Greetings,
I’m planning to set up a headless desktop computer to run some of my heftier OpenMC simulations. Likewise, I was wondering if there were any particularly important specs I should look out for when buying a desktop computer (i.e. CPU speeds, RAM, bus data speed, etc.).
Any advice as to which specs I should be keeping an eye out for and what range they should be within would be immensely appreciated. Thank you and happy holidays!
Hi @Josh_Guertler and welcome to the forum. OpenMC runs quite well multi-threaded, so I would recommend getting a CPU with as many cores as possible, even if it means that each core has a lower frequency. The performance is also generally pretty sensitive to the memory subsystem — the more data you can fit in cache the better, so all else being equal look for a CPU that has a large L3 cache. As for the total amount of RAM needed, that will entirely depend on the specifics of the problem you want to run so it’s hard to make a blanket recommendation there. Obviously the more RAM you have, the larger a problem (in terms of number of tallies, etc.) you’ll be able to run.
Thank you for your response.
Is 20 MB of L3 cache enough for depletion simulations with multiple channels? Thank you.
The amount of cache won’t actually impact the size of the problem you’re able to run, but it does impact the performance — a larger cache will lower the average memory access time. The total amount of RAM determines how large a problem you’re able to run. When you say “multiple channels”, I assume you mean multiple depletable materials? How many materials are you looking to handle?
My apologies for not specifying; I’m fairly new and getting used to some of the lingo. By channels I meant literal reactor channels, or depletion zones.
In terms of materials, I’m likely looking at around 10 different starting isotopes and dealing with at least 20 intermediate or output isotopes.
If you have a modest number of depletion zones, you should be able to get by with 4 or 8 GB of RAM. However, it’s always best to test out such assumptions on another machine (i.e., measure your actual memory use) to get a better feel for how much memory you’ll actually need.