SetsulNot really. There's 4400+ RAM on the QVLs. It's 2:1 for the IF clock beyond 3733 though so it only benefits bandwidth, not latency.
You asked for benchmarks with 3733, so now you don't get to act like 3200 vs 4133 is a completely fair and realistic scenario.
Either way different demos and different configs so I'm not sure where you're going with this. Or did you really think that just pushing it to 5.0 and using 4133CL17 would make the 8350K >50% faster than a 3600X?
The only other results (equally unfair) are actually higher than yours. So I don't get where the "pounding all these Zen2 chips" talk is coming from.
Clocks != voltage. 4.9 is less common than 4.8, 5.0 is less common than 4.9 and so on. Big surprise. That doesn't tell you what they consider excessive vcore or if they are limited by voltages at all, not thermals.
On top of that 8350Ks that make it to 5.2 by silicon lottery's standards exist and I can absolutely guarantee that those don't need excessive vcore for 4.9. I mean they claimed 82% to 4.9, 59% to 5.0. Does that mean they always use excessive vcore? That'd be your definition then, not theirs because they obviously think that vcore was fine.
Didn't realise people were using different benchmarks now, I just assumed everyone was posting benchmark1.dem results so there would be a common reference point.
And above 1.4v is excessive vcore for Coffee Lake because it's well beyond the optimal range of efficiency. Beyond 1.4v you start getting very large increase in heat generation for little improvement in clocks. Not unsafe (1.45+), but excessive.