Will a Pentium G4560 bottleneck 3 GPUs when rendering with Cycles or Luxrender?

I plan to build a rendering machine to use mostly the GPUs for rendering and I’m looking for the best bang for the buck. I’m almost convinced to use the GTX 1080 because it has pretty large ammount of Cuda cores and it just went down in price due to the release of the GTX 1080 Ti. My main concern here is if that low cost CPU will bottleneck my 3 video cards for rendering. If you need more info to know the answer just ask.

Hi, CPU does not influence GPU render performance.
If scene build is ready and send to GPU it render as fast as any other CPU like i7.
BUt big scenes need a lot of CPU power to build so BVH building need may more time than render. :wink:

Cheers, mib
EDIT: Looking at https://ark.intel.com/products/97143/Intel-Pentium-Processor-G4560-3M-Cache-3_50-GHz
your CPU is not that bad.

I don’t actually have the CPU yet. I’m just planning to build a second PC to use it only for rendering. The cost of that CPU saves some money that can be spent in GPUs instead than CPUs. I can save about $250 USD by buying that CPU vs i7-7700K without decreasing rendering times. Will that be true also for socket 1150? I have seen some nice prices on used motherboards and DDR3 RAM making the building cost even lower.

Regarding BVH building, I plan to put my second PC as slave with my main PC which has a i7-4790k, GTX 970 and GTX 1070 as main computer. Does that mean that the scene making will rely on my main PC making the rendering easier for the slave PCs?

It depends on the project. The CPU still must deliver instructions. If the system is intended for gaming as well, I recomend a seperate machine for these GPUs. It will certinanly help to get a CPU that exceeds 4 cores for even ONE GTX 1080.

Skimping on the CPU to that level and going all-in on GPU’s may speed up rendering and a few other things (the viewport will join with a dramatic speed up in 2.8), but it will dramatically slow down almost everything else compared to an i5, i7, or Ryzen processor (such as modeling tools, modifiers, UVmapping, ect…).

The key here is balance so you can have reasonable performance for every step in the process of 3D.

The Pentium will be installed on a second PC used only for rendering. I already have my main PC which has a i7-4790K, 16GB of RAM, 750 watt PSU, Asus Z97-A/USB 3.1 motherboard, one GTX 1070 and one GTX 970. The modeling and scene building will be made on that PC and rendering will be made via network. The second PC and any PC I decide to add after that will be slaves. I didn’t mentioned it, but I’ll be adding the GTX 1080s first to my main PC and then build the second PC. I just wanted to ask if the CPU won’t bottleneck the 3 cards since it’ll be used for rendering and not gaming or modeling or anything else.

The short answer is it will to a fairy small degree. The effects will be noticable in the long run.

I would personally suggest the i3 7100, its 3.9GHz single score speed will help with any single threaded components, and the hyper threading will give it some advantage over the Pentium. It’s still a decent bit cheaper than the i5 (which is slower in single threaded parts).

If rendering is your main objective, than I think the i3 would be a good fit, and the higher clock speed better than a slow i5. A faster i5/i7 would be better, but your paying more than twice the i3.

Whilst 3 cards will make the render go faster, with animations, it is MUCH more efficient to split the render up over 3 blenders (1gpu per blender) then to get all 3 gpus working on the same frame. Why? because say there is one particularly hard tile to render (much more complex then all of the others), that could mean that 1 gpu is cranking away on that, whilst the other two are sitting idle. In our real world scenarios, if 1 gpu is 100% speed, 2 gpus, same frame is 145% faster, whilst 2 gpus different frames gave us a speed up of 189%. This is on first gen i7 quad core (core i7 920).

If you are sinking this much into a rig, you dont want it to be bottle necked. I would highly suggest switching to a quad core (ryzen 3 sticks out to me) type scenario.

> If rendering is your main objective, than I think the i3 would be a good fit, and the higher clock speed better than a slow i5. A faster i5/i7 would be better, but your paying more than twice the i3.

A lot of the CPU bound processes with rendering (BVH building / image loading / compositing) are multithreaded which would take advantage of the more cores. Especially if there are 3 gpus with 3 separate blenders running you dont want to skimp out on this!

EDIT: Because people are asking about how to do this.

the easiest way is to save your file with output directory set (and rendering to frames) and that the interface gets locked during render (makes it go faster – https://gooseberry.blender.org/wp-content/uploads/2015/02/locked_interface.png ).

open it up in a secondblender, in each blender in the user prefs select one GPU (so for blender 1, select only gpu 1, blender 2 select only gpu.

Then hit render animation in both blenders. Alternatively instead of hitting render animation, you can render via python console (slightly faster) by typing in the command bpy.ops.render.render(animation=True)

Can you explain in more detail how can I do that? Perhaps also point me to a tutorial?

I just browsed eBay for used CPU/Motherboard/RAM combos and found some good offers for i7-4770K/90K, motherboard and 16GB of RAM. I’m sure that when the time comes to build my second PC, I’ll be able to find a good offer there. For now I’ll concentrate in filling my main PC with GPUs. Thanks everybody for your answers.