Instead of pasting the wall of text, there is a very good writeup about it in the HWiNFO forum.
But the TL;DR of it is that AMD's spec allows the motherboard manufacturers to basically lie in their reporting to the CPU as to how much power it's actually drawing which allows for higher clocks at the expense of power consumption and heat production, and is no doubt a cause of a lot of these "WTF IS WITH THESE TEMPS WITH THE STOCK COOLER AND STOCK SETTINGS OMG!!!" types of threads we have seen on here.
Now, AMD knows about this deception, but they really can't do anything about it. The article states they "actively put pressure" on them to stop, but they don't really have the power to do it.
Now, the way you use it is to download it, which is available only as a portable version right now, which isn't a big deal since it won't mess with your current install, and let it run while you run a processor intensive task, they recommend Cinebench R20 which really is the most processor intensive yet realistic load you can put on it. Now, I'm overclocking and undervolting which explains the greater than 100% figures, but when you start seeing figures of 90%, 80%, or even 50%, that's where it's reporting up to half of the actual power consumption
And again, to quote:
Until now, you just couldn't actually see it.
TomsHardware put out an article on it which is a bit easier to understand perhaps than my summation
Anandtech has a very wonderful article on this subject as well, with a detailed explanation as to why it's not that detrimental on the lifespan because of the other limits AMD has in place, unlike Intel which basically give motherboard manufacturers free reign.
One has to wonder too if this is the real cause of those 1.45-1.50v readings as well. Even though electromigration isn't going to be a problem in the typical lifespan of a CPU these days, heat and excessive power will cause an already weakened component to fail faster. When you're talking about a CPU as complex as a Ryzen processor, especially Ryzen 3000 series, that's a lot of potential points of failure, and likely the reason AMD actively discourages it.
But all processors have built in Maximum safety limits that will prevent the processor from being damaged. So even if the Motherboard manufacturers are using unfriendly electronic tactics or workarounds (unethical practices in my opinion) to give the illusion that the processor is not running as hot nor using as much power on their motherboards, The Ryzen's built in Maximum safety limits will kick in and still prevent the processor from going past it Maximum safety limits in voltages or temperatures.
But as you mentioned, if the Ryzen, as an example, is running at a vCore voltage near its maximum safety limit but under its Maximum safety range and running hot and the Users believe the vCore voltages is 50% less this will eventually shorten the CPU's lifespan from the processor running at a higher than normal vCore voltages and temps associated with it.
Also as you mentioned if by any chance the Processor is weak for what ever reasons, the higher voltages and temps will cause it to fail much earlier than if it would have run within its normal safety range and not it maximum safety range.
The only trouble is the Users will believe something is defective with either the CPU or Motherboard when the CPU runs hotter than it should be or shuts down automatically when the Temps are within normal range.
anyway that is the way I personally see it from what I read in the articles mentioned. But my view is probably flawed technically since I don't have an Bachelor's in Science in Computer Electronics.
I spelled out BS - Bachelor's of Science not to be confused with a bull's feces (-: .
Out of curiosity I dropped my overclock and reset everything to auto, and took these screenshots right after Cinebench R20 ended. The Crosshair VI Hero isn't an egregious violator either, average 8.6% variance.
And the full display with overclock is
Interesting figures to say the least. CPU VID (Effective) reduced from 1.35v to 1.1v results in 9*C lower temps despite a 275mhz per core increase, and about 5w lower power consumption per core as well. Something I never paid attention to since I tend to use AIDA64 and not HWiNFO...
Anyway yea, I don't have a degree in computer science or electrical engineering either, just a couple decades of building, tweaking, and F'ing things up.
It's also a bit unnerving that AMD was aware of this practice yet didn't disclose it...
AMD probably did but in private not publicly. AMD probably doesn't want to upset or embarrass some major Motherboard manufacturers by going public.
Your Ryzen Maximum Temps under Stress are very good. They maintained within a certain close range of just a few Cs apart from Minimum and Maximum Temps.
EDIT: HWINFO64 - Power Reporting Deviation unfortunately is for Ryzen processors only. But ran Cinebench 20 and according to HWINFO64 my maximum CPU Temperature for my FX8350 was 53C while in CoreTemp it was 59-60C. But my Asus Motherboard CPU sensor was slightly hotter reporting a Maximum CPU Temperature of 55C.
So my Cooler Master Hyper 212 EVO is actually working better than I had imagined. I have 2 fans on the Cooler Master. One controlled by the motherboard (PWM) and the other connected to a case fan port with no PWM (3 pins). The one connected to the Case fan has a minimum and maximum of 1980 RPM. While the PWM Fan had a minimum of ~840 RPM and maximum of ~1984 RPM.
My Cinebench results were poor needless to say. HWINFO64 had me disable an embedded Asus Motherboard Sensor plus I have a huge amount of background apps running.
Corsair H115i isn't exactly weak, though it's warmer than I'd like to see, but that's just me speaking with so many years experience with AMD CPUs which had very low thermal maximums as compared with Intel's which were always high, especially the FX-8350 which has a Tmax of 61 despite being a large, inefficient, slow piece of silicon. The (VERY expensive) maglev fans I have on there, Corsair CO-9050044-WW ML140, work wonders as well. Even under Cinebench R20 or Handbrake video encoding (non GPU assisted for max load) it's not bad, and all low pitched so no annoying whines.
So in a benchmark review comparing a Ryzen CPU and a competing Intel CPU it's possible the Ryzen is technically overclocked? Wouldn't this skew the results in favor of the Ryzen? Maybe that is part of the reason AMD didn't want to go public with this.
No, because motherboard manufacturers do the same thing with Intel CPUs as well. The Anandtech article explains it better than I can:
As well as the Gamers Nexus article on how motherboard manufacturers exploit Intel's MCM and technically overclock without actually overclocking.
What this says to me is it's impossible to compare two different CPU's. Unless you have extensive knowledge of these things. Which I'm betting 95% of people don't. I sure wasn't aware that a mainboard could do that and I've been messing with them for over 25 years. There's no way to say Ryzen is better or Intel is better without including the mainboard they've been tested with. Sounds like a big can of worms.
AMD could change their Turbo functionality to match that of Intel's with turbo tables and time limits, but then we'd see these low TDP processors blow way past that, like we see with the 9900K and 10900K, then again we would also likely see AMD's stated maximum Turbo frequency actually able to be sustained, but not unless they stop shooting enormous amounts of voltage to obtain it.
I'll use TomsHardware's chart from the 10900K review as an example as well as the excerpt which talks about TVB which is supposed to limit the CPU frequency after it passes a certain temperature.
The key words being "functionality will vary by board".
Thankfully in the overall picture there are other differences which cause AMD or Intel to be the winner, be it absolute performance, power consumption, higher core counts, feature levels, and cost to performance ratio, among others, so you can't just point to just one metric, and this can of worms stays fairly innocuous.
The only real solution I can think of is for AMD and Intel to require board partners to manufacturer a to-spec "Reference Edition" motherboard, and the AIBs would just scoff at that.
I have a power monitor on the wall plug that monitors power load. This is the only real measure.
CPU and GPU devices are so complicated that only an estimate of power consumption can be made.
Media conversion uses my CPU and GPU and power consumption can be high.
So, seeing this message pop in my HWINFO I started looking into it, only to see there really is nothing that can be done. Kind of sucks, because apparently this may be the reason my machine crashes when i attempt to run OBS to try streaming. I've replaced the CPU with an upgrade, added more DIMM, and upgraded the GPU. A game that i should be able to stream with no issue, IE RedM or Farm Sim 22 and if i have too much open it crashes, and i have to have Steam open, OBS, and if running RedM i have to have the R* launcher open and i usually my Discord open all the time.
I'm honestly seriously beginning to think this has been the culprit this whole time Here's HOPING AMD will somehow put out a fix patch to fix this issue. It hurts multitaskers, but i won't hold my breath.