GrassHopper CPU Optimization

Hello,

from what I have seen in my computer in Windows Resource Manager, GH only uses up one CPU core. I have a Intel Xeon X5450 @ 3.00GHz which nowadays is not much. But even considering a newer chip like the Intel Core i7-3770 @ 3.40GHz its not possible for GH to use every core. So it would make more sense to use an overclocked CPU like the Intel Core i7-2600K @ 3.40GHz which can easily be overclocked @ 4.0GHz-4.5GHz.

Or is it possibe to make GH use more than one core?

Thanks!

Replies to This Discussion

Permalink Reply by Hannes Löschke on November 6, 2012 at 3:03am

For now GH is single threaded. As far as I know this will not change too soon since Rhino5 still in not entirely guaranteed to be thread safe in every aspect.

Until all of Rhino5 is thread safe, there will probably be no multithreaded GH.

Permalink Reply by Jesus Galvez on November 6, 2012 at 3:16am

So I guess its better to use an overclockable CPU.

Well at least Rhino 5 won't run out of memory and suddently exit the program.

Permalink Reply by Sander Mulders on November 6, 2012 at 3:14am

For more info on MultiThreading:

http://www.grasshopper3d.com/forum/topics/is-there-a-plan-to-suppor...

For me an overclocked I7-2600K @4.4Ghz made the most difference + using Rhino 5 64 bit to use a lot of RAM.

Permalink Reply by Jesus Galvez on November 6, 2012 at 3:19am

Oh thanks for the reference! Yeah, that's the system I've been looking on building.

Permalink Reply by Thibault Schwartz on November 6, 2012 at 4:33am

You can also use multi-instances definitions and play with UDP messaging to communicate data between them.

Running this kind of things on a main computer + a slave, both with i5 2500K @ 4.8Ghz / 16Gb, you can do whatever you want. Better than the 2600K (also have one at the office) because the cores only have one thread each on the 2500K (and of course it's cheaper).

Permalink Reply by Jesus Galvez on November 6, 2012 at 5:05am

Do you mean opening the GH file from two different computers? How is the workload then shared? I guess that would involve the master + slave configuration you speak of.

I do have another system which, though not the same, runs at more or less the same speed. But out of curiosity, because I'm not really going to setup such a system, would both computers need to have the same CPU, memory...?

Permalink Reply by Thibault Schwartz on November 6, 2012 at 5:26am

Let's say that with the system I am describig, you can open as much GH files and rhino instances as you have threads, of course you can add more than 8 but I don't think it's necessary, we are already talking about huge amount of data with those kind of specs) and dispatch the work between them, the same way you would do with real multi-tasking (one work to do for each instance/thread). So It can also work with only one computer with multiple threads available. If fact you could aso run a render farm made of old computers with this logic if you want, it's working the same.

You just need to properly dispatch the things to compute on your different threads (and definitions) so that if you work with another (slower) computer, the amount of data to manage is acceptable for the processor & RAM. You don't need anything special to do that.

This is just some kind of cheap distributed solving system that you can use when you have a few computers or threads that you can use and that you have for exemple to have one thread running Kangaroo, another one running Galapagos, maybe another one running Firefly, etc. If you want to merge the results then you will have to split the definition so that each step can evolute depending on the results of the other ones, so it's quite helpful.

RSS