Update: Titan V Results now added
When I first started researching the switch from CPU to GPU rendering, I couldn’t find a lick of tests and when I did it was only based on 1 or two cards. Some in the forum claim to have been posting tests for years but #crickets (so tired of everyone with an opinion but don’t want to offer help or information)
All in all I hope this helps someone like it has helped me. Each render was ran up to 5 times to make sure there were no bottlenecks or loading issues, often render time ranged varied by 1 second so I just took the best time.
The combinations for me was to determine which cards will go in which node. I couldn’t just fill up a 4GPU board and leave another node half performing. trying different combinations allowed me some insight on how the cards worked together and their scaling. This is also important for those who buy one card at a time and wondering how your scaling would look, should you shoot for another card. Initially the plan was to have 4x GPU per node but was scaled down to 3x due to the size of the Titan XP collectors edition body being too big by 5+mm
The render specs are irrelevant since I was not concerned with super super fast renders. I can’t see the margin if everything is rendering under a minute. So I made sure there was enough buffer. This helps me determine the ratio for the final render which will be optimized.
and for those who think i’m not legit because I don’t have an active license. I plan to buy more than one, so I need to make sure I have my nodes worked out so I can determine how many I need. That was by far the least of my concern but apparently its super important to some here.
Rendered 8k images (not 1080 like most) because this is a real project, 360/VR render for motion playback in VR headset – 4k each eye.
Why 2 of everything? well #nvidia’s cruel purchase joke. After I purchased the 2x 1080ti’s and 2x 1070 ti’s they sold out a week later and based on inflation, the Titan XP and Titan V have been made more attractive. I personally don’t see the 1080ti prices going down unless Nvidia releases a new card that forces their rate down and from what I’m told by a friend, April at earliest, but don’t quote me.
All in all this has been pretty fun.. expensive but fun. I personally don’t care what anyone else results look like because my results is what i’ll have to work with during my project(s). However if I would have believed the speculation talk, I might not have chosen Redshift. I chose to test Redshift because the team encouraged me to run my own tests. (Thanks Panos) – regardless of any one forum ego telling you what you should buy or try, just run your own tests.
I went with Redshift over Nvidia Iray because I could not get in contact with anyone who would answer basic setup questions and the workflow was too new for the timeline but after this major project is done (April) I will go back and do a proper speed test between Iray and Redshift.
www.redshift3D.com is the company
(See attached Excel Image)
Be safe and happy rendering!
I’m not hard to find if you have any additional questions.
- Jamiel Boling