Different results on render farm
#1
At least Physx simulations simulate different on our render farm.
I didn't cached anything, tf is so fast, thought I wouldn't need it.
There are differences between Intel Xeon E5-2630 v3 2,4 GHz and Xeon E5-2687W 3,1 GHz both dual socket 16 cores.
Both Windows 10 prof and cheap Geforce GT 710 graphics, same nvidia 388 driver.

Will test some "simple" particle interactions without PhysX.
By the way tyflow is the first simulation software I've used which use all my 32 cores to 100% for particle physics at least.
Amazing work, I bow in front of you.

Perhaps an option for saving the automatic cache with the 3sa max file or a companion file (f.i. scenename.ty) would be good ideas.
  Reply
#2
Thanks, if you check the roadmap (http://roadmap.tyflow.com) you can see that PhysX determinism problems across machines is on the bug list.

For now a fix is to cache our your flows to tyCache, and render the tyCache instead Smile
  Reply
#3
Hi vklein, 
I haven't been able to keep tyFlow from locking up my render farm. Max will just freeze. Did you have any issues with that, or has tyFlow worked on your farm from the start? It's good to see network rendering has worked for other people, even with the PhysX problems!
  Reply
#4
Hey andy,

Could you provide a few more details about what kind of setup you're using on the farm, what rendernode manager (backburner, deadline?), which version of tyFlow you're using, etc?
  Reply
#5
This is awkward...

I downloaded the latest tyFlow build to be sure I had all the loopholes for error closed before taking this further with you, and now it appears it's working! I didn't try this earlier because after viewing the changelogs, I was sure none of the fixes would solve my problem, but I'm wrong. Thank you, Tyson!
  Reply
#6
Glad it's fixed! If it happens again, let me know.
  Reply


Forum Jump: