Aug. 17, 2019
One question remains: what will you do with all this time?
The team (Valeria Gerontopoulos, Alice Girard Segaud, and Cedric Marchessoux) decided to get into the details of Substance Painter’s bakers improvement, and see just how much time artists would gain with the Substance Painter Summer 2019 release.
How We Organized The Tests
Scenario 1:
Baking a low poly mesh and a high poly mesh
The team used a low poly mesh (with 10 734 polygons & 21 342 triangles) and a high poly mesh (with 1 406 657 polygons & 2 784 010 triangles).
With this dataset, we decided to compare the baking time between SP 2019.1 and SP 2019.2 for 3 maps: Normal, Ambient Occlusion, and Thickness.
As variables, we chose the output resolution range with 64 secondary rays.
Scenario 2:
Baking one mesh
The team used a high poly mesh with 3 819 560 polygons. This dataset was the basis to compare the baking time between SP 2019.1 and SP 2019.2 for 2 maps: Ambient Occlusion, and Thickness.
As variables, we settled on three parameters: the output resolution range, the secondary rays range, and Anti-Aliasing.
The Results:
CPU Improvement
This test compares results for the 1st scenario on a GPU AMD RX Vega, and analyses the difference in processing time between CPU before the bakers improvement and after. The results vary between 11% and 74% faster. On average, the impact is significant.
After testing with basic setup, we decided to see what would be the improvement with RayTracing. And the difference became staggering.
OptiX improvement
We first tested the 1st scenario on a NVIDIA GTX 1650, then the 2nd scenario on a NVIDIA RTX 2080. In both cases, we compared CPU before the latest Substance Painter release and GPU with the new bakers acceleration.
On the GTX 1650, baking time goes from 9% to 65% faster – which represents a significant gain of time – but that is even more noticeable when equipped with a RTX 2080, as the baking time is 95% to 98% faster!