I’m new to the Unreal Engine and I’m having difficulty finding info on how the Editor uses CPU cores but maybe my search skills are weak.
I’m building myself a workstation for lots of different hobbies I have. I do some 3D rendering which will take advantage of as many cores as I can get, but how does the Unreal Editor use these?
I’ll probably build a dual Xeon e5-2011 system as I have a dual socket motherboard and 64GB of ECC from my previous setup. Would I be better off with fewer faster cores or will the Unreal Editor use all available cores.
You definitely need multiple cores as “background” jobs are sent to those cores all the time. The clock speed itself contributes to the final outcome but, I would say that higher clock less cores gets balanced/overpowered by lower clock more cores. Does those 200-300MHz more really make a difference overall, except that they costs more money per core.
In detail
My own setup is similar to what you plan to get; dual Xeon e5 at 2.6GHz, 8 cores on each CPU. Albeit, it’s a e5 v1, and since you are getting a v3, you could probably aim for a 10 to 14 cores per CPU, and let those rip through shaders and renders.
ShaderCompileWorkers will utilize every CPU core you have so the more the better (and it does make a lot of difference when you need to compile 2,000+ shaders ).
If you are going to be compiling C++ code, multiple cores help there as well, and I know this from my own example. While getting the Linux Editor to work past year, having a 7 minute full compile time versus 35-50 minutes made all the difference to me (I wouldn’t have had the patience to wait otherwise).
Depending on how much money you plan to spend, another option could be to get a single 18 core CPU and leave the other socket empty until you are ready to buy/need another one.
With 18 cores you are already ahead of the most dual CPU setups and, more power will always be available by adding another CPU. Sadly, the drawback would be the price of these CPUs, which is prohibitively high for most people.
Regarding 3D rendering, unless you do GPU rendering in Octane for example, you should see the benefit of all these cores, regardless of the clock speed they are running at.
I am running the editor on a duel Xeon E5-2650 V2 2.6GHz 32 core system and the engine uses all cores with every large render, so yes, it does seem to be capable of the total core/hyper-thread setup of the system.
The thing that, by far, uses the most CPU, is the lightmass light map calculator. That is a highly threaded application (and it also can run across a network, if you have more computers available.)
).