Hi. I’ve access to two workstations (Win10) with 2x RTX 3090 in each.
With CR 0.4.2 installed in Blender 2.93, and everything synced, I tried rendering an animation using only the GPU’s on both machines.
Problem is that the render with CR (and potentially 4 GPU’s) is way slower than the internal render with only 2 GPU’s..
I’ve tried different tile sizes, which reduces the render time, but never faster than the internal render. I would expect something like half the rendertime, at least for this to make sense.
Am I doing something wrong here?
Plz tell me if you need anymore info. Best, R
Hi Rasmus, are you aware of the load balancer menu in the Crowdrender panel? I would suggest setting load balancing to manual and then set the amount of the screen rendered by each machine to be 0.5 each.
The logic of this is that if you have two machines of equal power, then they should render half or 0.5 of the total area each. I suspect what may have happened here is that you have used automatic mode.
Auto mode has a few quirks, one of which is that if you render on just one of the computers initially, then connect a second computer and render with the second computer enabled. The load balancer, for the first frame, will not give much of the screen to the second machine. This results in a slower render than normal.
Using manual mode you can get around this.
One other potential cause is that by default, the CPU and GPU will render. Since CPUs grab a lot of tiles, if you set large tile sizes, then what can happen is that the CPU will grab a lot of tiles and effectively you'll be CPU rendering MOST of the image.
So, try going into the settings for each of your nodes, and making sure that you're not using the CPU and the GPU at the same time without realising it. Chances are, you will be.
Try these tips out and please write back if they help?