topic Re: Increase speed of Archicad through distributed computing in Visualization

Increase speed of Archicad through distributed computing?

Anonymous — Thu, 11 May 2023 10:04:00 GMT

I know that a lot of 3D renderers out there support distributing renders over multiple nodes in a render farm. For an arhictect, rendering a 3d model is not the most common task throughout the day, and so I would like to know of ways to increase the speed of other common tasks.

Some tasks I could perceive would benefit from distributed computation over multiple render nodes would be section and elevation renderings, as these could take minutes on my 8-core 16GB computer for a complex model imported from Rhino.

Any tips on how to set this up; if at all possible? Or perhaps this is not implemented, but in the planning phase?

For it to make any sense, of course the nodes would need to share a high-speed local network.

Re: Increase speed of Archicad through distributed computing

Laszlo Nagy — Tue, 13 May 2014 15:30:17 GMT

I am not aware of any such distributed computing thing being possible in ArchiCAD.
However, the program will use all the cores and threads your computer has. Since AC17 GS recommends 4-8+ cores because it utilizes them for such things as Section generation. So you machine is probably doing it best, assuming you are on AC17.

Re: Increase speed of Archicad through distributed computing

Bricklyne Clarence — Thu, 15 May 2014 20:27:49 GMT

Additionally, adding more cores (or nodes) will not necessarily increase speed in the manner you imagine.

Your computing power or " raw speed" is always limited by an upper limit as defined by your fastest core or cores.

All that more cores (or nodes, if such a thing were possible in ArchiCAD which it's not) do is increase the efficiency of how the program completes it's task by allowing such "tricks" like background processing and taking advantage (in other programs) of features like 'Virtualization' to perform multiple tasks in parallel rather than serially.

Think of it another way.

If I have a car with a top speed of 200 MPH (or if you prefer, 200 km/h) hauling a load behind it at top speed; having 10 cars with that top speed is not going to enable me to pull that load at 2000MPH.
Your top speed maximum it always going to be the top speed of your fastest car.
Having more cars would only allow you to carry more loads at the same top speed (200MPH) in a more efficient manner (rather than one-by-one as you'd have to do with only one car).

It's a crude analogy but that's basically how computing power works.

You don't get more "speed" by adding more cores.
Only more efficiency and load capacity.

In which case, distributed nodes would likely not be useful from the perspective of getting more speed from a program like ArchiCAD.