Ryzen Power Reporting Deviation metric in HWINFO?

[maflynn]

Thoughts on this? I stumbled upon this metric within HWINFO the other day, when it was read at 86%. Oddly enough, I disabled what my bios called "Core Performance Boost" which seems like Intel's turbo boost. So instead of 86%, my now "slower" computer is reporting a larger variance, of 24% to 40% (the screen shot happens to show the low end) Underload it can go up to 80%

HWINFO has the explanation here: Explaining the AMD Ryzen "Power Reporting Deviation" -metric in HWiNFO

Toms Hadware has a somewhat sensualized story that indicates you will burn out your CPU
Ryzen Burnout? AMD Board Power Cheats May Shorten CPU Lifespan (Updated)

This was largely refuted by Gamer Nexus.

In all truthfulness, I'm not sure what to make of this, what does the 24.5% mean? Does it imply that the CPU is only getting 25% of the 65 watts or 25% more?

I'm largely siding with Gamers Nexus and thinking this is probably not a huge deal, and for my own peace of mind, I'll not use HWINFO.

 

[maflynn]

I'm still trying to get my brain wrapped around this but seeing this article helps to some extent
Updated: HWInfo Application beta Introduces Power Reporting Deviation Sensor (Cheats)

For example, if we take a Ryzen 7 3700X CPU that has 65W TDP and 88W default power limit (PPT), and use it on a board which has declared only 60% of its actual telemetry reference current, we'll end up with effective power limit of ~ 147W (88 / 0.6) despite running at stock settings (i.e. without enabling manual overclocking or AMD PBO). While the 3700X SKU used in this example typically cannot even reach this kind of a power draw before running into the other limiters and limitations, the fact remains that the CPU is running far outside the spec without the user even acknowledging it. This exploit can also cause additional cost and work to the consumer, who starts wondering about the abnormally high CPU temperatures and starts troubleshooting the issue initially by remounting the cooling and usually, eventually by purchasing a better CPU cooler(s).

So using the above equation, my 24.5% translates to 374W (88 / .245) which makes absolutely zero sense. Turning off "turbo" which definitely slowed my computer down, so its consuming less power then when I had it enabled and the metric was reporting 86% (88/.86) = 102W Even that seems too high for what the cpu can handle.

btw, my CPU is running very cool, which would not be the case if it was pulling in 374 watts.

 

[I7guy]

So from what I've read and can interpret, if the Ryzen cpu is fully loaded, let's say through running Cinebench benchmark, the deviation reported by HWINFO should be near zero%.

As I can gather by reading the links and watching the GN video, if the deviation is under 100%, GN gives a 5% deviation as ok, the cpu is being fed more voltage than necessary as the cpu believes there is more headroom.

So something seems amiss.

 

[GoldfishRT]

Edit: Just watched the GN video explaining it. My values would be invalid.

 

[maflynn]

I couldn't figure it out, but you're welcome to see my metrics (CPB disabled as well)

From my research, its a metric that HWINFO made up, but its based on the fact that board makers pass certain values to the CPU, so in effect they can "fool" the CPU, so it will draw more power then it ought too. I guess Intel does this too in different forms.

Overall, its not a huge issue, unless you're really over-clocking the CPU. If you're using it normally, then its largely a non-issue I believe.