Actually if you read the entire article, including the cartoonish waveforms which are proof of nothing, the whole point of even that article is overshoot volts. If I set bios to 1.4, and vdrop is to 1.35 idle, and vdroop is to 1.3v load, then providing mobo manufacturer followed intel specs, overshoot spikes are to 1.4 (bios setting) or .1v.
Now if I use loadline enabled, and set bios to 1.35 and idle is 1.33 and load is 1.3, the overshoot from same load is still going to be .1v or to 1.4v from 1.3. As long as you can do the simple math and not fool yourself into thinking the overshoots are not there...LLC by itself is not going to kill anything. But if you are using 1.5LLC vcore with overshoots to 1.65V thinking it is same as setting 1.5 bios without LLC, then you have a problem.
According to the anandtech article the above is correct in terms of overshoot volts. There conclusion based on 1 mobo, 1 older vrm circuit that volts were more erratic b/c overclock needed little more vcore, is not going to persuade me of anything. 50 mobos and take an average with 50-100K multi capable of accurately measuring waveforms and detecting slight volt differences, and we will talk. But even then, I can tell you on my mobo, LLC makes it easier not more difficult to OC, so I have different experience than them.
All of which is irrelevant anyways. Until real world experience shows me that vcore I am using with LLC enabled is harming my chips, I will continue to do so.
As an aside feeding a loaded chip less volts than unloaded is idiotic. It is simply cheaper to implement that piss poor strategy via circuitry than better much more expensive one of designing circuitry with less overshoot. But as long as I can add and subtract, I will be using LLC.
Bookmarks