VMD-L Mailing List

From: Frederico De Santana Pontes (F.Pontes_at_leeds.ac.uk)
Date: Tue Nov 21 2023 - 07:39:39 CST

Hi Diego,

Well, I just tested the OpenGL stuff, and the performance is really below that you reported:

1772 frames in 5.0 seconds = 354.249 FPS
1891 frames in 5.0 seconds = 378.146 FPS
1887 frames in 5.0 seconds = 377.296 FPS
1815 frames in 5.0 seconds = 362.838 FPS
1886 frames in 5.0 seconds = 377.110 FPS
1885 frames in 5.0 seconds = 376.931 FPS

Thank you for instructions about correct installation of VMD. I don't have permission to install a software in my desktop. Everything is done by IT team. I'll try an alternative way to install, but I already asked help for the IT. I included some questions raised here in my report and see what they can do.

Thank you again,

Fred


________________________________
From: Diego Gomes <diego.enry_at_gmail.com>
Sent: 21 November 2023 13:27
To: Frederico De Santana Pontes <F.Pontes_at_leeds.ac.uk>
Cc: Vermaas, Josh <vermaasj_at_msu.edu>; vmd-l_at_ks.uiuc.edu <vmd-l_at_ks.uiuc.edu>
Subject: Re: vmd-l: How to evaluate VMD performance properly?

nvidia-smi all looks good.
I forgot to mention a basic OpenGL check: glxgears

dgomes_at_csm-chicago:~/Downloads$ glxgears
13804 frames in 5.0 seconds = 2760.712 FPS

BTW, did you try installing VMD yourself?
1) Get the latest Alpha build: VMD 1.9.4a55 for RHEL 7+ Linux, 64-bit Intel x86 (x86_64), RTX RTRT (Oct 19, 2021):
https://www.ks.uiuc.edu/Development/Download/download.cgi?PackageName=VMD

2) extract the tarball
tar xfz vmd-1.9.4a55.bin.LINUXAMD64-CUDA102-OptiX650-OSPRay185-RTXRTRT.opengl.tar.gz

3) Edit the configure file with the appropriate paths
# Directory where VMD startup script is installed, should be in users' paths.
#$install_bin_dir="/usr/local/bin";
$install_bin_dir="/home/dgomes/software/vmd/bin";

# Directory where VMD files and executables are installed
#$install_library_dir="/usr/local/lib/$install_name";
$install_library_dir="/home/dgomes/software/vmd/lib/$install_name";

4) Run the configure script
./configure

5) Install
cd src
make install

6) Launch from where you installed.
 /home/dgomes/software/vmd/bin/vmd


If it is still slow, showing the text message as in the a attached screenshot might help debug.

Friendly,
Diego.

On Tue, Nov 21, 2023 at 3:39 AM Frederico De Santana Pontes <F.Pontes_at_leeds.ac.uk<mailto:F.Pontes_at_leeds.ac.uk>> wrote:
Hi guys!

Thank you for using your time to answer my question.

Josh,
I put at the end of this message the output on screen after type 'nvidia-smi'. I don't see anything wrong. When I describe that I had problems to visualize/rotate a 500k system is just one structure file without a loaded trajectory.

Diego,
After your message and from Josh, I suspect something wrong with the VMD installation related to NVIDIA and OpenGL. I did the test you suggested me and I could not rotate smoothly my simulation box. VMD installation in my desktop was made by the IT team from the University and I'll try to get details about it with them.

thank all,

Fred



| NVIDIA-SMI 535.54.03 Driver Version: 535.54.03 CUDA Version: 12.2 |
|-----------------------------------------+----------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+======================+======================|
| 0 NVIDIA T1000 Off | 00000000:65:00.0 On | N/A |
| 35% 36C P8 N/A / 50W | 571MiB / 4096MiB | 1% Default |
| | | N/A |
+-----------------------------------------+----------------------+----------------------+

+---------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=======================================================================================|
| 0 N/A N/A 3616 G /usr/libexec/Xorg 193MiB |
| 0 N/A N/A 3803 G /usr/bin/gnome-shell 124MiB |
| 0 N/A N/A 4291 G /usr/lib64/firefox/firefox 145MiB |
| 0 N/A N/A 5775 G ...70647646,1451225651515119381,262144 40MiB |
| 0 N/A N/A 6447 C ...2rjo3fzfzoskw5/lib64/vmd_LINUXAMD64 62MiB |
+---------------------------------------------------------------------------------------+
________________________________
From: Diego Gomes <diego.enry_at_gmail.com<mailto:diego.enry_at_gmail.com>>
Sent: 21 November 2023 04:30
To: Vermaas, Josh <vermaasj_at_msu.edu<mailto:vermaasj_at_msu.edu>>
Cc: Frederico De Santana Pontes <F.Pontes_at_leeds.ac.uk<mailto:F.Pontes_at_leeds.ac.uk>>; vmd-l_at_ks.uiuc.edu<mailto:vmd-l_at_ks.uiuc.edu> <vmd-l_at_ks.uiuc.edu<mailto:vmd-l_at_ks.uiuc.edu>>
Subject: Re: vmd-l: How to evaluate VMD performance properly?

Hi Fred,
I must agree with Josh, 500k particles should not be an issue for VMD to represent as Lines.
If it is not RAM or issues with nvidia-drivers, can you try the following?

Extensions->Modelling->Add Solvation Box. Mark on "WaterBox only" and make a 200x200x200 box.
It should give you >700k atoms, and rotate smoothly. I tried a 300^3 box with 2.6 million atoms that rotates smoothly on my laptop (M1 mac).

Please let us know if that doesn't work.

Friendly,
Diego.



On Mon, Nov 20, 2023 at 10:41 AM Vermaas, Josh <vermaasj_at_msu.edu<mailto:vermaasj_at_msu.edu>> wrote:

Hi Fred,



How many frames do you have loaded when this happens? 500k particles really isn’t that much to be drawing lines for normally, although some have more bonds drawn than others. I think there are two possibilities:



  1. Under RAM pressure (like when you have a lot of trajectory loaded), many linux systems are set up to use hard drive space as swap. This is terrible for performance, as the latency for most hard drives is much higher than RAM. You can check this pretty easily with “top” to see how much memory you still have available.
  2. Are you *sure* you are using NVIDIA drivers? Does “nvidia-smi” return something sane? Admittedly a T1000 is on the older side, but if you are using the open-source NVIDIA drivers, you will have a bad time in OpenGL applications.



-Josh



From: <owner-vmd-l_at_ks.uiuc.edu<mailto:owner-vmd-l_at_ks.uiuc.edu>> on behalf of Frederico De Santana Pontes <F.Pontes_at_leeds.ac.uk<mailto:F.Pontes_at_leeds.ac.uk>>
Date: Monday, November 20, 2023 at 8:03 AM
To: "vmd-l_at_ks.uiuc.edu<mailto:vmd-l_at_ks.uiuc.edu>" <vmd-l_at_ks.uiuc.edu<mailto:vmd-l_at_ks.uiuc.edu>>
Subject: vmd-l: How to evaluate VMD performance properly?



Hi everyone!



I have some difficulties to work with VMD 1.9.3 on my machine. Rotating a system (~500k Coarse-grain beads)in a default configuration (Coloring Method: Name; Drawing Method: Lines; Material: Opaque) is a really hard task.



I know that my system is quite large, but I'm running VMD locally and my machine has the following features:



OS: Red Hat Enterprise Linux 8.8

Memory: 62.3 GiB

Processor Intel Xeon(R) W-2245 CPU @ 3.90 GHz x 16

NVIDIA T1000/PCle/SSE2

GNOME v.3.32.2



I don't know if the difficulties to manipulate/rotate my system in VMD is due to the size system or some other reason, which I could improve on a new installation. Maybe I could make some tests to evaluate VMD performance, but I don't know which kind of test.

Someone could help me in this situation?



Kind regards,



Fred




--
Diego Enry B. Gomes, PhD
Department of Physics at Auburn University &
NIH Center for Macromolecular Modeling and Visualization
Leach Science Center - Ste. 3182 - Auburn, AL
dgomes_at_auburn.edu<mailto:dgomes_at_auburn.edu>


--
Diego Enry B. Gomes, PhD
Department of Physics at Auburn University &
NIH Center for Macromolecular Modeling and Visualization
Leach Science Center - Ste. 3182 - Auburn, AL
dgomes_at_auburn.edu<mailto:dgomes_at_auburn.edu>