Loading Scene into Octane Tweet

[rappet]

Hi Paul,

When the plugin is Loading Scene into Octane it can take a while when having to Process 10K elements.
Only the CPU is the bottleneck on this right?
I have been monitoring the CPU workload while processing and I noticed that multi threads are used, but not to the full.

What can I do (now on current PC or in future PC) to reduce this waiting?

Is this a general thing of the Standalone or is it also something you might be able to optimize in the plugin?

greetz,

[whersmy]

Standalone has the same part, RAM as well fills up

You could OC the cpu a bit, don`t think it would improve much, but in general I think these processes are mainly single-threaded?

[Lewis]

Its same in plugin(s) , most of time it's either single-threaded or just using few % of several cores but on my machine i've not seen anything more than total 10% CPU usage (Dual Xeons - 80 Threads). I hope DEVs will try to optimize that 'coz CPUs are getting more and more cores and less GHz so it's important to optimize all to multi-threading as much as possible.

[rappet]

Hey Lewis,

I also monitored the CPU core behaviour while generating heavy floorplans.
I have noticed that one core has high workload, and other cores have a small workload.

This might indicate that having many more cores the gain is relatively small coz' the extra cores will add some small amounts.
this examples in numbers (adding % workload);
my current i7 cpu 12 threads): 1x 70%+11x10%=180%
dual Xeon (40 threads); 1x 70%+39x10%=480%.
That would indicate that a dual Xeon could be 2,5 faster for generating (heavy) floorplans?
Do you think this assumption is correct?

cheers,

[Lewis]

Probably not 'coz you didn't include GHZ core speed, Architecture of each CPU/core etc. etc... It' s a bit more tricky than that .

[rappet]

Probably not 'coz you didn't include GHZ core speed, Architecture of each CPU/core etc. etc... It' s a bit more tricky than that .

Oh yes, I forgot about that in this calculation... xeon=2.2 / i7=3.0...
then it would indicate that a dual Xeon could be 1,6 (instead of 2,5) faster for generating (heavy) floorplans?
Then cost/value approach is not that attractive comparing i7 versus dual xeon.

cheers

[face_off]

Hi Jeroen - the plugin will extract all the geometry from ArchiCAD in a single thread, load that info into Octane in a single thread, but then Voxelize the data using multiple threads.

Paul

[rappet]

So it is not so easy to determine what would be better CPU config for ArchiCAD-Octane user... While i7 cores have higer frequency for that one thread, but xeon can have a dual cpu and many more threads.

It would be nice to have sort of benchmark with loading (heavy) scene, 2d and 3d plan generating and loading scene into plugin, but that would be a hard one, right?
But we can make/use a test scene and compare with eachother as users with digferent cpu's dual.

Cheers,

[rappet]

Hi Jeroen - the plugin will extract all the geometry from ArchiCAD in a single thread, load that info into Octane in a single thread, but then Voxelize the data using multiple threads.

Paul

Hi Paul,

I do not know it is so, but I got the impression that the renderslave (with less cpu) needs less time to load the scene.
Can that be (because of master has done the hard job)?
How does the master and slave work regarding loading scenes?

Greetz, Jeroen

[face_off]

I do not know it is so, but I got the impression that the renderslave (with less cpu) needs less time to load the scene.

The slave is being sent the Octane NodeGraph - so yes, a lot of the hard work has already been done.

Paul