Blender multi core support

So im looking to upgrade my computer for the purpose of running blender and decreasing render times.

I am looking at an amd 3.2 quad black edition. I know rendering supports up to 8 cores but do other feature like modifiers and such use all 4 cores?

Thanks for the help guys.

Simulations (Fluids, smoke), sculpting, rendering, and compositing all support multi-threading/ multiple cores, I believe modelling, texturing, animation do not support multi-threading, but usually they aren’t very processor intensive anyway.

There will probably be some other things, but I’ve forgotten.

Also the internal supports 64 threads, which means it’s capable of supporting 64 cores.

EDIT: Never write a thread while watching Black Books, it’ll end up with embarrassing grammatical errors :frowning:

EDIT2: The AMD X4 is a good choice of processor, although if your budget can stretch a little further I’d highly recommend the AMD X6 processors.

haha so pretty much anything that need multi core support can use multiple cores. awesome. thats good news. thanks bro!

Not my experience, at least using 2.49. Only raytracing of the internal render seems to be multi-threaded. How often I have looked my quad core only to see 25% usage. Baking and compositing is only single threaded.

Atom, 2.5 is much better in that regard.

‘I’m a prostitute robot from the future!’

.b

also 2.5 doesn’t seem to support multi core in baking and physics simulations (as also some part of rendering such as occlusion processing and other stuff). apart from pure rendering, my i5-750 is always stucked to 25%, in some case 40%. this is quite frustrating :frowning:

if you get one of the builds from graphicall.org that has OPENMP enabled, many of the features Atom mentioned seem to magically become multithreaded.
pretty good trick if you ask me

I had heard some time back that the cloth sims were much faster in 2.5x, I thought that was because of multithreading. Is that correct?

i heard of it many times, but i could never find any build working like this… i have always been a little bit dumb for graphicAll build researches… :smiley:

EDIT: a big doubt is coming up to me… should i install something concerning OPENMP in order to make openMP blender build to work properly???

All recent official Blender releases for Windows have OpenMP enabled. I also ensure that for upcoming releases this will be turned on.

/Nathan

that’s odd… i tried the last build two days ago from graphicall and it was written that openMP was enabled, but in cloth simulation the cpu was still far from 100% usage. as i wrote before, should i install some libraries on my computer in order to use them properly in blender?

thanks

With OpenMP I think you won’t see 100% CPU usage for all cores combined. But it should be larger than the percentage for one core.

/Nathan

ok… i thought that i could take fully advantage of the multicore… so for now the multicore is still limited to rendering.

:slight_smile: Not that bad, allready… 2, 4, 8 times faster rendering is yet good to take… :yes: But if bakery and particles will follow, it’ll be quite allright for me too! :yes: :yes:

Just for reference, rendering can be done in parallel, so each thread or core can take a piece of the image and process it then put it together (100% across the board). As far as I know, when doing physics calculations one thing relies on another (this happens to make this happen to make that happen and so on) so has to be made or calculated serially therefore it may be the reason that each core breaks up 100% of the bandwidth = 25% workload each. And then there is the way that information is piped to the processor, I’ve done many things with many different pieces of software and it is always the case where one multithreaded process pegs the CPU and another uses 1/2 or a 1/4. Replies that I have been given are above, and in many cases deals with the coding itself or whether or not the process can be serial or parallel in nature.

short question:

In know from blender 2.4x, that you had to choose somerwhere an option to support multicoring.
Is it automatically in 2.5 or am i just not finding the Button?

On topic; the AMD x6 processors are fine for rendering. I recently assembled a new rig and I seriously considered these. However, according to performance charts available around online, it seems the x6 processors aren’t all that great for anything other than applications that utilize all 6 cores at once. Its because there are evidently more cores, each single core does not have a great clockspeed. But then again, its all based on a few seconds and margins; nevertheless, check them out.

Unless you’re doing more of rendering and stuff, I suggest you go for a Phenom x4; or if you plan to change your motherboard too, consider Intel i7 950 and if you want to stay budget friendly, go for an i7 760. Its the same price of an x6.

Actually would be even nicer to have GPU acceletarion, wich seems to be the fastest method, just check the GPU renders out there (smallGPUlux, octane render, etc…) so imagine to use this amount of power in other areas than rendering.