Hi I read in this post that phoenix uses only 1-2 cores when simulating. Is it true? I was about to replace 10 core i9 with 18 cores one . Can anyone from chaos reply?
“What is the reason to pick the 7900X when for the same money and even less you can get the 16 core/32 threads AMD Threadripper 1950X which is ~40% faster than 7900X in multi-threaded workloads? Except for fluid simulations (Phoenix FD) where calculations are always done on only 1-2 cores, Threadripper 1950X will always be faster by like ~40%.”
This is completely not true and you can check this really easy in the Task manager Phoenix will use all of the cores that are available. Though, it’s really important to note that processor speed is more important than core count. Also, NUMA and multiprocessor architectures do not perform well with Phoenix. Simulating on one NUMA node is often faster than waiting for multiple nodes to synchronize.
You can read about the recommended hardware for simulation here.
Please have in mind that the processor is not the only factor for the speed of the simulation and the utilization of the cores may vary depending on your scene. In general Phoenix FD will try to utilize all the cores that are available, however some tasks like reading data from 3ds Max or Maya have to be done in single thread as the API of Max/Maya enforces that. If your 10 core processor gets a good utilization with you scenes chances are that the 18 core one will give you even better performance, however fast enough memory and high enough bus speed are required to feed all the cores with data. You can observe the load of the cores during simulation with your typical setup in the Task manager before you decide whether to upgrade.
It’s gotta be noted in the Intel docs for that particular CPU if it is either NUMA or it’s not - it’s very specific if it’s one way or the other - it’s not a matter of consideration
Is there a setting i have to check in order to speed up rendering? I am using a threadripper 1950x with 16+16 cores and in task manager it usually uses 20% off all the cores.. and overall load is 20% on the whole CPU . And calculation is kinda slow.
I came to the conclusion that when a simulation starts its much faster if i limit it to 6-10 cores, but when we have a whole lotta water(which happens in after 40-50 frames) its faster when i turn off the limiter and get the threadripper to use all its cores.
Hey, here is the screenshot while resim of wavelet on i9-7940x which has 28 cores. All of them are used, which makes phoenix very nice to play. Also, a m.2 drive or 500mb/s (actual, tested) ssd is recommended for grids denser than 25.000.000 cells. Also, at least 64gb of ram for really dense grids. Thats my experience.
Phoenix will use all of the cores that are available. Though, it’s really important to note that processor speed is more important than core count. Also, NUMA and multiprocessor architectures do not perform well with Phoenix. Simulating on one NUMA node is often faster than waiting for multiple nodes to synchronize.