VMD-L Mailing List

From: John Stone (johns_at_ks.uiuc.edu)
Date: Wed Jun 10 2020 - 11:11:36 CDT

Bart,
  I think that your "black screen" issue with the interactive OptiX
rendering path is related to DCV, but the memory consumption problem
is almost certainly unrelated to DCV.

I've been continuing my OptiX/RTX tests this morning and I seem to have
found the cause of the OptiX rendering performance anomaly I described, and
it may be related to your memory consumption problem. OptiX RTX mode has
a special feature intended to improve performance, which is supposed to
release GPU memory associated with RTX triangle geometry buffers as soon
as the acceleration structure has been built. Apparently there are cases
where it will choose note to free the GPU memory until much later, but
the application has no way of knowing if/when this has occured.

I've modified my OptiX rendering code to disable this RTX-specific mode,
and that seems to have cured the performance anomaly I had been observing
in my local testing. It's possible that the same code is what has been
causing your memory consumption problem. If so, I should be able to give
you a test build of VMD to try and you can report if the solves that part of
your problems or not.

Best,
  John

On Wed, Jun 10, 2020 at 05:58:01PM +0200, Bart Bruininks wrote:
> This occurs with OPTIX on and two graphics cards over a NIVE DCV service.
> If you would like, I could ask which exact version we are running for
> NICE. I added a short video which hopefully illustrates the issue clearly.
> I will see if the suggested helps and report the results when I have them.
>
> Op wo 10 jun. 2020 om 17:06 schreef John Stone <[1]johns_at_ks.uiuc.edu>:
>
> Bart,
> Â I haven't had the GPU run out of memory, but I've been observing an
> unexplained performance loss anomaly with a simple movie rendering
> test that just renders a virus scene in a tight loop a few hundred
> times. I haven't seen GPU memory use climb noticably in my test, but
> I will keep an eye out for this.Â
>
> If I understood your symptoms, you encounter the problem when RTX is
> enabled, but NOT if RTX is disabled?
>
> What happens if you set this environment variable before you run VMD:
> bash:
> export VMDOPTIXDESTROYCONTEXT=1
> csh:
> setenv VMDOPTIXDESTROYCONTEXT 1
>
> Best,
> Â John
>
> On Wed, Jun 10, 2020 at 03:05:10PM +0200, Bart Bruininks wrote:
> >Â Â Dear VMDers,
> >Â Â I am using the 1.9.4a42 build to render videos of trajectories on
> a
> >Â Â server. There used to be the issues that after a while frames
> would not be
> >Â Â written anymore. However, atm it seems all frames are written,
> but the
> >Â Â internal rendering data structures are not cleaned rendering
> between
> >Â Â frames. This causes my VRAM to get filled during successive
> renders and
> >Â Â eventually my GPUs kill themselves crashing VMD and sometimes
> taking some
> >Â Â other stuff with them (window manager etc.). I can understand
> that having
> >Â Â no free VRAM could result in such issues, but I guess it
> shouldn't build
> >Â Â up like that in the first place.
> >Â Â VMD:
> >Â Â 1.9.4a42
> >Â Â Nvidia Drivers:
> >Â Â NVIDIA-SMI 440.82 Ã* Driver Version: 440.82 Ã*Â Ã*Â Ã*Â CUDA
> Version: 10.2
> >Â Â GPU's:
> >Â Â OptiXRenderer) Creating OptiX window: 672 x 805..
> >Â Â OptiXRenderer) VMD TachyonL-OptiX Interactive Ray Tracer help:
> >Â Â OptiXRenderer) ===============================================
> >Â Â OptiXRenderer) Using 2 devices:
> >Â Â OptiXRenderer) [0] Quadro RTX 8000 Ã*Â Ã*Â Ã* CUDA[0], 47.5GB
> RAM, KTO
> >Â Â OptiXRenderer) [1] Quadro RTX 8000 Ã*Â Ã*Â Ã* CUDA[1], 47.5GB
> RAM
> >Â Â Cheers,
> >Â Â Bart
>
> --
> NIH Center for Macromolecular Modeling and Bioinformatics
> Beckman Institute for Advanced Science and Technology
> University of Illinois, 405 N. Mathews Ave, Urbana, IL 61801
> [2]http://www.ks.uiuc.edu/~johns/Â Â Â Â Â Â Phone: 217-244-3349
> [3]http://www.ks.uiuc.edu/Research/vmd/Â Â Â
>
> References
>
> Visible links
> 1. mailto:johns_at_ks.uiuc.edu
> 2. http://www.ks.uiuc.edu/~johns/
> 3. http://www.ks.uiuc.edu/Research/vmd/ 

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