From: John Stone (
Date: Wed May 28 2014 - 14:33:45 CDT

You'll need to run one of the new pre-release versions of VMD 1.9.2 in order
to use CUDA on the latest GPUs, because the VMD 1.9.1 builds were done
using an older CUDA toolkit and at that time we didn't know how to ensure
forward binary compatibility. I expect to have a new CUDA-enabled Windows
build posted in the next few days, accessible by the steps outlined here:

  John Stone

On Wed, May 28, 2014 at 07:06:23PM +0000, Matthew J. Comstock wrote:
> Hello,
> I am having trouble getting CUDA to work with my new desktop graphics card
> in Linux.
> I just recently assembled a desktop for fairly basic manipulation and
> visualization of protein structures in VMD. I have been doing these tasks
> (fairly well) on a MB Air. I have been making movies from trajectories
> using the quick surf rep and the Tachyon internal ray tracer. The movies
> look awesome (at least I think so) but take a lot of time. My goal with
> the desktop was to speed up the process in part with more main RAM (I have
> 32 GB now) and a discrete graphics card with CUDA processing. I see that
> both quick surf and Tachyon are sped up with CUDA.
> I bought components and assembled the desktop with seemingly no trouble. I
> have Win7 installed on a partition without issues. The latest NVIDIA
> drivers (337.88) are installed along with VMD 1.9.1. The graphics card is
> a Gigabyte Geforce GTX770. In windows, VMD correctly identifies my
> graphics card, including the CUDA processor, and runs without error. I do
> in fact see great speed up playing trajectories with quick surf using the
> gpu (beautifully smooth high speed playback now - thanks!). But of course
> in Windows I am limited to only 4 GB of my RAM (only a 32-bit VMD build
> seems to be available). And who wants to run VMD in Windows anyway?
> I have also installed Fedora (both versions 18 and 20) and Ubuntu. For
> unknown reasons no version of Fedora I install can see my ethernet card. I
> canšt solve that mysterious problem (and not related to VMD) - hence the
> Ubuntu (has no such problems). On Ubuntu I have installed the latest
> (current - 304.117, also a few older versions from packages, and the
> drivers downloaded directly from NVIDIA included in the CUDA 6.0 install)
> NVIDIA drivers, which all appear to run just fine. I installed VMD 1.9.1
> 64-bit, which also runs just fine, recognizing my card and the CUDA
> processor. EXCEPT on startup it gives a CUDA error. Herešs the startup
> output:
> Info) VMD for LINUXAMD64, version 1.9.1 (February 1, 2012)
> Info)
> Info) Email questions and bug reports to
> Info) Please include this reference in published work using VMD:
> Info) Humphrey, W., Dalke, A. and Schulten, K., `VMD - Visual
> Info) Molecular Dynamics', J. Molec. Graphics 1996, 14.1, 33-38.
> Info) -------------------------------------------------------------
> Info) Multithreading available, 8 CPUs detected.
> Info) Free system memory: 31023MB (96%)
> Info) Creating CUDA device pool and initializing hardware...
> CUDA error: invalid device symbol, line 62
> Info) Detected 1 available CUDA accelerator:
> Info) [0] GeForce GTX 770 8 SM_3.0 @ 1.19 GHz, 4.0GB RAM, KTO, OIO, ZCP
> Warning) Detected X11 'Composite' extension: if incorrect display occurs
> Warning) try disabling this optional X server feature.
> Info) OpenGL renderer: GeForce GTX 770/PCIe/SSE2
> Info) Full GLSL rendering mode is available.
> Info) Textures: 2-D (16384x16384), 3-D (2048x2048x2048), Multitexture (4)
> Info) Dynamically loaded 2 plugins in directory:
> Info) /usr/local/lib/vmd/plugins/LINUXAMD64/molfile
> If I try to do a QuickSurf rep I get the following error:
> QuickSurf: R*1.0, I=0.5, H=1.0 Pad: 3.6 minR: 1.5 maxR: 1.8)
> GridSZ: ( 115 155 134) BBox: (-29.7 -10.8 -32.4)->(84.3 143.2 100.6)
> CUDA error: invalid device function , line 783
> Computing density map grid on CPUs .Running marching cubes on CPU...
> Info) QuickSurf: 0.126 [pre:0.001 vol:0.065 gr:0.015 mc:0.03 mcv:0.012]
> On a Tachyon internal render I get no errors, but also no indication it is
> using either the GPU or the CPU.
> Of course Išve looked around the list and elsewhere to try to figure out
> what I might be doing wrong, but Išm at a loss now. Has anyone else
> encountered and solved a similar problem? Did I just pick a bad graphics
> card with a CUDA driver problem? That would be frustrating. I tried to
> find a note somewhere about recently verified builds (i.e., hardware known
> to have no issues) but couldnšt find anything like that.
> Any advice would be greatly appreciated. Thank you.
> Matt Comstock

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