From: Rune Thomas Kidmose (
Date: Wed Mar 07 2018 - 12:59:54 CST

Hi John,

I just did a couple of quick test, and this was exactly what I needed! Thank you very much for your help!

Now I got a bit curious regarding what else could be tuned through such commands.

I found this document written by you (I assume from the name):

Can I use this document as a good introduction to all "under-the-hood" settings of tachyon, or do you know a better source?

Again thx!


Fra: John Stone <>
Sendt: 7. marts 2018 17:19
Til: Rune Thomas Kidmose
Emne: Re: vmd-l: TachyonL-optix rendering question

  If you want to use advanced ray tracing features like depth of field
and the like, you'll need to increase the default number of samples
used for the batch mode renderings. When you run the interactive ray tracer,
you'll see that as it shoots more and more stochastic samples, it prints
the accumulated sample count in the VMD text console. When you want to
reproduce the same sample depth in a batch mod ray tracer, you need to tell
it to use that same sample count, as follows (the same syntax works for
the other Tachyon-derived ray tracing engines also):

# Set antialiasing sample count (number of primary rays per pixel)
# The default is 12
render aasamples TachyonLOptiXInternal 1024

# Set ambient occlusion lighting samples (AO feeler rays per hit)
# The default is 12, which is good when only using 12 AA rays per pixel,
# but when using depth of field or other features that require very large
# sample counts, it is best to set AO samples to 1, to avoid OptiX/CUDA
# kernel timeouts (RT kernel launches that take more than 0.7 seconds total)
render aosamples TachyonLOptiXInternal 1

  John Stone

On Wed, Mar 07, 2018 at 03:49:40PM +0000, Rune Thomas Kidmose wrote:
> Hi,
> Lately I have been playing around with the very powerful tachyonL-optix rendering on relative big models (70S ribosome and 100S ribosomes) in complex scenes.
> I prefer to do all my final movies and images from scripts, so I am also calling tachyonL-optix this way. However, I have noticed a big change in quality between what the interactive TachyonL-optix engine gives me (from GUI of course) and what I get from the internal tachyonL-optix engine via my script.
> When I test in the GUI on large complex scenes ( movie of ribosome trajectories, a lot of glass, shadows, reflection and very large molecules etc) I have to let the interactive tachyonL-optix scene render for quite a while, to get everything looking smooth and nice. At that point I stop the render and save the image. When I do the exact same scene via the internal tachyonL-optix from scripts (but also from the GUI !) the end results is far from the same quality regarding smoothness and details. It applies the same materials, lightning view matrix display size etc. but the rendered image looks much worse.
> I guess it is has something to do about how long time the image is allowed to render before being written as a file. Is there any way of increasing this time via an extra tachyon render command or flag I can state? Or is it something completely different that is the cause?
> Regarding the hardware I test it on (if relevant) it is a dell linux workstation with dual GPU (quadro M5000 and nvidia 980TI) plenty of memory and 16 cores. I can add details if needed of course.
> And I am using VMD 1.93 from ubuntu 16.04
> Cheers
> Ruki

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