CPU for FEM - Numerical simulations

An FX8320 should be great value for money for computations.

Thank you very much for your reply!

Do you think that FX8320 would be better than a FX8350? Because if the 8350 is a better performer in computations and a PC with 8350 is in the budget, I think I will go for that solution. Am I missing something?

Another question, some people told me that the AMD 8350 is not a true 8 cores because there are some parts shared every 2 cores and so they are not independent, for example there are just 4 floating points units instead of 8, and they told me that these aspects can affect in a negative way the calculation time.

For these reasons they suggested me to consider an i5 Intel CPU, is this advice a good one? What do you think?

I think that the best solution would be an i7, but I also think that it is out of budget. Am I wrong?

The objective is to minimize the calculation time because I need to run simulations as fast as possibile (always considering the available budget).

Thank you very much again!

Well my point is the 8320 is the same chip only cheaper. Just OC it to match the 8350.

What they have told you about shared resources is true. However if your computations are multithreaded (as they should be), the FX should beat the i5.

An i7 might be possible within your budget but a tight fit indeed.

I am not familiar with FEM programs...

It is correct that the AMD FX 8xxx series are not true 8 core CPUs because they share resources. However, from a technical standpoint it is better than Intel's Hyper Threading (HT) technology which more or less waits for a moments when the core(s) is idling to processes a different thread. The caveat is that Intel's Ivy Bridge CPUs can execute more instructions per clock cycle than AMD CPUs; called IPC. HT can boost performance anywhere from 10% - 40% depending on how well the program has been designed to use HT. Perhaps HT support has not been implemented at all which means 0%.

You should do a little research into the FEM program(s) you are using. 1st to see if HT is supported and 2nd to see if it supports the AVX instruction set. AVX can help improve performance of financial and scientific modelling calculations because it doubles the number of vector registers’ width from the 128-bit to 256-bit for floating point calculations. By doubling the register size to 256 bits, AVX fits eight 32-bit values into each register, enabling 16 floating-point operations per cycle. This effectively, in one fell swoop, doubles the peak performance (under ideal situations).

I believe AMD has licensed the AVX instruction set from Intel, however, due to higher Intel's CPU can execute more instructions per clock cycle than AMD CPUs, spending more money on an i7-3770k can cut down on calculation times.