Dramatic increase in render speed -- HOW??
Posted: Fri Aug 09, 2019 3:20 pmFormer system: Gigabyte AORUS RTX 2080 Ti Xtreme + NVIDIA Quadro P4000
Current system: 2 x Gigabyte AORUS RTX 2080 Ti Xtreme connected with NVLink
OctaneBench (former system): 269.42 (RTX 2080 Ti only); 360.34 (RTX 2080 Ti + Quadro P4000)
OctaneBench (current system): 560.12
4K test scene (former system): 242 sec (RTX 2080 Ti only); 185 sec (RTX 2080 Ti + Quadro P4000)
OctaneBench (current system): 114 sec
I mean, don't get me wrong... I am delighted by the dramatic increase in render speed. I'm just trying to understand how these numbers are possible. The RTX-only before and after numbers are especially puzzling. Going from one RTX 2080 Ti to two, I was expecting somewhere less than twice the increase in speed (due to various diminishing returns factors); instead, speed increased by 108% on OctaneBench and 112% on my 4K test scene. Is this because the Quadro P4000 was acting as a boat anchor of sorts? Or is this because of NVLink (which of course was not there with the old setup)? If NVLink is the cause of this dramatic jump in speed, why are published OctaneBench scores always somewhat less than a 100% increase for two RTX 2080 Ti cards vs. one?
Again, delighted with the results, which exceeded my high-end expectations. I'm just trying to understand what's going on here.
Current system: 2 x Gigabyte AORUS RTX 2080 Ti Xtreme connected with NVLink
OctaneBench (former system): 269.42 (RTX 2080 Ti only); 360.34 (RTX 2080 Ti + Quadro P4000)
OctaneBench (current system): 560.12
4K test scene (former system): 242 sec (RTX 2080 Ti only); 185 sec (RTX 2080 Ti + Quadro P4000)
OctaneBench (current system): 114 sec
I mean, don't get me wrong... I am delighted by the dramatic increase in render speed. I'm just trying to understand how these numbers are possible. The RTX-only before and after numbers are especially puzzling. Going from one RTX 2080 Ti to two, I was expecting somewhere less than twice the increase in speed (due to various diminishing returns factors); instead, speed increased by 108% on OctaneBench and 112% on my 4K test scene. Is this because the Quadro P4000 was acting as a boat anchor of sorts? Or is this because of NVLink (which of course was not there with the old setup)? If NVLink is the cause of this dramatic jump in speed, why are published OctaneBench scores always somewhat less than a 100% increase for two RTX 2080 Ti cards vs. one?
Again, delighted with the results, which exceeded my high-end expectations. I'm just trying to understand what's going on here.
