Different computation speeds can have multiple reasons. First of all, are you sure that both of your computations used the same basis? If I look here:

https://molmod.ugent.be/deltacodesdft

both packages should use different orbital types and of course those affect the SCF convergence immensely. Depending on your functional and integral calculation method, I would expect that your calculations will scale with the basis set size N by N^3 or N^4 and of course that's a difference.

Thanks for attention

But another question create! In Abinit for example, is there any option to choose basis set? I never do that

I mean, in ATK we can choose basis set like LACO or Plane-Wave and when we choose Plane-wave for similar material with same cut-off energy and K-points mesh as Abinit, the results are same but time is much quicker in VNL. I can't understand this!

Jurgen,

in VNL basis set is plane waves.