[Spiros Kouzoupis]

Hi,

I want to know, i have an Intel i7 8700k 6 cores 12 threads, now i use ht on, my cpu works 100% when i run my simulation. In UI in ansys fluent i put 12 cores (meaning 12 so i can tell fluent that there are 12 logical cores) and 1 gpu. Would it run faster if i turned off ht and so put 6 cores and 1 gpu? Also would help to upgrade to i9 9900k?

[evcelica]

Why not try turning off HT and see for yourself?
Both those CPUs only have 2 memory channels, so i suspect you are memory bandwidth limited, not CPU limited, and there would not be too much difference between the CPUs, as it won't scale well past a few cores anyways. I would pick up a CPU with at least 4 memory channels, which would be roughly 2x the speed of your machine. CFD is usually memory bandwidth limited.

[Spiros Kouzoupis]

I switched off ht, but because other problems i cant compare the result in time. Every simulation cost me 1 month, if i want to compare i have to run the same simulation 2 times, so 2 months at least. Also ansys sometimes crashes by it self so its not possible to compare. Im Just asking by techincal point of view.

[Eng.Meshal]

Hi,
for 2D pipe flow, we useTb (bulk temperature) formula,
lengthInt(Velocity u*Y*Temperature)@radial/ lengthInt(Velocity u*Y)@radial
How to put the formula in comment mode to find results?
Please help

[flotus1]

Originally Posted by Spiros Kouzoupis

I switched off ht, but because other problems i cant compare the result in time. Every simulation cost me 1 month, if i want to compare i have to run the same simulation 2 times, so 2 months at least. Also ansys sometimes crashes by it self so its not possible to compare. Im Just asking by techincal point of view.

Step 1 is usually to get a test case that can be run in minutes, but is still representative of your actual application.
But with a run time of 1 month you can probably skip that entirely. You need something way more substantial than the small difference between SMT off/on. What that might be, depends on why your simulation takes this much time.

Maybe the cell count is so high that the simulation does not fit in memory. In that case, you would need more RAM. And better CPUs than a mainstream 6-core processor.
Maybe it's because your simulation runs a very high number of iterations/time steps. In which case there is not much you can do about it from a hardware perspective. Careful evaluation of the simulation parameters is the better approach here.
Or maybe your simulation does a lot of file-I/O, some of which might be unnecessary. Here a faster storage solution might help.

Tell us more about the simulations you are trying to run.