I'm thinking to build a new PC. I currently have two Xeon E5-2620 v2 (6 cores each, 2.10 GHz) and I"m considering building a PC with an IntelCore i9-7920X X-Series 2.9 GHz.
Do you think this is a significant upgrade and it will allow me to work faster with Arnold, or is not worth it?
The rest of the components are similar: SSD drive, 32GB ram and I'll go from a current Quadro to a 1080ti (which should be a good upgrade for the next Arnold with GPU support).
The more cores the better.
See this for some general advice on choosing hardware:
https://arnoldsupport.com/2017/03/14/choosing-hardware-for-rendering/
Thank you, Stephen!
What it's difficult for me to figure out is how much I can gain with the new setup: twice as fast? Less? More?
The thing is, the new Intel would have a higher clock speed, but cores would be the same, 12 total so I'm kinda stuck there, not sure if there will be a significant difference.
I have an intel core i9 7900X overclocked to 4.6 GHz - it gets a 2306 cinebench score. It is very, very fast
If you would like to share a simple scene I am happy to use it as a benchmark and test on my computer - it should give you some idea of speed increase. The speed gains of an overclocked i9 vs. a xeon can be significant
I was thinking of doing exactly the same, but purchasing a Titan with the i9 - 7980 XE, and a second 980 TI.. currently have a 2012 Mac Pro, two 6 core Xeon 5690's @ 3.46 ghz, Nvidia GTX 980 Ti, and 64 gigs of ram. Cinebench 1560 😞 A new machine is in the air. I'll try the scene as well for comparison if it is posted ...
Titan's are a waste of money. Only have 1 GB more VRAM than 1080ti and similar performance for 1.5x the cost. or better yet get 2x 1070ti
For the same price get a Ryzen 1950x 16 core.
http://www.guru3d.com/articles-pages/amd-ryzen-threadripper-1950x-review,10.html
Cinebench will not necessarily tell the whole story when you're rendering to Arnold which scales very well with additional cores. Not sure how well physical scales.
It is getting hard to choose.
I'm reading many good opinions about the Threadripper 1950X and it seems that after all, the difference with the i9-7920X is not that much, but it's cheaper, which is good for my pocket 🙂
Anyway, I prepared a very simple scene (maybe too simple?). I'm not sure what's the best approach in this case, so I'm sorry if this is not the best way to run this type of tests.
I didn't render the full res myself yet, but I'm planning to do it soon, then I'll post my results.
I wasn't sure about the Cinebench either, as Shawn pointed out.
P.S. I'm not sure if it's OK to use the ufile service. I didn't know how else to post the file as it exceed the maximun file size allowed here.