Okay so correct me if I am wrong, but I am under the assumption that the 4000 series cards will downclock when their memory becomes unstable instead of artifacts/crashing, therefore in order to properly overclock the 4000 series you should instead be benchmarking when you overclock the memory to make sure the score is not going down, is this method correct?
For example a week ago, I was overclocking the 4070ti ventus memory and running simultaneously a superposition 4k benchmark to make sure that every time I increased the memory clock by +100 in afterburner, that superposition score wasn't going down
Here was my results: https://imgur.com/a/ae87WKh
Basically my benchmark score kept going up (except from +1200 to +1300) until I reached +1500, then at +1600 it got slightly worse and at +1700 it crashed. So I decided that +1500 was probably the best I could push on this card and now I am stress testing the memory with port royal + games in order to push the tensor and RT cores not just CUDA.
Assuming +1500 is stable after more in depth stress testing, should I try to get stable at +1600 or because the superposition benchmark showed the score decreased and because +1700 is for sure unstable, therefore +1500 is probably best for this card?!