While it is good to standardise the heat sync factor for comparison test purposes, I don't see it as that important.
The CPU performance has dropped less than 10% over a sustained high use period.
10% CPU difference is not subjectively detectable in normal use.
I agree. It was just nice to confirm that the i5 wasn't behaving abnormally. The main reason my initial m3 had it doing better over time than the i5 was because I was using it on a stone countertop. Once I switched to wood, the m3 and i5 behaved similarly. However, the bigger point was that it appears there is no real point in getting the i5 (unless you need more storage), or the i7 for that matter... at least according to this test.
Over 10 runs, the fastest speed the i7 achieves is 275 (near the beginning of the 10 runs), and the slowest speed the m3 achieves is 246 (at the end of the 10 runs). The difference there is 29, which is just 11% slower than 275.
Furthermore, the biggest difference in this test on the same run between the i7 and the m3 is 268-250=18, or all of just 7%. As you said, subjectively there should be no discernable difference.
That said, for very bursty stuff, the i7 may feel a bit faster. According to Geekbench 4, it can achieve approximately 8450 multi-core, whereas the m3 achieves about 7050. That's a 17-20% difference. There is still no point for the i5 though, since it only achieves about 7480, which is only 6% difference vs the m3.
BTW, here again are Notebookcheck's 2016 Cinebench numbers:
2016 m3: 218 --> 185 (-15.1%)
2016 m5: 260 --> 229 (-11.9%)
2016 m7: 251 --> 207 (-17.5%)
Versus my m3 over the same 22 runs:
2017 m3: 265 --> 240 (-9.6%)
