ABINIT Discussion Forums

3 years ago, I did a GW calculations for 24 atom system, I first did it in Abinit and it took me more than 8 hours unfinished GW calculation. (I think its taking huge time in solving the Self Energy Sigma?) My supervisors advised me to use VASP since that is what was installed in the Lab, I am the only one insisting to use Abinit, while the rest were using VASP. Vasp was donated to the Lab so I guess that is why they Insisted that we make use of it.

So I did my study in VASP and it just took me two hours done. As I recall, in Abinit I can adjust the energy bands at which GW will be calculated, In VASP all energy bands is calculated in GW but still VASP is faster!?

Now I am in a different Lab and I am configuring a new system much powerful than my previous Lab, and I am going to use Abinit, Yey!

This time, I want to be able to get rid of the things that is making my GW calculations slow, but I dont know how?

Things that speed things up in GW VASP

1. Lower the GW cut off energy

I did the same thing in Abinit before but nothing could make it comparable to the VASP speed, so I am completely clueless! Anybody could please help?

If its important,

1. Abinit is on MPI and ground state calculation was fast, comparable to VASP
2. Both were using Atlas linear algebra

Dear Dominic,

In order to gain speed when using abinit, you have to compile it with mpiifort compiler and mkl libraries.

Furthermore, it could be possible that VASP has a better symmetry detection, and hence only for a smaller number of k points QP energies are calculated.

If you are interested in fast G0W0 + BSE calculations, i recommend you YAMBO.


Bests
LUCA