Although they have similarities, the internal implementation of Auto Mask differs from Magnetic Mask. Auto Mask is incredibly fast at defining the mask, but incurs more runtime overhead. If you select that clip and do CTRL+R to render it to cache, it should then play smoothly. But rendering the clip may take some time.
By contrast, Magnetic Mask's "Analyze" feature can take time, but when it finishes, it tends to play more smoothly without needing to render to cache with CTRL+R.
This difference is fully expected. But if you have a case where Auto Mask has sluggish playback even after rendering to cache (IOW the render dots vanish above that clip), then please report more details.
If so, is it Auto Mask itself causing it, or is it some combination of effects plus Auto Mask? There are some known cases (not limited to Auto Mask) where certain combinations of effects may not get the expected performance benefit of rendering to cache. You can help us see if that case applies by selecting that clip, Edit > Remove Attributes, then remove all attributes except for one using Auto Mask.
Since Auto Mask is often applied as an option on other color effects such as Color Wheels, as a troubleshooting procedure, you may have to temporarily remove or disable each of those except the one using Auto Mask. IOW it may not appear as a stand-alone effect. But the overall area to examine is if the clip contains other effects besides the Auto Mask-enabled effect. If so, selectively remove those as an investigative step to see what combination of effects cause this. This would also apply to computationally-intensive timeline operations such as retiming with Machine Learning.