Hi,
I would like to build a small cluster. Cluster should solve CFD / fluid dynamics / simulation tasks (diff.equations iteratively). I would like to build it from workstations (up to 15 depends on budget and answer on question no.3) with a head-station which should store and distribute the task.
1. Which of the CPUs are the most suitable for such an application ?
2. How to measure their ratio - price/comp.power ?
3. What should I think to have - more cores and less station or more stations with less cores ?
4. Is there an optimum in RAM size per core, or amount od DIMM per core ?
5. Is it necessary to have ECC registered RAM to obtain corect solution after runing the simulation for ~hundred days ?
6. Should I prefer server over desktop solution while running some tasks continously for a long period ?
7. Should I prefer intel over AMD from the 'energy consumption' point of view? Do they heat more / cooling needs of the cluster ?
8. What about having non-heterogenous nodes in cluster ? To have some cheaper mid-end nodes for common computing and to have few pro-nodes.
I could choose anything (up to my budget). Having any working hardware gives me the profit of HPC. Since I have to choose one I stucked. Please, help me to make any kick-off decision.
thanks
I would like to build a small cluster. Cluster should solve CFD / fluid dynamics / simulation tasks (diff.equations iteratively). I would like to build it from workstations (up to 15 depends on budget and answer on question no.3) with a head-station which should store and distribute the task.
1. Which of the CPUs are the most suitable for such an application ?
2. How to measure their ratio - price/comp.power ?
3. What should I think to have - more cores and less station or more stations with less cores ?
4. Is there an optimum in RAM size per core, or amount od DIMM per core ?
5. Is it necessary to have ECC registered RAM to obtain corect solution after runing the simulation for ~hundred days ?
6. Should I prefer server over desktop solution while running some tasks continously for a long period ?
7. Should I prefer intel over AMD from the 'energy consumption' point of view? Do they heat more / cooling needs of the cluster ?
8. What about having non-heterogenous nodes in cluster ? To have some cheaper mid-end nodes for common computing and to have few pro-nodes.
I could choose anything (up to my budget). Having any working hardware gives me the profit of HPC. Since I have to choose one I stucked. Please, help me to make any kick-off decision.
thanks
Facebook
Twitter
Instagram