I am using gaussian 9W 32-bit for some research purposes with this
%NProcShared=2
%Mem=1.5GB input
But, I simply can’t extend number of processors and RAM to make the calculations faster. I saw several researchers talking in researchgate about using LindaWorkers to make the calculation faster. I also saw some comments where some of them mentioning 12 processors and 4 GB RAM (%mem) where I can only use 1650 MB with 4 processors maximum.
So, can anyone please tell me how to use more memory and processors in Gaussian 9 and what is LindaWorkers? Is this a separate software than Gaussian? Should I need to buy it even if have Gaussian 9 installed? and how can I use it?
If you're facing resource constraints and require quick calculations, I would highly recommend considering Orca and RI-DFT. The great news is that Orca 5 is available for free to academics, and it leverages Gaussian-like syntax. (Orca: https://www.orcasoftware.de/tutorials_orca/first_steps/install.html)
Here's a helpful link you may find useful about Linda: http://wild.life.nctu.edu.tw/~jsyu/compchem/g09/g09ur/m_linda.htm
Based on my experience, I haven't used Gaussian with Linda, even when working with 64 core nodes. The Gaussian manual indicates that you must compile the code with Linda, which will enable you to execute calculations across the nodes. I should point out that running jobs across multiple nodes is not standard practice for a Gaussian calculation, as the community typically considers about 100-150 atoms as the limit for a conventional DFT calculation. Beyond that limit, we'll apply truncated models or QM/MM approximations.
Furthermore, if you're running jobs on SMP multiprocessing protocol in a single node, Linda will not necessarily result in faster calculations. Here's a useful resource to learn more about SMP:
https://www.techtarget.com/searchdatacenter/definition/SMP
I hope this message helps!
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size