you lost me there, I have no idea what Turbo LLC is
I have noted down your post, I did already try undervolting once, but that was with the stock cooler still. Scores did get worse but temps stayed sucky.
Basically, LLC compensates for droop. Say you have the CPU running at 4GHz, 1.2V. While you're idle, everything looks normal. But when you fire up a full load benchmark like Prime95 Small for example, Vcore might drop to say 1.125V when the CPU is at 100% load. And so whether you're stable at 4GHz isn't dependent on the voltage that you've set (1.2V), it's dependent on the actual voltage after droop (1.125V). That's droop, and is engineered into VRM design.
LLC usually has a bunch of settings which vary from board to board, from default to really aggressive. The more aggressive you set LLC, the less you droop, but you increase the chance that you get an errant, larger and larger spike in Vcore in the instant when that benchmark begins and load is applied. Say that you still have that 1.2V Vcore, but now you set LLC on a relatively high setting; now when you fire up P95, your Vcore might only drop to 1.175V instead, even though your actual Vcore is still the same (1.2V) as before. But at maximum LLC settings, the spike introduced by LLC can become so large as to become a concern for chip longevity, especially if you're already running a fixed overclock with high voltage on any chip. In a split second when the benchmark starts, 1.2V may jump all the way up to 1.3V then immediately back down to 1.175V.
Because Matisse is self-regulating and always adjusting its Vcore (sub-0.3V) at idle, there's a lot that we don't know about how LLC or undervolting affects idle (from what I can see, it makes very little difference). But as with any other CPU, once that chip is at full sustained load, the same principles regarding Vcore, LLC and droop apply.
As a theoretical example, if you try -0.05V offset/High LLC and Stock Vcore/Default LLC, they may come out to the exact same load voltage and performance, but one runs tamer at idle.
Because we don't know exactly how Matisse is working at stock, it's hard to tell how these settings are working, but in my rig, the difference between -0.075V/Turbo and 0/Default is night and day. One is 10-15C cooler at idle, and always sounds like it's actually idle, while the other constantly sounds like it's ramping to 100% load on nothing. And I lose no performance at load compared to stock without PBO. I prefer to run -0.05V/High now for less aggressive LLC, it functions much the same and doesn't spike.
Basically, the issue with what a lot of people are doing with "undervolting" in the SFF space is that they're reducing the PPT value, or setting an aggressive undervolt (eg. -0.075V) without any LLC. If you reduce PPT, you simply prompt the chip to reduce clocks and performance, which may be beneficial in sub-10 litre cases where you can only make do with a crappy NH-L9a, but is otherwise undesirable. And undervolting without LLC, as we know, will reduce clocks and performance if Vcore is allowed to droop below the essential voltage that it needs for a specific clockspeed.
But every chip is different, so it's up to you to explore the voltage limits of yours and thus, how much undervolting room you have. LLC performance depends on the board too; some boards may take an extremely aggressive stance, while others may require max LLC just to make any difference to Vdroop. If you aren't too bothered about the idle behaviour, just leave it as it is. As long as 100% load temperatures don't reach 80C, you're fine.
. To be honest I think this stuff is a little out of my league, I don't wanna make matters worse by tweaking the wrong stuff, but I did watch some very informative bullzoid vids about it.
.
Don't have experience with either but I'd go for the Godlike. Your also invested in corsair from what I recall in your first post and that has the Corsair integration into that motherboard from what I saw in the manual.