Re: [SIESTA-L] why this kind of model spent so much time on calculating?

Dear Riccardo : Have you ever done this kind of model? for electrode with a scatter region at center. if yes, how did you set MD.ConstantVolume=true MD.VariableCell=false?? ​since I want to get the .DM file for TranSiesta calculation. Thank you!​ -- Best Regards 邱芳瑜 Chiu Fang Yu 國立成功大學 材料科學與

Re: [SIESTA-L] why this kind of model spent so much time on calculating?

>>> If I use KPOINTS MonkHorst Pack 2 2 2 for first calculation and then chenge it to KPOINTS MonkHorst Pack 4 4 4 for second calculation. should I use the .XV .DM .CG file for the second calculation?? thank you! If you want the second calculation to be a continuation of the first calculatio

Re: [SIESTA-L] why this kind of model spent so much time on calculating?

Dear Riccardo: If I use KPOINTS MonkHorst Pack 2 2 2 for first calculation and then chenge it to KPOINTS MonkHorst Pack 4 4 4 for second calculation. should I use the .XV .DM .CG file for the second calculation?? thank you! -- Best Regards 邱芳瑜 Chiu Fang Yu 國立成功大學 材料科學與工程學系碩一 MOBILE：0930287221

Re: [SIESTA-L] why this kind of model spent so much time on calculating?

>> when you do this two step for calculation (1) relax with DM.Tolerance 1E-3 (2) reconverge the electronic structure down to DM.Tolerance 1E-4 will you save .XV .DM .CG file for the second step calculation? Of course, otherwise the second calculation would start again fro scratch. Also y

Re: [SIESTA-L] why this kind of model spent so much time on calculating?

Dear Riccardo: when you do this two step for calculation (1) relax with DM.Tolerance 1E-3 (2) reconverge the electronic structure down to DM.Tolerance 1E-4 will you save .XV .DM .CG file for the second step calculation? thank you very much! -- Best Regards 邱芳瑜 Chiu Fang Yu 國立成功大學 材料科學與工程學系碩一

Re: [SIESTA-L] why this kind of model spent so much time on calculating?

It depends on a number of different parameters, like the number of k-points, for instance. The textbook answer is that you should run tests to find out the optimum number of cores for your problem. Having said that, if this is a Gamma point calcualtion and you have access to some "modern" co

Re: [SIESTA-L] why this kind of model spent so much time on calculating?

Dear Riccardo: so according to my system (94 atoms) how many cores you will use to do the calculation? thanks for replying! -- Best Regards 邱芳瑜 Chiu Fang Yu 國立成功大學 材料科學與工程學系碩一 MOBILE：0930287221(中華) GMAIL：joyce79...@gmail.com 2014-09-17 15:38 GMT+08:00 Riccardo Rurali : > On 9/17/14 3:57

Re: [SIESTA-L] why this kind of model spent so much time on calculating?

On 9/17/14 3:57 AM, joyce79928cc . wrote: Dear Riccardo: so in your case, when you do the calculation with DM.Tolerance 1E-3 and MD.MaxForceTol0.04 eV/Ang. how many CPUs you use? thank you! Well, of course it depends on the size of the system. The converge parameters indicated above are not

Re: [SIESTA-L] why this kind of model spent so much time on calculating?

Dear Riccardo: so in your case, when you do the calculation with DM.Tolerance 1E-3 and MD.MaxForceTol 0.04 eV/Ang. how many CPUs you use? thank you! -- Best Regards 邱芳瑜 Chiu Fang Yu 國立成功大學 材料科學與工程學系碩一 MOBILE：0930287221(中華) GMAIL：joyce79...@gmail.com 2014-09-16 16:55 GMT+08:00 Riccardo Ru

Re: [SIESTA-L] why this kind of model spent so much time on calculating?

Dear: And I want to ask another question. Should this model set MD.ConstantVolume=true MD.VariableCell=false?? Thank you Mads Engelund 於 2014年9月16日星期二寫道： > Dear Chiu, > > Your force tolerance is the maximum force you allow, so it depends how > accurate a relaxation you need. > The lower force

Re: [SIESTA-L] why this kind of model spent so much time on calculating?

Dear: And I have use a 32core CPU to calculate this model. But it shows no faster than two 8 core cpu What's the reason for this? Thank you! Mads Engelund 於 2014年9月16日星期二寫道： > Dear Chiu, > > I am mostly running on 8 cores myself. Not knowing the full details, I > think 2 days does not sound un

Re: [SIESTA-L] why this kind of model spent so much time on calculating?

Dear Chiu, Your force tolerance is the maximum force you allow, so it depends how accurate a relaxation you need. The lower force tolerance you have, the more accurate your calculation have to be to calculate the forces. And as Ricardo pointed out, you can try first a less accurate relaxation,

Re: [SIESTA-L] why this kind of model spent so much time on calculating?

dear: you mean I should set MD.MaxForceTol with a higher number like change from 0.04 to 0.1 eV/Ang? thanks! -- Best Regards 邱芳瑜 Chiu Fang Yu 國立成功大學 材料科學與工程學系碩一 MOBILE：0930287221(中華) GMAIL：joyce79...@gmail.com 2014-09-16 16:43 GMT+08:00 Mads Engelund : > Dear Chiu, > > I am mostly runnin

Re: [SIESTA-L] why this kind of model spent so much time on calculating?

Hej Mads. On 9/16/14 10:43 AM, Mads Engelund wrote: I think DM.Tolerance 1E-3 is too high for relaxation if your force tolerance is below 0.1 eV/Ang. This will make things take longer, sorry. Not in my experience. I run all my Siesta relaxations with DM.Tolerance 1E-3. When convergence on for

Re: [SIESTA-L] why this kind of model spent so much time on calculating?

Dear Chiu, I am mostly running on 8 cores myself. Not knowing the full details, I think 2 days does not sound unreasonable for the system you are showing. I think DM.Tolerance 1E-3 is too high for relaxation if your force tolerance is below 0.1 eV/Ang. This will make things take longer, sorry

[SIESTA-L] why this kind of model spent so much time on calculating?

Dear: I have my model and input file like attached. I have set the flags with KPOINTS: MonkhorstPack = 2,2,4 MD.ConstantVolume=T MD.VariableCell=F DM.MixingWeight 0.1 #default:0.25 DM.NumberPulay 10 # Pulay convergency accelerator DM.Tolerance 1E-3 # EDIFF ElectronicTemperature 400 K # Defaul