Im using that chip after upgrading from a phenom quad core. I reduced a test scene I had from 24 to 10 minutes to render.
Would you mind doing a test for me? That same render that took 10m on this system—what does it take if you turn off multiple cpus? (menu: File/Preferences, ‘General’, uncheck ‘Use Multiprocessors’).
I just tested on a Xeon 6-core and saw improvement of about x6—to be expected because it has six cores even though the operating system reports 12 due to hyperthreading being enabled.
The thing with the new AMD architecture is that while it is heavier than hyperthreading each ‘core’ is not complete: they are grouped in pairs with two cores sharing an FPU. Performance improvement will depend on how FPU dependent the Carrara renderer is, and is something I’m interested in knowing. What is in the scene and render options may affect that as well (e.g., even if the renderer is overall integer bound it might use the FPU for caustics—I have no idea).
So if you don’t mind I’d appreciate timings for same-setting renders of the same scene with and without multiple cores.
For the record, these were my results (I’m rerunning the single threaded as until the last render I had a VM allocated two cores running which is very likely why it exceeded a six-fold improvement)
single core: 13:43 (tile size 128) x1
12 core (w/ hyperthreading): 2:42 (tile size 128) x5.08
12 core (w/ hyperthreading): 2:25 (tile size 64) x5.68
12 core (w/ hyperthreading): 2:06 (tile size 64) x6.53