Other than performance (rendering speed), is there any "quality" difference if you have an Nvidia or AMD card when using the PhysX engine?

For example, soft cloth/hair... will you get the exact same rendering results with an AMD card as an Nvidia card?

I put a shawl on a character, and in spite of proper collision shapes on the avatar (turned On), the cloth tends to go through the arms a lot.  I tried numerous changes to the relevant cloth and avatar settings, but never got satisfactory results.  The dress behaves okay, but it is a heavier cloth.  The shawl is supposed be light and fluttery, like gauze.

If I should expect the exact same results with my AMD card, then I'll start another thread about the issues I encountered.  But I'd be happy to blame it on the card.

---

iClone 7... Character Creator... Substance Designer/Painter... Blender... Audacity...
Desktop (homebuilt) - Windows 10, Ryzen 9 3900x CPU, GTX 1080 GPU (8GB), 32GB RAM, Asus X570 Pro motherboard, 2TB SSD, terabytes of disk space, dual  monitors.
Laptop - Windows 10, MSI GS63VR STEALTH-252, 16GB RAM, GTX 1060 (6GB), 256GB SSD and 1TB HDD

Collision shapes on avatars are extremely limited.  At least from within iClone.
Just for fun, I tried attaching an additional object to the avatar to wee what would happen, but that's off-topic.
FYI... Reallusion limits it to 16 active collision shapes on an avatar.  That's an iClone restriction, (probably for performance reasons, I suspect) not an Nvidia PhysX restriction.


BACK TO PHYSX...

Anyway, the thought behind my question is... I was wondering if *maybe* PhysX was slightly crippled when running without Nvidia hardware.

When you are on non-ideal hardware, code can branch into a "low-quality" mode, for example.  I've seen stuff like that before, and was curious if my low-powered AMD card was compounding my issue (not the sole source, but adding to the challenge).  Thus the question.

---

iClone 7... Character Creator... Substance Designer/Painter... Blender... Audacity...
Desktop (homebuilt) - Windows 10, Ryzen 9 3900x CPU, GTX 1080 GPU (8GB), 32GB RAM, Asus X570 Pro motherboard, 2TB SSD, terabytes of disk space, dual  monitors.
Laptop - Windows 10, MSI GS63VR STEALTH-252, 16GB RAM, GTX 1060 (6GB), 256GB SSD and 1TB HDD

Exactly!!!  You get it now.
So.. back to the original question...

Other than "speed," might I actually get better rendering results by switching to an Nvidia card?

---

iClone 7... Character Creator... Substance Designer/Painter... Blender... Audacity...
Desktop (homebuilt) - Windows 10, Ryzen 9 3900x CPU, GTX 1080 GPU (8GB), 32GB RAM, Asus X570 Pro motherboard, 2TB SSD, terabytes of disk space, dual  monitors.
Laptop - Windows 10, MSI GS63VR STEALTH-252, 16GB RAM, GTX 1060 (6GB), 256GB SSD and 1TB HDD

Sober answer: Physx is proprietary tech and NVIDIA has gone to great lengths to ensure it doesn't work directly off AMD cards (AMD relies on CPU workaround) so you lose benefits of GPU acceleration and that definitely impacts on the quality of results. 

I've tried turning off PhysX on my card. Then PhysX runs on your CPU. Still gives the same results.

There are a couple additional PhysX cloth features I would like to see implemented. One is the aforementioned "backstop", and the other is the transfer of physics from one layer to another without having to run physics on both layers. What that means: Think of a thick overcoat that you only compute physics on the inside layer, and the outside layer just follows it with an offset. It seems that would be better for multi-layer clothing than using a physics collision on each layer.

Without an Nvidia card, PhysX will still run, but on you CPU instead of the graphics card, so no acceleration of course.  Knew that.

Different "quality" of results?  You say "Yes," but see Rampa's test results below...




That was a clever way to perform the test.
Same results on the CPU as on the GPU = Yes.
Thank you.  :)

Later I'll play with the cloth/collision settings some more.  At least now I know my lack of Nvidia is not aggravating the situation..

---

iClone 7... Character Creator... Substance Designer/Painter... Blender... Audacity...
Desktop (homebuilt) - Windows 10, Ryzen 9 3900x CPU, GTX 1080 GPU (8GB), 32GB RAM, Asus X570 Pro motherboard, 2TB SSD, terabytes of disk space, dual  monitors.
Laptop - Windows 10, MSI GS63VR STEALTH-252, 16GB RAM, GTX 1060 (6GB), 256GB SSD and 1TB HDD

Ok. :)

But then he wouldn't be using iClone.
And we'd be discussing this in a different forum.

---

iClone 7... Character Creator... Substance Designer/Painter... Blender... Audacity...
Desktop (homebuilt) - Windows 10, Ryzen 9 3900x CPU, GTX 1080 GPU (8GB), 32GB RAM, Asus X570 Pro motherboard, 2TB SSD, terabytes of disk space, dual  monitors.
Laptop - Windows 10, MSI GS63VR STEALTH-252, 16GB RAM, GTX 1060 (6GB), 256GB SSD and 1TB HDD

It really all depends on where iClone sits in your foodchain and what else you have in your pipeline.

If iClone is low in hierarchy and your multimedia software (3D, NLE, GPGPU etc) likes openCL, by all means go with AMD.

However if you want to future-proof your hardware for iClone, as well as having software that exploits CUDA you need to go with NVIDIA.

If you want to end up with two expensive but incompatible Graphics cards down the line.. base your decision on Rampa's "test".

If you have too much money to burn... share some with me. I take DA points.:D

Its really not that complicated.

PhysX is on by default with the Nvidia drivers. It can be turned off for making such a test.

What Viking wanted to know was whether or not the end result would look the same if PhysX was computed on his CPU. It will, but it might be a little longer before he can view those results.

I would encourage any iClone user looking to upgrade their GPU to go with Nvidia, as that is more in line with iClones current requirements. I'm guessing even more so towards the future.