User Notes for VASP

There are several versions of VASP installed on the LCRC servers. We recommend that you use the most recent builds of VASP 6.1.2.   These builds have better parallel performance, new features, and bug fixes.

There is a pure MPI build of VASP 6.1.2 optimized for the Broadwell (BDW) processors in /soft/vasp/6.1.2/bdw.  An example script is provided in /soft/vasp/6.1.2/bdw/ to run multithreaded on the BDWs.

There is a hybrid MPI/OpenMP build of VASP 6.1.2 optimized for the BDW processors in /soft/vasp/6.1.2/bdw_omp.  An example script is provided in /soft/vasp/6.1.2/bdw_omp/ to run multithreaded on the BDWs.

There is a hybrid MPI/OpenMP build of VASP 6.1.2 meant for use on the Knights Landing (KNL) processors in /soft/vasp/6.1.2/knl_omp. An example script is provided in /soft/vasp/6.1.2/knl_omp/ to run on the KNLs.

There are three VASP binaries for each architecture vasp-gam (gamma-point only), vasp-std (multiple k-points), and vasp-ncl (noncollinear spins).  Please use the binary which is appropriate for the type of VASP calculation you plan to run.

In general, the hybrid MPI/OpenMP build will be slower than the pure MPI build except when running on more than one core per atom. Above one core per atom, the hybrid MPI/OpenMP build will give better parallel scaling.

We have built VASP for the A100 GPUs on Swing.  Instructions for using VASP on Swing can be found at this link.

General notes:

• VASP is subject to license conditions that restrict its use. You need a VASP license for version 5 to become a member of the “vasp52” Unix group to use version 5 or 6 of VASP on Bebop. Please email with your license number if you are interested in using VASP on the LCRC clusters.

• Note that access to VASP 6 for members of the vasp52 will be temporary.  At some point in the future, you will need to upgrade your VASP 5 license to run the VASP 6 binary.  Contact your VASP vendor for more details.

• As a VASP 5 registered user, your access to VASP 5 will continue as per your license terms.  You do not need to upgrade to a new license to continue using VASP5! We recommend using 5.4.4.pl2 which is the recent release of vasp 5.4.4, if you prefer.  There example scripts for vasp.5.4.4 in /soft/vasp/5.4.4.pl2/{bdw,snb}/

• You should consider using the following tricks to increase the throughput of your calculations. Exploiting these tricks can increase your throughput by over an order of magnitude.◦ If you are running a calculation with only the gamma-point in the “integration” over reciprocal space, you can save a factor of two in elapsed time by using the gamma-point only version of VASP.

◦ The RMM-DIIS iterative matrix diagonalization algorithm (ALGO = Very_Fast) will give the best parallel performance. It approximately twice as fast (per iteration) as the Davidson algorithm (ALGO=NORMAL) ALGO=Fast uses RMM-DIIS most of the time and is recommended for metallic or reactive systems.

◦ The parallel performance of VASP depends on the NCORE parameter on the Broadwell nodes. You can improve performance by a factor of two (or more) by using an optimum value of NCORE. Our tests with the NCORE = 6 gives good performance on the Broadwells. However, you should optimize NCORE and the number of nodes to get the best performance for your model. The best performance will be a divisor of 18.  Note that if you are using multiple threads per process with hybrid MPI/OpenMP build of VASP the NCORE parameter will be ignored.

◦ The LPLANE input variable should be set to “.FALSE.” for best performance.

◦ If you are using multiple k-points, the KPAR parameter should be used to improve the parallel scaling. See section 6.57 of the VASP manual for more details (

◦ When performing a geometry optimization or molecular dynamics for a big system consider using low precision (PREC=LOW). You can save a lot of time in the initial stages of a geometry optimization or equilibrate your molecular dynamics simulations with lower precision and convergence criteria. You can always use higher precision to get final results once the geometry is close to converged or the trajectory is equilibrated with low precision. Verify that the drift in energy is not too large in an NVE trajectory at low precision before using higher precision in molecular dynamics. Please see section 6.2 of the VASP manual for more details (

◦ If you are doing a simulation involving water, an oxide, or any other system containing oxygen, consider using the soft oxygen pseudopotential (O_s). This will allow you to use a lower energy cutoff for the plane waves and save you significant time. The errors due to the softer pseudopotential are generally smaller than the errors due to the DFT approximation. But one should always verify the errors are not larger than you can tolerate.

The file vdw_kernel.bindat, for calculations involving the vdW-DF functional of Langreth, Lundqvist et al., is available in /soft/vasp.

Useful Tools:
• VESTA is a free crystal structure viewer and builder which can read and write POSCAR and CONTCAR files. VESTA can also visualize 3D data such are charge densities, potentials, and orbitals from CHG, CHGCAR, PARCHG, LOCPOT, and ELFCAR files from VASP. See
• R.I.N.G.S. can extract pair distribution functions, mean square displacements, and other properties from the trajectory files generated by VASP molecular dynamics runs. See
• Graeme Henkelman’s group, at UT Austin, maintain a collection called the “Transition State Tools for VASP.” These tools provide methods for finding saddle points, evaluating transition state theory (TST) rate constants, Bader charge analysis, and Adaptive Kinetic Monte Carlo for VASP. See
• VMD can be used to visualize structures and trajectories from the VASP XML file. See
• CIF2Cell is a tool to generate a POSCAR from a CIF (Crystallographic Information Framework) file. The program currently supports output for many popular electronic structure programs. See

• See the VASP resources page at for a list of additional useful tools for VASP.

User Notes for VASP Sidebar