We don't expect big performance gains from switching OT to Vulkan (not CPU bound), and even if we very much liked to switch to a clean modern API, it's hard to justify the switch that would take us several months. For us it's mainly the question if Vulkan can really bring in better quality drivers, so we'll follow it closely and evaluate it.
I will be curious as to how this affects their ongoing performance testing.
Our perf tests aren't struggling on the CPU side at all, it all seems to be on the GPU side. Similarly whole OT is rather GPU bound, and the agreement is that Vulkan won't help much here. However, since it's a simpler and more explicit model, the drivers *might* be better optimized. On the other hand, explicit model is more complex, and if coders start to emit bad code, vendors will
likely try to optimize it in their drivers to gain over their competition ... that would be bad.
