Is it worth waiting to see how intel’s arc gpu’s do in machine learning compared to Nvidia already supported gpu’s? Or is the amount of library support for Nvidia outweighs the upcoming support for intel arc?

16gb of vram in the A770 vs 12gb of vram in the 3060

I did read that tensowflow and PyTorch are working to use intel’s arc XMX, but there is no date of when that will be done