Their comparison charts showed the performance of mobile GPUs, not the desktop ones. So, I wouldn’t call this “practical”. Most likely depends on what kind of models you are building and what software you use and how optimized it is for M1.