VMD-L Mailing List

From: John Stone (johns_at_ks.uiuc.edu)
Date: Mon Mar 08 2010 - 11:19:00 CST

Filippo,
  Although the environment variable is working for you, I'm still
curious why setting the device to compute mode prohibited did not
work, since it works fine in my testing here, even with the same
model of GPU. Can you send me the output of these commands:
  nvidia-smi -a -q
  nvidia-smi -a -q -s

Cheers,
  John stone
  vmd_at_ks.uiuc.edu

On Fri, Mar 05, 2010 at 09:38:40AM +0200, Filippo Federici wrote:
> Hi,
>
> I set the VMDCUDANODISPLAYGPUS environment variable and
> now the Quadro does not show up in the GPU list.
> No more "line 62" error at startup either.
> Thanks to everybody for helping.
>
> Cheers
>
> Filippo
>
> On Thu, 2010-03-04 at 13:54 -0600, John Stone wrote:
> > Hi,
> > The existing VMD code checks both for computeModeProhibited and
> > for the kernel execution timeout flag (used for display), so unless
> > the nvidia-smi tool and driver 'forgot' the setting by the time you
> > ran VMD (there was a bug of this sort some time back), this should not
> > affect the behavior at all. If setting the VMDCUDANODISPLAYGPUS
> > environment variable makes a difference, please let me know, as that
> > would be an unexpected behavior given how I wrote the code.
> >
> > Cheers,
> > John Stone
> > vmd_at_ks.uiuc.edu
> >
> > On Thu, Mar 04, 2010 at 02:16:58PM -0500, Axel Kohlmeyer wrote:
> > > On Thu, 2010-03-04 at 19:54 +0200, Filippo Federici Canova wrote:
> > > > Hi,
> > > >
> > > > I actually tried both with the Quadro on and off,
> > > > and by off i mean i disabled it with the
> > > > nvidia-smi command line program to set it on
> > > > 'prohibited'. the driver is 190.53 and the
> > > > latest cuda sdk was installed.
> > >
> > > that may explain the error message. you see.
> > >
> > > try setting VMDCUDANODISPLAYGPUS instead or
> > > at the same time to exclude the NVS from
> > > within VMD.
> > >
> > > cheers,
> > > axel.
> > >
> > > >
> > > > cheers
> > > >
> > > > Filippo
> > > >
> > > > Quoting John Stone <johns_at_ks.uiuc.edu>:
> > > >
> > > > >
> > > > > Hi,
> > > > > I suspect what is happening there is that the Quadro NVS 295 is
> > > > > running out of memory. I would suggest excluding it from your pool
> > > > > of CUDA GPUs. Also, what driver version are you running on your machine?
> > > > >
> > > > > Cheers,
> > > > > John Stone
> > > > > vmd_at_ks.uiuc.edu
> > > > >
> > > > > On Thu, Mar 04, 2010 at 06:29:44PM +0200, Filippo Federici wrote:
> > > > >> Hi,
> > > > >> I just tried to plot some isosurfaces of molecular orbitals,
> > > > >> with a very fine grid. Since the CPU is also very fast
> > > > >> (Core i7) is hard to tell if there is a performance improvement.
> > > > >> I still get that error on startup:
> > > > >>
> > > > >> Info) Detected 3 available CUDA accelerators:
> > > > >> Info) [0] Tesla C1060 30 SM_1.3 @ 1.30 GHz, 4095MB RAM, OIO,
> > > > >> ZCP
> > > > >> Info) [1] Quadro NVS 295 1 SM_1.1 @ 1.30 GHz, 255MB RAM, KTO
> > > > >> Info) [2] Tesla C1060 30 SM_1.3 @ 1.30 GHz, 4095MB RAM, OIO,
> > > > >> ZCP
> > > > >> Info) Creating CUDA device pool and initializing hardware...
> > > > >> CUDA error: no CUDA-capable device is available, line 62
> > > > >>
> > > > >> so why 2 Tesla C1060 are not CUDA capable devices according to line 62?
> > > > >>
> > > > >> thanks
> > > > >>
> > > > >> Filippo
> > > > >>
> > > > >> On Thu, 2010-03-04 at 08:11 -0600, John Stone wrote:
> > > > >> > Hi,
> > > > >> > Only the electrostatics, molecular orbital rendering, and
> > > > >> implicit ligand
> > > > >> > sampling features are accelerated by CUDA in VMD 1.8.7, so those are the
> > > > >> > only VMD tests that are meaningful for measuring CUDA performance.
> > > > >> >
> > > > >> > Try using the commands "volmap coulombmsm", "volmap ils" or the "Orbital"
> > > > >> > representation, all of these will use CUDA automatically.
> > > > >> >
> > > > >> > Please see this page for more information:
> > > > >> > http://www.ks.uiuc.edu/Research/vmd/vmd-1.8.7/cuda.html
> > > > >> >
> > > > >> > Let me know if you have questions.
> > > > >> >
> > > > >> > Cheers,
> > > > >> > John
> > > > >> >
> > > > >> >
> > > > >> >
> > > > >> > On Thu, Mar 04, 2010 at 09:18:14AM +0200, Filippo Federici wrote:
> > > > >> > > Good morning,
> > > > >> > >
> > > > >> > > we have installed a CUDA workstation and we are done
> > > > >> > > testing it, but we have some issues with VMD.
> > > > >> > > I downloaded the binary (LINUX64 CUDA) and apparently
> > > > >> > > it starts but i see no GPU improvements.
> > > > >> > > On startup the output tells:
> > > > >> > >
> > > > >> > > Info) -------------------------------------------------------------
> > > > >> > > Info) Multithreading available, 8 CPUs detected.
> > > > >> > > Info) Free system memory: 5269MB (66%)
> > > > >> > > Info) Detected 3 available CUDA accelerators:
> > > > >> > > Info) [0] Tesla C1060 30 SM_1.3 @ 1.30 GHz, 4095MB RAM, OIO,
> > > > >> > > ZCP
> > > > >> > > Info) [1] Quadro NVS 295 1 SM_1.1 @ 1.30 GHz, 255MB RAM, KTO
> > > > >> > > Info) [2] Tesla C1060 30 SM_1.3 @ 1.30 GHz, 4095MB RAM, OIO,
> > > > >> > > ZCP
> > > > >> > > Info) Creating CUDA device pool and initializing hardware...
> > > > >> > > CUDA error: no CUDA-capable device is available, line 62
> > > > >> > > Warning) Detected X11 'Composite' extension: if incorrect display occurs
> > > > >> > > Warning) try disabling this optional X server feature.
> > > > >> > > Info) OpenGL renderer: Quadro NVS 295/PCI/SSE2
> > > > >> > > Info) Features: STENCIL MSAA(4) MDE CVA MTX NPOT PP PS GLSL(OVF)
> > > > >> > > Info) Full GLSL rendering mode is available.
> > > > >> > > Info) Textures: 2-D (8192x8192), 3-D (2048x2048x2048), Multitexture
> > > > >> > > (4)
> > > > >> > > vmd > Info) VMD for LINUXAMD64, version 1.8.7 (August 1, 2009)
> > > > >> > > ---------------------------------------------------------------------
> > > > >> > > as you can see we have 2 Tesla GPUs and a Quadro NVS 295 for rendering.
> > > > >> > > The quadro is set to 'prohibited', but I tried to leave it on too with
> > > > >> > > no luck.
> > > > >> > > The driver is 190.53, the CUDA SDK samples run correctly, and I can run
> > > > >> > > some piece of code I wrote with no problem.
> > > > >> > > Any suggestion?!
> > > > >> > >
> > > > >> > > thanks in advance
> > > > >> > >
> > > > >> > > Filippo
> > > > >> > >
> > > > >> > >
> > > > >> >
> > > > >
> > > > > --
> > > > > NIH Resource for Macromolecular Modeling and Bioinformatics
> > > > > Beckman Institute for Advanced Science and Technology
> > > > > University of Illinois, 405 N. Mathews Ave, Urbana, IL 61801
> > > > > Email: johns_at_ks.uiuc.edu Phone: 217-244-3349
> > > > > WWW: http://www.ks.uiuc.edu/~johns/ Fax: 217-244-6078
> > > > >
> > > >
> > > >
> > > >
> > >
> > > --
> > > Dr. Axel Kohlmeyer akohlmey_at_gmail.com
> > > http://sites.google.com/site/akohlmey/
> > >
> > > Institute for Computational Molecular Science
> > > Temple University, Philadelphia PA, USA.
> > 

-- 
NIH Resource for Macromolecular Modeling and Bioinformatics
Beckman Institute for Advanced Science and Technology
University of Illinois, 405 N. Mathews Ave, Urbana, IL 61801
Email: johns_at_ks.uiuc.edu                 Phone: 217-244-3349
  WWW: http://www.ks.uiuc.edu/~johns/      Fax: 217-244-6078