Yes, filling is always fast (from someone’s experience), you can fill up the 16GB with data in no time, even on a much smaller bus. And there will always be that if you exceed the total capacity, you will have obvious performance issues. So you want to avoid this at all times.

Memory speed plays a different role and doesn’t have much to do with capacity, you want to match the total bandwidth (that’s not just the bus) as closely as possible to the GPU speed, since they both depend on each other. A 4060Ti with a 96-bit bus can choke the GPU to the point of potential loss. But this can be compensated for with a higher registered memory. As is often the case, many things play a role in the overall performance of a video card, so saying that 96-bit is too little is an understatement.

I wouldn’t be surprised if the 4060Ti can work just fine with a 96-bit bus because this generation is optimized very differently in this area. You can see that the 4070 with half a bus can actually provide the same performance as the 3080. A larger cache may play a role here.

All of this doesn’t change the fact that when memory overflows there will always be a performance issue, no matter how fast or slow the memory is. So the 16GB 4060ti will in some cases be more resistant to certain settings than the 4070ti. While the overall top speed is of course lower, the video card can in some cases outpace the 4070ti and this is certainly an odd choice from nVidia.