With a few Intel Arc graphics cards launched, some interesting news related to RT performance has been confirmed recently.
The content was discovered while doing an optimization patch by Intel’s Linux graphics driver engineer Lionel Landwerlin, and it was reported that the performance improved by nearly 100 times through a simple coding improvement.
Lionel Landwerlin said that, in the existing Vulkan RT, memory allocation was allocated to system memory instead of GPU memory.
In other words, as the bandwidth speed of the system memory is lower than that of the GPU memory, the bottleneck element was allocated to the GPU memory and resolved, resulting in a high performance improvement of nearly 100 times.
