Quit messing about with small random changes. Use a temperature tower that will change the temp for each of multiple levels and then pick the level that works best. One tower will probably take less time than you spend on one failed print based on a guess. If the best value is at the very top or bottom, you may have found a better temperature, but not yet the best. Re-run the test with that value in the middle, but with maybe a smaller range. There are plenty of available temp tower models available.
Better yet, use OrcaSlicer, which has a temp tower and other necessary calibration tests built in. Then you can slice the model in a familiar interface with more options than Studio offers.