So I have a crazy idea about building a render rig from an old mining setup. Here is the premise:

I use a mining motherboard that has 19 pci 1x slots, I know that the loading of textures will take time because it's 1x however, but I was wondering if iClone will see and utilize all the GPU cards to render the image/animation?

Does anyone have any thoughts?  I currently have a rig with 4 980ti gtx's and it works great so I thought man if I was able to get 19 going!

FYI, I render with iRay plugin.

Iclone will only detect/use only one, iray on the other hand can use all of them

i missed a key piece of info you provided, the slots are pcie x1. I would ask in support or maybe some forum mod can chime in on this as i don't know if there would be any issues regarding the pcie x1 slot.

@GrumpyOtis - Raxel is correct, is first answer is as complicated as it gets ultimately.

iClone's realtime renderer will only be using your 1 main MOBO Display GPU.
Iray, however, can use all of the GPUs your rig has.  It will not matter if they are PCIE 1x, all this will mean is that the load time to feed data to that GPU, per frame, for that card will be 1x, which would not be as fast as a 16x, 8x, or 4x slot.
So, you will be bottle necking your load time into Iray for all cards at 1x, per frame with use of any 1x slot.  The trade off then is that you would potentially have many cards rendering once the data is loaded into the cards, so it can be faster.

I know this first hand as I have a rig now with 10 cards, 8 of which are connected PCIE 1x.  Iray, Indigo and Octane all load data to GPU slower with 1x attachments, but they all render the frame WAY faster with more cards. 
So, it usually is better to have as many cards as possible with a CUDA raytrace renderer like Iray. 

There is ONLY 1 exception to this paradigm - Unreal RTX raytrace, which uses 1 card, like iClone, for its raytrace processing.  So, in that case you want the most powerful card you can get your hands on, and have that be your MOBO Gpu.   (ie 2080Ti RTX, Titan RTX)

That’s interesting! So I guess my other question is, does it take the same amount of time to load textures to the card for each frame or is it a long load time initially and then the renders follow pretty quickly after the textures are loaded?

If you use animated objects list with refresh only, all the static meshes and textures should stay in memory and only animated objects are loaded for each frame