How is the performance of 1080 ti ? How is it the effiency of using 2x vs 3x vs 4x 1080 ti? No information so far in the blenchmark for multiple 1080 ti .
Also, anyone using ryzen7 1800x /1700x ? Any problem ? How is it compared to Intel ? Any suggestion?
we have thrown 2 1080s into one machine, and have noticed the following:
rendering 1 frame on 1 gpu, it is 100% speed or say 100 frames of an animation in 100 minutes
rendering 1 frame on 2 gpus, it is 146% speed, 100 frames in 68.49 minutes
rendering 2 frames at the same time, allocated to 1 gpu each, it is 189% speed, so 100 frames in 52.91 minutes
The reason for it is if there are complex tiles on a frame, it is making one gpu render intensively for it, and one gpu just sits idle. waiting for the next frame.
the 1080ti is approximately 30% faster then the 1080. blenchmark is not a reliable source of benchmarks either.
We run the 1700x on our workstations here. It works fine. no problems. fast. reliable.
Cool… glad to hear it. Btw , do you ever experienced any problem with ryzen using other 3d app or support app (other than blender) ? my biggest concern is i’m afraid there’s hidden bug of the ryzen. Btw did you OC your 1700x or your GPU ? or running at default mode. What motherboard you’re using ? i found so far the crosshair VI hero (x370) as the best choice for ryzen 7 , but maybe you have other opinion ? if you don’t mind can you list your main specs of your workstation for my reference/guide ? i want to build one, but seems that too many consideration makes me nervous to pick.
Also are these multiple GPU can benefit on rendering only ? can it help on my viewport etc ? like in modo , the viewport can only benefit from single gpu. The multi gpu will be used for v-ray/octane render.
I am literally in the same boat as you. Planning my first PC build, and wondering if Ryzen plays well with Blender. Some people say that single thread speed is also important for some unknown to me aspects of blender. After doing lots of research, I’m going to pick Ryzen. Best of luck!
I’m also very interested in the benefits of dual GPUs for Blender. Do they erase the whole pc freezing up during rendering problema I’ve read so much about?
Single thread performance is useful for all other tasks not related to Cycles render, such as simulations, modifiers, edit mode, sculpt mode. Even for those tasks that are supposed to be multi thread, like this particles benchmarck (you see results below in users messages):
I get around 2:40 min on my old i7-3770 baking that simulation.
You can improve viewport rendered view by running one GPU as a render device and one as a display like I do - makes things much smoother and less prone to locking up. If you’ve got two identical cards you can just set them both to render when you don’t need the pc to be completely usable.
Other than that, no, not really. A lot of viewport is CPU driven - which is consequently one reason for high single-threaded performance.
The other reasons include operations that aren’t multi-threaded and thus benefit from the uber-high clock speeds or Intel’s 7700k’s and newer etc.