A recent article by Real World Tech has revealed why Nvidia’s Maxwell and Pascal GPUs really are so special.
It has long been understood that the new partitioning wasted much less power on interconnects. It is also known that Maxwell implemented a significantly altered SM structure so that it was easier to reach peak utilization and that they had increased the L2 cache size and optimised transistors.
But Nvidia has always held back information surrounding the technology and it’s “secret sauce” – until now:
By playing around with some DirectX code specifically designed to look at triangle rasterization, he has come up with some solid evidence that NVIDIA’s handling of tringles has significantly changed since Kepler, and that their current method of triangle handling is consistent with a tile based renderer.
“The significance of tiling is that by splitting a scene up into tiles, tiles can be rasterized piece by piece by the GPU almost entirely on die, as opposed to the more memory (and power) intensive process of rasterizing the entire frame at once via immediate mode rendering.”
This was a technology first developed for Nvida’s mobile GPUs in order to see a performance boost from the weaker cards but comes with its own drawbacks, including compatibility issues.
It seems that Nvidia has found away around these issues and is now confident in having the technology in its graphics cards.
Nvidia is currently the only GPU manufacturer using tile-based rasterization.