Currently I am using a core i7 cpu which has 4cores/8threads and 2channel memory. I want to know if I upgrade it to a 6core/12threads cpu with 4channel memory access, how much the simulation speed will change? Is there any maximum number of cores supported by lumerical fdtd solutions.
I can’t make any promises with respect to the amount of performance increase you will get by upgrading, but here are a few general guidelines.
The number of cores is usually not the factor that limits the simulation speed. Instead, the limiting factor is more often the speed of the memory bus that connects the RAM and CPU. In most computers, the memory bus is not fast enough to keep all of the cores running at full speed. For this reason, simply adding more cores is unlikely to significantly change the simulation speed. However, if you can increase the memory bus speed by upgrading to 4-channels or by using RAM with a faster clock speed, the simulation will probably run faster.
There is no limitation on the number of processes that you’re allowed to use, but in practice, the computer hardware will limit the number of processes that give good performance.
I agree with @ckopetski. From my experience on running on a 6-core machine and 4 core machine, the 6 core machine does perform better (Intel 2011v3 series). It also depends on the simulation size and I have generally employed the total FDTD solver speed. My 4 cores, Core i5 machine yields around ~120 Mnodes/s while the 6 core machine (based on the Intel 2011 series) typically gave around 220 Mnodes/s. The six core machine had more memory interfaces and hyper threading enabled.