New AMD Ryzen "Power Reporting Deviation" metric in HWiNFO

[windwhirl]

Just saw this around.

IMPORTANT - Explaining the AMD Ryzen "Power Reporting Deviation" -metric in HWiNFO

Ryzen CPUs for AM4 platform rely on external, motherboard sourced telemetry to determine their power consumption. The voltage, current and power telemetry is provided to the processor by the motherboard VRM controller through the AMD SVI2 interface. This information is consumed by the processors...

www.hwinfo.com

Some motherboard manufacturers intentionally declare an incorrect (too small) motherboard specific reference value in AGESA. Since AM4 Ryzen CPUs rely on telemetry sourced from the motherboard VRM to determine their power consumption, declaring an incorrect reference value will affect the power consumption seen by the CPU. For instance, if the motherboard manufacturer would declare 50% of the correct value, the CPU would think it consumes half the power than it actually does. In this case, the CPU would allow itself to consume twice the power of its set power limits, even when at stock. It allows the CPU to clock higher due to the effectively lifted power limits however, it also makes the CPU to run hotter and potentially negatively affects its life-span, same ways as overclocking does. The difference compared to overclocking or using AMD PBO, is that this is done completely clandestine and that in the past, there has been no way for most of the end-users to detect it, or react to it.

HWiNFO will display "Power Reporting Deviation" metric under the CPUs enhanced sensors. The displayed figure is a percentage, with 100.0% being the completely unbiased baseline. When the motherboard manufacturer has both properly calibrated and declared the reference value, the reported figure should be pretty close to 100% under a stable, near-full-load scenario. A ballpark for a threshold, where the readings become suspicious is around ±5%. So, if you see an average value that is significantly lower than ~ 95% there is most likely intentional biasing going on. Obviously, the figure can be greater than 100%, but for the obvious reasons it rarely is

As stated before, this metric is only valid during a relatively stable near-full-load condition. That is due to the typical measurement accuracy of the VRM controller telemetry, and also due to the highly advanced and fast power management on Ryzen CPUs, that not only result in extremely low idle, but also in extremely rapidly changing power consumption. A suggested workload to get a stable and reproducable deviation metric is Cinebench R20 NT, with the HWiNFO sample rate set to less or equal to 1000ms.

I'd like to stress that despite this exploit is essentially made possible by something AMD has included in the specification, the use of this exploit is not something AMD condones with, let alone promotes.
Instead they have rather actively put pressure on the motherboard manufacturers, who have been caught using this exploit.

I was not aware that motherboard makers tampered with the metrics in such a blatant way.

By the way, this feature is only available in the latest beta version of HWiNFO, v6.27-4185.

 

[tabascosauz]

So...does this affect the power consumption reported by the CPU, or is the wattage still roughly accurate because the board is only lying to the CPU? Well, this is a Gigabyte board, where DRAM is overvolted up to an automatic +0.3V beyond what you set it, so I can't say I'm very surprised at the 76% result.

Somehow, I don't feel as though Cinebench is telling the whole story. All these benches behave very differently, and CB tends to set a temp target of 70C for itself, as well as reduce clockspeeds and Vcore as the test progresses. However, I can confirm that it is the same 76% reading in P95 Smallest, which by default runs lower clocks because of AVX but draws the same power.

 

[windwhirl]

So...does this affect the power consumption reported by the CPU, or is the wattage still roughly accurate? Well, this is a Gigabyte board, where DRAM is overvolted up to an automatic +0.3V beyond what you set it, so I can't say I'm very surprised at the 76% result.

View attachment 158301

Somehow, I don't feel as though Cinebench is telling the whole story. All these benches behave very differently, and CB tends to set a temp target of 70C for itself.

I myself don't know what to think of the short testing I did. I got 96.5% minimum and 155.3% maximum, with an average of 117%. Ironically, it was nearly stable at 100% during the Cinebench run, but after it sort of went bananas.

 

[Space Lynx]

and here I was thinking MSI meant it when they said Return to Honor... sigh. unless MSI is not doing this, though I suspect all of the mobo makers do it? can anyone confirm? AMD seems to not approve regardless lol

 

[windwhirl]

and here I was thinking MSI meant it when they said Return to Honor... sigh. unless MSI is not doing this, though I suspect all of the mobo makers do it? can anyone confirm? AMD seems to not approve regardless lol

Here is an practical example recorded on MSI X570 Godlike motherboard, using the most recent 1.93 beta-bios version.
For this bios version MSI has declared 280A reference current, when the correct value that produces near 100% result (i.e. no deviation) and also a matching power draw compared to other boards (same CPU and workload) is 300A. This means that the board allows 7.14% (300/280) higher power draw for the CPU than AMD specifications state. Compared to the worst violators (up to 50%) this is minor infraction, so MSI deserves a benefit of a doubt whenever this is intentional or a honest error.

The guy tested that board only in that article, but maybe this will be of interest to you

As of now, outside of certain MSI motherboards, the biasing isn't end-user controllable. In case there is clear evidence of biasing taking place on certain motherboards or their bios versions, please contact the manufacturer and ask them to remove the telemetry biasing from the bios. The biasing can be implemented in different ways, it can be tied to a specific setting(s) (known as an "auto-rule") in the bios or be fixed in a certain bios version or in all available bios versions.

 

[EarthDog]

Feels like what they do with intel power limits... ish. Or what card partners did/do with nvidia power limits. Lol

 

[thesmokingman]

Finally we can call BS on these BS motherboard speed tests results. Note I'm referring to board makers not those that run the tests. The boards technically should all perform the same but in reality they are damn varied.

So...does this affect the power consumption reported by the CPU, or is the wattage still roughly accurate because the board is only lying to the CPU? Well, this is a Gigabyte board, where DRAM is overvolted up to an automatic +0.3V beyond what you set it, so I can't say I'm very surprised at the 76% result.

View attachment 158301

Somehow, I don't feel as though Cinebench is telling the whole story. All these benches behave very differently, and CB tends to set a temp target of 70C for itself, as well as reduce clockspeeds and Vcore as the test progresses. However, I can confirm that it is the same 76% reading in P95 Smallest, which by default runs lower clocks because of AVX but draws the same power.

Wattage is not reliable since the MB is faking a low output to the cpu there in turn the cpu raises available power draw.


Cinebench cannot do any of what you suggest, it's an app. It doesn't have any power to control the algorithms in the cpu unlike the MB. Whatever it is doing the cpu is trying to stay inside it's power draw limit. P5 small is a much greater load than Cinebench so of course it is going to use a lower clock to maintain said power limit. You can see this in effect by running Small vs Smallest for yourself.

 

[R0H1T]

Does the reading rely on actual TDP (rated) for the processors or what?

Besides that we've seen power readings on various reviews, even if motherboard (vendors) are lying to the processor as long as the temps are in check it shouldn't matter too much. AMD enforces aggressive TDP limits on their desktop processors & it can hurt performance in some cases.

 

[AsRock]

The beta been removed or some thing ?, all links go to last stable.

 

[R0H1T]

Try this ~

Download Portable HWiNFO64

Download Portable HWiNFO64 7.72 Build 5355 / 7.73 Build 5385 Beta - Displays the hardware configuration of your computer, providing support for HTML, XML, CSV or text reports that can be saved on the local PC

www.softpedia.com

 

[tabascosauz]

Does the reading rely on actual TDP (rated) for the processors or what?

Besides that we've seen power readings on various reviews, even if motherboard (vendors) are lying to the processor as long as the temps are in check it shouldn't matter too much. AMD enforces aggressive TDP limits on their desktop processors & it can hurt performance in some cases.

Last AM4 CPU that I can think of actually kinda abided by TDP is the 2700, which actually sipped power but a little too much.

Ryzen CPUs all abide by PPT, TDC and EDC for their power and current limits. Thus it makes a lot of sense why this misreporting would be a problem. PPT is already considerably higher than TDP, approx 88W for 65W and 142W for 105W parts. Moreover, SMU power draw regularly exceeds even that number, without PBO, on full load for my chip.

@thesmokingman I just want to know if the SMU wattage readings under the CPU sensors in HWInfo are skewed because of this, or if it just affects the CPU's boost algorithms internally. Because despite purportedly being 25% off, the wattage readings coming from the IR35201 controller don't usually appear to be absurdly off in comparison to the SMU power draw. Misreporting CPU wattage is one thing, but I find it hard to believe that the IR35201 would be skewed on this as well as to the power it's providing to the socket.

 

[thesmokingman]

The beta been removed or some thing ?, all links go to last stable.

Link works for me. When you click thru the beta is the grey option.

Last AM4 CPU that I can think of actually kinda abided by TDP is the 2700, which actually sipped power but a little too much.

Ryzen CPUs all abide by PPT, TDC and EDC for their power and current limits. Thus it makes a lot of sense why this misreporting would be a problem. PPT is already considerably higher than TDP, approx 88W for 65W and 142W for 105W parts. Moreover, SMU power draw regularly exceeds even that number, without PBO, on full load for my chip.

@thesmokingman I just want to know if the SMU wattage readings under the CPU sensors in HWInfo are skewed because of this, or if it just affects the CPU's boost algorithms internally. Because despite purportedly being 25% off, the wattage readings coming from the IR35201 controller don't usually appear to be absurdly off in comparison to the SMU power draw. Misreporting CPU wattage is one thing, but I find it hard to believe that the IR35201 would be skewed on this as well as to the power it's providing to the socket.

I think you are asking if the reading from the motherboard/vrms is accurate? I suppose they would be as the report didn't specify that the MB was giving lower readings to the user facing side. Then again as you wrote, 105w parts are drawing 142w and so it lends to the notion that we are seeing real numbers on our side. Percentage wise that is a grip that the board is pushing the cpu up.

It is hella sketchy imo, with certain boards already goosing it by that much and then people jack PBO up to max on top of that thinking they are within safe parameters...

 

[tabascosauz]

Link works for me. When you click thru the beta is the grey option.

I think you are asking if the reading from the motherboard/vrms is accurate? I suppose they would be as the report didn't specify that the MB was giving lower readings to the user facing side. Then again as you wrote, 105w parts are drawing 142w and so it lends to the notion that we are seeing real numbers on our side. Percentage wise that is a grip that the board is pushing the cpu up.

It is hella sketchy imo, with certain boards already goosing it by that much and then people jack PBO up to max on top of that thinking they are within safe parameters...

No no, the higher PPT for all chips is by design and normal behaviour. But yeah, I just find it hard to believe that the IR controller also misreports. My rig is hooked up to a UPS that constantly reports power draw; taking into account power draw of fans and drives and the GPU, I don't think I've ever seen this ~30W full load discrepancy that a 76% reading would imply, on either IR or SMU readings.

Stilt suggests that this delta is fixed, which means that even though the actual limit is 140W+, no sane 3700X on stock settings will ever hit that without running into its temp or voltage limits.

 

[thesmokingman]

I just went thru bios and set power draws to 100%, disabled PBO in tweaker and advanced cbs/oc menus. I am getting 96%-98% on deviation which I assume to mean it is on target. If you are only seeing 76%... I'd check your bios settings. My board is a Strix-E btw. It'd be hilarious though if my board was cheating since it was Stilt who pointed this out, lol. Gonna run r20 again to be sure.

So, if you see an average value that is significantly lower than ~ 95% there is most likely intentional biasing going on. Obviously, the figure can be greater than 100%, but for the obvious reasons it rarely is...

*edit

Yeap, my deviation % stayed between 96%-98% thru the whole R20 run.

I myself don't know what to think of the short testing I did. I got 96.5% minimum and 155.3% maximum, with an average of 117%. Ironically, it was nearly stable at 100% during the Cinebench run, but after it sort of went bananas.

Don't look at average unless you reset the counter. But you wrote yours went up to 155%? That's a huge arse overshoot.

 

[windwhirl]

Don't look at average unless you reset the counter. But you wrote yours went up to 155%? That's a huge arse overshoot.

I monitored Hwinfo while I was doing Cinebench. During the benchmark it more or less remained around 100%. When Cinebench stopped and I started doing other things while I let the software keep recording data the Power Reporting Deviation started jumping up (I had completely forgotten about it and I caught it going over 140% when I minimized some apps). The capture I posted was from over half hour of HWinfo monitoring the sensors.

View attachment 158303

To be frank, I'm not sure if it's just HWInfo being still a little "green" so to speak or if I have some erroneous BIOS/UEFI configuration to fix. EDIT: I sort of have to go to sleep now, so I guess later I'll reset everything or something...

 

[thesmokingman]

I monitored Hwinfo while I was doing Cinebench. During the benchmark it more or less remained around 100%. When Cinebench stopped and I started doing other things while I let the software keep recording data the Power Reporting Deviation started jumping up (I had completely forgotten about it and I caught it going over 140% when I minimized some apps). The capture I posted was from over half hour of HWinfo monitoring the sensors.



To be frank, I'm not sure if it's just HWInfo being still a little "green" so to speak or if I have some erroneous BIOS/UEFI configuration to fix.

Oh I think you are ok then. What we're concerned with is the % when under a full load.

 

[Mussels]

well this could explain why some people have over-cooked ryzen builds that wont behave

makes me wonder if some of the 'enhancement' options in the BIOS control this

"As stated before, this metric is only valid during a relatively stable near-full-load condition " - ahah, cinebench time!

Asus did good

 

[Space Lynx]

well this could explain why some people have over-cooked ryzen builds that wont behave

makes me wonder if some of the 'enhancement' options in the BIOS control this

"As stated before, this metric is only valid during a relatively stable near-full-load condition " - ahah, cinebench time!

Asus did good

I'm leaning towards Asus B550 now, trust is important to me.

 

[AsRock]

Oh I think you are ok then. What we're concerned with is the % when under a full load.

and goes below 100%, as if i read it right any thing over +-5%. So if it's reading like mine is around 93% which indicates 2% extra power going to the cpu.

 

[thesmokingman]

and goes below 100%, as if i read it right any thing over +-5%. So if it's reading like mine is around 93% which indicates 2% extra power going to the cpu.

Stilt mentioned 95%-100% is on target. If you get drastically below 95% (like 75%) then shenanigans are at play. I think you are ok on that front.

 

[AsRock]

Stilt mentioned 95%-100% is on target. If you get drastically below 95% (like 75%) then shenanigans are at play. I think you are ok on that front.

Thought as much, i aint worried about 7% lol.

EDIT: Added pic

 

[GoldenX]

TFW you have a 3000 series Ryzen but it isn't a Zen2. Oh well, instructions unclear, pushed PBO scalar to 10x anyway.

 

[Mussels]

TFW you have a 3000 series Ryzen but it isn't a Zen2. Oh well, instructions unclear, pushed PBO scalar to 10x anyway.

try the various steps in between, dont just assume max is best

 

[R0H1T]

The issue could be major if the motherboard is pumping high voltage at peak loads consistently ~ that's when this becomes a big deal.

 

[Assimilator]

https://twitter.com/i/web/status/1270245205090934784

"For AMD users, it's a nothingburger."