1.
> 2.
> 3.
> 4.
> 5.
> 6.
> 7.
> 8.

NIH Center for Macromolecular Modeling and Bioinformatics
Beckman Institute for Advanced Science and Technology
University of Illinois, 405 N. Mathews Ave, Urbana, IL 61801           Phone: 217-244-3349