New AMD Ryzen "Power Reporting Deviation" metric in HWiNFO

[Thanh Trần]

Hi , i have this situation. When i start my comuter and checking with HwInfo v6.27-4185 Sensor Status , My Power Reporting Deviation ~250% , when i running Cinebench R20 to get full load 100% CPU , My Power Reporting Deviation ~140% - 150% . So is there anything wrong with my computer?

 

[AsRock]

Take a gander at this.

 

[tabascosauz]

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.

An update: I've been doing a little bit of testing with my BR1500MS UPS, which shows real time power draw on the battery-protected sockets. I've only got my monitor and my main rig connected to battery power.

• At idle, the CPU is at about 20-30W power draw and the GPU (1070 right now because 2060S on RMA) is low, at about 5-10W. So we'll say 40W.
• On the UPS, idle power draw is about 45-48W with the monitor off and just the PC drawing power.

So at the very least, we know that the three fans in the case and three drives are barely drawing any power. The CPU is still doing its usual thing at idle.

• When we move to Prime95 Smallest, the CPU is fully loaded with AVX instructions, which causes clocks to remain at between 39.5x and 40x. HWInfo always reports that the CPU is at a steady "90W" of power draw, and temperatures range from 75-80C on a Dark Rock Pro 4.
• However, at load, the power draw on the UPS shoots up all the way to 170W. Same deal, monitor off. So at this point, while the CPU and GPU accounted for almost all of the wattage draw at idle, in Prime95 we have almost 70W of extra power draw.

The drives remain at millivolts' worth of power consumption, as they're not doing anything. The NF-A14iPPC-2000 is ramping up slightly and the Silent Wings 120mm and 135mm are up to full speed (although still quite slow and shockingly inaudible), but even the industrial grade A14 draws a maximum of 2.1W at full speed, which it isn't doing. So worst case unattainable scenario, the three fans pull 2W each for 6W total, and we'll give the drives another 5W of leeway at full power (which, again, they're not).

That still leaves us with 60W of unexplained power draw, considering the GPU is still idling at 5-10W, VRM losses included obviously. This board reports a deviation value of 76%.

I get the feeling that although the X570 Master has been mostly cleared of culpability, Gigabyte's B450 series boards may not turn out quite the same...

I'm starting to think that the reason why the B450I Aorus Pro WIFI suffers current overload with a 3900X and high VRM temps with everything else is not because its 4-phase IR35201 setup with IR3556 PowIRs is incapable of handling a 12-core, and instead because Gigabyte is putting 33% more power than is necessary through the VRM...

 

[TheoneandonlyMrK]

An update: I've been doing a little bit of testing with my BR1500MS UPS, which shows real time power draw on the battery-protected sockets. I've only got my monitor and my main rig connected to battery power.

• At idle, the CPU is at about 20-30W power draw and the GPU (1070 right now because 2060S on RMA) is low, at about 5-10W. So we'll say 40W.
• On the UPS, idle power draw is about 45-48W with the monitor off and just the PC drawing power.

So at the very least, we know that the three fans in the case and three drives are barely drawing any power. The CPU is still doing its usual thing at idle.

• When we move to Prime95 Smallest, the CPU is fully loaded with AVX instructions, which causes clocks to remain at between 39.5x and 40x. HWInfo always reports that the CPU is at a steady "90W" of power draw, and temperatures range from 75-80C on a Dark Rock Pro 4.
• However, at load, the power draw on the UPS shoots up all the way to 170W. Same deal, monitor off. So at this point, while the CPU and GPU accounted for almost all of the wattage draw at idle, in Prime95 we have almost 70W of extra power draw.

The drives remain at millivolts' worth of power consumption, as they're not doing anything. The NF-A14iPPC-2000 is ramping up slightly and the Silent Wings 120mm and 135mm are up to full speed (although still quite slow and shockingly inaudible), but even the industrial grade A14 draws a maximum of 2.1W at full speed, which it isn't doing. So worst case unattainable scenario, the three fans pull 2W each for 6W total, and we'll give the drives another 5W of leeway at full power (which, again, they're not).

That still leaves us with 60W of unexplained power draw, considering the GPU is still idling at 5-10W. This board reports a deviation value of 76%.

I get the feeling that although the X570 Master has been mostly cleared of culpability, Gigabyte's B450 series boards may not turn out quite the same...

I'm starting to think that the reason why the B450I Aorus Pro WIFI suffers current overload with a 3900X and high VRM temps with everything else is not because its 4-phase IR35201 setup with IR3556 PowIRs is incapable of handling a 12-core, and instead because Gigabyte is putting 33% more power than is necessary through the VRM...

Some of that extra power will be the cost of data on internal networks and the extra load on memory and other subsystems.
Think I'll retest with auto settings instead of default.
See if it shows up.

 

[tabascosauz]

Some of that extra power will be the cost of data on internal networks and the extra load on memory and other subsystems.
Think I'll retest with auto settings instead of default.
See if it shows up.

P95 Smallest and Small fit entirely in CPU cache. There's next to no memory load. That's why I ran it. Infinity Fabric usually draws a constant 10-15W at all times on Matisse (unlike Renoir where its frequency is dynamic) and is included in the package power number. And like I said, the drives and fans don't draw nearly that much power even if they were at full blast. VRM losses is the only culprit that I can think of, and I can't imagine it accounts for all of that already conservative 60W.

I'm already on "Normal" Vcore, which is slightly lower than the default Auto Vcore.

 

[thesmokingman]

An update: I've been doing a little bit of testing with my BR1500MS UPS, which shows real time power draw on the battery-protected sockets. I've only got my monitor and my main rig connected to battery power.

• At idle, the CPU is at about 20-30W power draw and the GPU (1070 right now because 2060S on RMA) is low, at about 5-10W. So we'll say 40W.
• On the UPS, idle power draw is about 45-48W with the monitor off and just the PC drawing power.

So at the very least, we know that the three fans in the case and three drives are barely drawing any power. The CPU is still doing its usual thing at idle.

• When we move to Prime95 Smallest, the CPU is fully loaded with AVX instructions, which causes clocks to remain at between 39.5x and 40x. HWInfo always reports that the CPU is at a steady "90W" of power draw, and temperatures range from 75-80C on a Dark Rock Pro 4.
• However, at load, the power draw on the UPS shoots up all the way to 170W. Same deal, monitor off. So at this point, while the CPU and GPU accounted for almost all of the wattage draw at idle, in Prime95 we have almost 70W of extra power draw.

The drives remain at millivolts' worth of power consumption, as they're not doing anything. The NF-A14iPPC-2000 is ramping up slightly and the Silent Wings 120mm and 135mm are up to full speed (although still quite slow and shockingly inaudible), but even the industrial grade A14 draws a maximum of 2.1W at full speed, which it isn't doing. So worst case unattainable scenario, the three fans pull 2W each for 6W total, and we'll give the drives another 5W of leeway at full power (which, again, they're not).

That still leaves us with 60W of unexplained power draw, considering the GPU is still idling at 5-10W, VRM losses included obviously. This board reports a deviation value of 76%.

I get the feeling that although the X570 Master has been mostly cleared of culpability, Gigabyte's B450 series boards may not turn out quite the same...

I'm starting to think that the reason why the B450I Aorus Pro WIFI suffers current overload with a 3900X and high VRM temps with everything else is not because its 4-phase IR35201 setup with IR3556 PowIRs is incapable of handling a 12-core, and instead because Gigabyte is putting 33% more power than is necessary through the VRM...

Did you forget to calculate efficiency factor into it? The hwinfo reading is straight from the cpu. The reading at the UPS is at the wall not factoring in PSU efficiency, etc etc.

 

[tabascosauz]

Did you forget to calculate efficiency factor into it? The hwinfo reading is straight from the cpu. The reading at the UPS is at the wall not factoring in PSU efficiency, etc etc.

I have a Focus Plus Platinum 550. At idle, it should be at a paltry 70-75% efficiency, but the idle readings are pretty much on the money. At between 75-200W, it's at around 85-92.5% efficiency depending on who you ask.

Point is, not all of that 60W+ is going to come from this specific firmware problem, but none of these peripheral factors account for more than half of that extra wattage, combined.

In the reported amp range, the IR3556s should be at 90%+ efficiency.

 

[harm9963]

Stock 2700X vs OC 2700X

 

[TheoneandonlyMrK]

Can't Get the thing to show up on mine?.

Default or auto.

 

[AsRock]

Can't Get the thing to show up on mine?.

Default or auto.

Uninstall and delete all left overs after maybe ?.

 

[Iceni]

Just did a stock run on my 3600 and got 102% with the MSI x570 gaming plus. So if anything it's over reporting and I'm loosing a little power.

 

[Assimilator]

An update: I've been doing a little bit of testing with my BR1500MS UPS, which shows real time power draw on the battery-protected sockets. I've only got my monitor and my main rig connected to battery power.

• At idle, the CPU is at about 20-30W power draw and the GPU (1070 right now because 2060S on RMA) is low, at about 5-10W. So we'll say 40W.
• On the UPS, idle power draw is about 45-48W with the monitor off and just the PC drawing power.

So at the very least, we know that the three fans in the case and three drives are barely drawing any power. The CPU is still doing its usual thing at idle.

• When we move to Prime95 Smallest, the CPU is fully loaded with AVX instructions, which causes clocks to remain at between 39.5x and 40x. HWInfo always reports that the CPU is at a steady "90W" of power draw, and temperatures range from 75-80C on a Dark Rock Pro 4.
• However, at load, the power draw on the UPS shoots up all the way to 170W. Same deal, monitor off. So at this point, while the CPU and GPU accounted for almost all of the wattage draw at idle, in Prime95 we have almost 70W of extra power draw.

The drives remain at millivolts' worth of power consumption, as they're not doing anything. The NF-A14iPPC-2000 is ramping up slightly and the Silent Wings 120mm and 135mm are up to full speed (although still quite slow and shockingly inaudible), but even the industrial grade A14 draws a maximum of 2.1W at full speed, which it isn't doing. So worst case unattainable scenario, the three fans pull 2W each for 6W total, and we'll give the drives another 5W of leeway at full power (which, again, they're not).

That still leaves us with 60W of unexplained power draw, considering the GPU is still idling at 5-10W, VRM losses included obviously. This board reports a deviation value of 76%.

I get the feeling that although the X570 Master has been mostly cleared of culpability, Gigabyte's B450 series boards may not turn out quite the same...

I'm starting to think that the reason why the B450I Aorus Pro WIFI suffers current overload with a 3900X and high VRM temps with everything else is not because its 4-phase IR35201 setup with IR3556 PowIRs is incapable of handling a 12-core, and instead because Gigabyte is putting 33% more power than is necessary through the VRM...

Unless you're measuring physical power draw on the ATX connectors, your readings are going to be wrong, and there's no point in trying to draw inferences from bad data. This is true whether you are taking said readings from your UPS or HWINFO. They. Are. Wrong.

 

[EarthDog]

Unless you're measuring physical power draw on the ATX connectors, your readings are going to be wrong, and there's no point in trying to draw inferences from bad data. This is true whether you are taking said readings from your UPS or HWINFO. They. Are. Wrong.

...in many cases... they aren't that far off, however.

 

[Goro_2030]

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

I have a B550 from ASUS, and the power deviation while doing NOTHING is at 115%, with peaks under load of 140% ... that is SO too much

 

[agentnathan009]

AMD has already addressed this issue that keeps popping up again because people panic about it and think their CPU is going to spontaneously combust like a punctured cellphone battery... Your CPU is fine and the Stilt didn't do anyone any good by mentioning this issue. AMD has safety limits in place to keep your CPU from dying when running default settings. Mobo manufacturers goose your CPU a small amount to gain a lead in benchmarks. It is not in the mobo manufacturers best interest to kill your CPU by dropping the hammer on your CPU. Stop worrying about it and stop the testing because most of you are not testing it per the Stilt's instructions anyway. CPU has to be full 100% load which Cinebench can do and you only check the deviation when Cinebench is running. Any CPU settings that you changed from default will change the reported number. CPU HAS TO BE RUN AT DEFAULT SETTINGS. If you are OC'ing then you are doing more damage to your CPU than the mobo maker goosing your CPU a bit at stock settings.

 

[Goro_2030]

AMD has already addressed this issue that keeps popping up again because people panic about it and think their CPU is going to spontaneously combust like a punctured cellphone battery... Your CPU is fine and the Stilt didn't do anyone any good by mentioning this issue. AMD has safety limits in place to keep your CPU from dying when running default settings. Mobo manufacturers goose your CPU a small amount to gain a lead in benchmarks. It is not in the mobo manufacturers best interest to kill your CPU by dropping the hammer on your CPU. Stop worrying about it and stop the testing because most of you are not testing it per the Stilt's instructions anyway. CPU has to be full 100% load which Cinebench can do and you only check the deviation when Cinebench is running. Any CPU settings that you changed from default will change the reported number. CPU HAS TO BE RUN AT DEFAULT SETTINGS. If you are OC'ing then you are doing more damage to your CPU than the mobo maker goosing your CPU a bit at stock settings.

Well, I think that's not my case, I'm not overclocking the CPU , just enabled AMD's "built-in" Precision Boost to Enabled ... is that considered Overclocking for the author?

 

[mstenholm]

Well, I think that's not my case, I'm not overclocking the CPU , just enabled AMD's "built-in" Precision Boost to Enabled ... is that considered Overclocking for the author?

Post is empty.

 

[agentnathan009]

Precision boost is ok, PBO is not default.

 

[Space Lynx]

I have a B550 from ASUS, and the power deviation while doing NOTHING is at 115%, with peaks under load of 140% ... that is SO too much

I already got me a MSI X570 Tomahawk, and I love it. VRM temps according to HWINFO are not going over 38 celsius while gaming... this mobo is so much overkill. Not bad for $219

MY CPU si ryzen 3600, with plans to upgrade to 4800x.

 

[Goro_2030]

I already got me a MSI X570 Tomahawk, and I love it. VRM temps according to HWINFO are not going over 38 celsius while gaming... this mobo is so much overkill. Not bad for $219

MY CPU si ryzen 3600, with plans to upgrade to 4800x.

Which cooler do you have man?!?!

My 3700X with the stock cooler won't be lower than 60 at idle ... One other thing that's strange from the temps is how it goes up and down all the time (look at the little upticks, or triangles in the measurements of core-temp , attached )

I tried everything ... my CPU is not overclocked, nor even with PBO. Just D.O.C.P . Bought new high quality thermal paste that just helped reduce a degree

Any ideas?

 

[Mussels]

Which cooler do you have man?!?!

My 3700X with the stock cooler won't be lower than 60 at idle ... One other thing that's strange from the temps is how it goes up and down all the time (look at the little upticks, or triangles in the measurements of core-temp , attached )

I tried everything ... my CPU is not overclocked, nor even with PBO. Just D.O.C.P . Bought new high quality thermal paste that just helped reduce a degree

Any ideas?

fan curve, probably. high idles are normal on ryzen as i've been told it reads from the hottest core, its not the average over the dies

 

[Goro_2030]

fan curve, probably. high idles are normal on ryzen as i've been told it reads from the hottest core, its not the average over the dies

Yeap, maybe that ... as I have the fan curve to Ultra-Silent as I have an open case and it's really annoying . I have the stock fan at 1255 rpm up to 77 degrees Celsius.

Something that cought my eye as well is that Fan Xpert from ASUS reports a way lower temperature than Core Temp or HWInfo's TCTL/TDIA CPU TEMP ( I'm looking at it now and it's reporting 53 degrees, while the TCTL/TDIE temp is at 65 )

 

[Space Lynx]

Which cooler do you have man?!?!

My 3700X with the stock cooler won't be lower than 60 at idle ... One other thing that's strange from the temps is how it goes up and down all the time (look at the little upticks, or triangles in the measurements of core-temp , attached )

I tried everything ... my CPU is not overclocked, nor even with PBO. Just D.O.C.P . Bought new high quality thermal paste that just helped reduce a degree

Any ideas?

if you are using wraith prism, there is a little switch on top of the cooler that has L and H, switch it H, it default is at L. your temps will be better now. will need a flashlight to spot it.

 

[Goro_2030]

if you are using wraith prism, there is a little switch on top of the cooler that has L and H, switch it H, it default is at L. your temps will be better now. will need a flashlight to spot it.

Thanks for taking the time to responding so quickly! I did what you mentioned on day one, but that is not the issue. Having it on High allows the fan to go to 3700 rpm vs 3100 of the Low configuration. The curves do not change, sinceI control them with Asus'es software.

Basically my question is... If someone here does have a 3700x with a stock cooler and you at the fan to ultra quiet ( meaning at minimum rpm until it hits 75C)... At which temperature it hovers? I've seen anything from 35 idle to my 60 idle...

My room temperature is 21C, btw

 

[Mussels]

dont use ultra quiet, go set a manual curve of like 1000RPM til 60c, and raise it from there. If you have an option to slow down the changes i set like a 5s delay on my board, which means it smoothly ramps.