You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hello, I am looking to build a workhorse linux PC optimized for heavy meep use and would appreciate guidance/tips from those who know computer architecture, MPI, and meep better than I do. The goal is to build a suitable system for large 3D simulations.
Below are my initial estimates for RAM, CPU and GPU resources based on meep documentation and others. I would very much appreciate any guidance the community can offer, as I am by no means an expert on this.
--
RAM: 256 GB minimum, expandable to 512 and beyond.
Following the memory requirements estimate guidance, initial calculations of an expected use-case give a lower bound of 158 GB. Since this does not account for PMLs, materials, and others, I figure opting for a minimum of 256 GB of RAM, with the option to expand to 512 or beyond, is a reasonable guess.
CPU: 96 cores on a single cpu, expandable to 2x CPUs
Currently looking at the AMD EPYC 9654. The reason for this is based on my reading of the expected speed up from parallel meep. My take from this is that it's better to have more cores in a single node to avoid network communication delays and cache contention. But I am far from an expert on this and could be totally incorrect. I assume, but do not know for sure, that AMD CPUs support some form of MPI that's compatible with meep.
GPU: Whatever is the top of the line, commercial option up to $2k
Meep doesn't support GPU acceleration, so really a good GPU is only needed to render potentially large 3D models and results post-processing.
--
Thanks again for the help! I'm happy to add any clarity needed, as there is no one "best" build and is case-specific.
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
-
Hello, I am looking to build a workhorse linux PC optimized for heavy meep use and would appreciate guidance/tips from those who know computer architecture, MPI, and meep better than I do. The goal is to build a suitable system for large 3D simulations.
Below are my initial estimates for RAM, CPU and GPU resources based on meep documentation and others. I would very much appreciate any guidance the community can offer, as I am by no means an expert on this.
--
RAM: 256 GB minimum, expandable to 512 and beyond.
Following the memory requirements estimate guidance, initial calculations of an expected use-case give a lower bound of 158 GB. Since this does not account for PMLs, materials, and others, I figure opting for a minimum of 256 GB of RAM, with the option to expand to 512 or beyond, is a reasonable guess.
CPU: 96 cores on a single cpu, expandable to 2x CPUs
Currently looking at the AMD EPYC 9654. The reason for this is based on my reading of the expected speed up from parallel meep. My take from this is that it's better to have more cores in a single node to avoid network communication delays and cache contention. But I am far from an expert on this and could be totally incorrect. I assume, but do not know for sure, that AMD CPUs support some form of MPI that's compatible with meep.
GPU: Whatever is the top of the line, commercial option up to $2k
Meep doesn't support GPU acceleration, so really a good GPU is only needed to render potentially large 3D models and results post-processing.
--
Thanks again for the help! I'm happy to add any clarity needed, as there is no one "best" build and is case-specific.
Beta Was this translation helpful? Give feedback.
All reactions