Is there a max number of cpu cores that Honeybee runDaylightAnalysis can utilize?


#1

Up to how many can you set on numOfCPUs input?


#2

Hi @IasonBournas,

You can practically use all your CPUs, but I would recommend keeping at lest one of the CPUs free for other processses on your machine. This article tells you how to find out the number of CPUs your system has.


#3

Ok then, so no upper limit, thank you devang!


#4

@IasonBournas ,

It’s important to clarify that the component will take any number of CPUs and split up the analysis into that many sections but, if you select more CPUs than what your computer has, the analysis will be very slow and will potentially crash your computer. So you should limit the number you input to the component to the amount that @devang recommended.


#5

Thanks Chris,
the upper limit was my main concern,
your inputs are greatly appreciated!


#6

That is a good practice however in honeybee even if you input a higher number honeybee will adjust the number to the max available cpus.


#7

Hi Mostapha,
i think the following question is relevant to the topic so i did not create a new one,
correct me if i am wrong though:

(1)Is the maximum number of cpus always the optimum input?
(2)For instance, could it be that depending on the number of sensor points there could be a lower number of (the same) cpus that could calculate faster? A kind of calculation of the optimum number of cpus for specific models.

I am talking about an annual simulation DC case. Only the illuminance calculation. The scene & materials being written prior with rad files.
Do you have some input on this?


#8

Maximum number of cpus is not always the case. In most of the cases leaving one of the cpus free helps the overall performance. Also if the number of points is not so many using multiple cpus can actually add to the overall time for the calculation.

In Honeybee (and Honeybee[+]) the parallel run basically is separating the test points and using a separate CPU to run the analysis for each of them.


#9

Valuable info, thanks Mostapha!


#10

Hello everyone,

Very interesting thread! I have a side questions to the original.

What is the better option for the CPU regarding f.x annual daylight simulations, having more cores with less speed or less cores but faster?
For example I have two machines 1) 12 processors at 3.5 GHz or 2) 8 processors at 4.0 GHz