Try to switch off shaders, in Preferences (in the menu) in the main VNL window, on the Graphics tab. You need to restart VNL after this.This menu point is gray and unchecked. So, I assume it is off.
As a general note, your graphics driver is too old/simple to support high-performance graphics in VNL. If you want proper 3D graphics the machine should use vendor drivers (from NVidia or ATI or similar) rather than Mesa. That said, VNL should work anyway, we will look into why it fails - the shaders should have been automatically turned off.In 99% scientists use remote servers for calculations, don't have physical access to them, and don't need high performance graphics. I don't care about graphics at ll. Actually, I have only two options: 1) ssh connection to hpc cluster (or dedicated server), or 2) VNC on request (VNC is more convenient). In the first case, I will construct system and run it from command line using batch system. In the second case, after constructions of the system I can either run it in VNC (using 1 server with 16-80 CPUs) or submit it using batch system (unlimited CPUs).
Moreover, why are you running VNL over VNC? Why not run it on your own computer, that way the graphics will probably be a lot faster (a lot!!!).I don't need nice and fast graphics. I use VNL because it is convinient tools for the construction of nanodevoces. Running VNL locally on my laptop - is not a solution for many reasons.
In 99% scientists use remote servers for calculations, don't have physical access to them, and don't need high performance graphics. I don't care about graphics at ll. Actually, I have only two options: 1) ssh connection to hpc cluster (or dedicated server), or 2) VNC on request (VNC is more convenient). In the first case, I will construct system and run it from command line using batch system. In the second case, after constructions of the system I can either run it in VNC (using 1 server with 16-80 CPUs) or submit it using batch system (unlimited CPUs).
In my case, it is either dedicated HP DL585 with 4x16 cores, or HP SL230s with 2x8 cores. I even don't know what kind of graphics they have.
Actually, I have more questions:
1. How many CPUs VNL can use effectively? When it was running on DL585 with 64 cores it was using 1, 10 or 32 CPUs in different moments of time. Why not all 64?
2. When running on cluster, I can specify as many CPUs as I want. But taking into account what I noticed about CPU usage on 1 server, it seems that running on cluster will just waste of computational time. So, can VNL run effectively on cluster?
However, we offer you the possibility of using VNL to create your input files (both configurations from the Builder and scripts from the Scripter) and analyze your output files with all the dedicate plugins that the graphical interface offers you.Cluster has a license - I don't have, and I can't use cluster's license on my PC.
Unless you have a very good reason to do otherwise (do you?) you should do these operations from your local machine copying the files to and from your remote cluster.
It seems you are an expert user so if you have to modify your input files it won't be a problem to do that directly from the server without going through VNL.My area of expertise is a bit different, h-bar for me is mainly 0.
Also, if you do not care about graphics at all why do you run VNL?
I use VNL because it is convinient tools for the construction of nanodevoces. I like it.
1. How many CPUs VNL can use effectively? Why not all 64?
yes, of course. It seems you actually run in serial.
Please have a look to the Parallel Guide here http://quantumwise.com/support/manuals
mpiexec -localonly -n 64 ... Cluster has a license - I don't have, and I can't use cluster's license on my PC.
What I saw in the guide: "MKL will detect the presence of a multi-core processor and enable OpenMP automatically. It is recommended to let the software figure out itself how many threads to use optimally."
So, I asume that when I run calculations from GUI ("locally"), all available processors should be used.
If not (this is what I see now), should I use
mpiexec -localonly -n 64 ...
to force usage of all 64 cores?