The translation is more like a reimplementation, and sometimes that reimplementation is faster than native. But it’s also because the Linux kernel is faster in some areas, and typically more memory efficient too.
And it’s partly also the quality of GPU drivers, especially in the case of AMD (although they have been getting better on the Windows side in recent years).
Maybe it’s a case of less bloat in Linux over Windows?
The translation is more like a reimplementation, and sometimes that reimplementation is faster than native. But it’s also because the Linux kernel is faster in some areas, and typically more memory efficient too.
And it’s partly also the quality of GPU drivers, especially in the case of AMD (although they have been getting better on the Windows side in recent years).