Monday, June 8th 2020

HWInfo's Power Reporting Deviation Sensor Reveals How Motherboard Makers are Cheating Ryzen CPUs

HWInfo with its latest version, introduced a new sensor that reveals whether your motherboard is tricking your 3rd Gen Ryzen processor into thinking its power consumption is within normal parameters, while in fact sending more power to it, to sustain boost clocks better, or support overclocking. These enhancements take advantage of the fact that an AMD Ryzen processor relies on the motherboard's CPU VRM controller for power-consumption telemetry, so the processor's power-management co-processor can accordingly adjust boost frequencies to stay within AMD's power-consumption limits.

Motherboard vendors have allegedly figured out a way to trick the processor into thinking its power consumption is within normal parameters (when it's not), and HWInfo's developer has determined a way to calculate the deviation between power consumption value reported to the CPU, against that measured by the VRM controller. This sensor is called "Power Reporting Deviation," and is an integer percentage value. 95-105% deviation can be interpreted as normal behavior, where the motherboard is respecting AMD specs. Anything outside this range could indicate a motherboard-level power enhancement designed to maximize performance of even processors that users prefer to run at stock speeds, reducing their lifespan. Overclocking legend "The Stilt" wrote a detailed essay on Power Reporting Deviation, which can be read here.

DOWNLOAD: HWInfo v6.27-4185 Beta
Add your own comment

32 Comments on HWInfo's Power Reporting Deviation Sensor Reveals How Motherboard Makers are Cheating Ryzen CPUs

#26
evernessince
PanicLakeBut the article also say: "The issue with using this exploit is, that it messes up the power management of the CPU and potentially also decreases its lifespan because it is running the CPU outside the spec..."
So TPU title even if click-baity is not false.
Running a CPU out of spec doesn't lower life span per say. For example, there are plenty of Intel motherboards that boost for periods longer then Intel spec, even with MCE disabled. You'd really need to prove that degradation would indeed occur at whatever settings a particular motherboard is using.
Posted on Reply
#27
employee24601
Wow. I'd love to see a list of offending boards & manufacturers compiled at some point (noting BIOS revisions). I'm still rocking a GTX 1080 Ti & 4790K, but my next build is absolutely going to be Ryzen. Frankly, I'd rather support a company that doesn't risk or reduce the longevity of their their customers hardware by gaming the numbers like this.
Posted on Reply
#28
TheoneandonlyMrK
Is this just on X570, I cannot get a value out of hwinfo for my x470, see below?

no sign of it, I checked in settings , it is not hidden or unmonitored, it isn't there?.

two motherboard bios 3004,3101 tested, no sign.

Posted on Reply
#29
thesmokingman
evernessinceRunning a CPU out of spec doesn't lower life span per say. For example, there are plenty of Intel motherboards that boost for periods longer then Intel spec, even with MCE disabled. You'd really need to prove that degradation would indeed occur at whatever settings a particular motherboard is using.
If that were the only thing wrong however you and Ian it seems fails to realize that before this ya didn't know your board could be lying to the cpu. Then you go to do your overclock and further run your chip into the redzone all the while not knowing how close you are to it. That's just not cool. And on the topic of degradation the Stilt's already shown us maximum FIT voltage to prevent it and it is not hard to degrade a matisse cpu. Stilt's had degradation running as low as 1.33v in about a month.
theoneandonlymrkIs this just on X570, I cannot get a value out of hwinfo for my x470, see below?

no sign of it, I checked in settings , it is not hidden or unmonitored, it isn't there?.

two motherboard bios 3004,3101 tested, no sign.

I think it is Matisse cpus only. I tested my kids 1600AF on x570 and did not see the deviation reading either.
Posted on Reply
#30
EarthDog
thesmokingmanI also think motherboard reviewers need to read this asap. They need to dig into the bios and disable all the auto oc features so we get a real feel for boards.
I disagree with this sentiment.

Most users simply don't and will not do such a thing. IMO, testing should be done setting optimized defaults and XMP/DOCP. This is the way you get 'out of box' performance as everyone else would. What people tweak and such is impossible to get consistent.
Posted on Reply
#31
TheoneandonlyMrK
EarthDogI disagree with this sentiment.

Most users simply don't and will not do such a thing. IMO, testing should be done setting optimized defaults and XMP/DOCP. This is the way you get 'out of box' performance as everyone else would. What people tweak and such is impossible to get consistent.
I agree though the likes of gamer's Nexus prefer to force intel and AMD reference specs on board's.
Disabling MCE ,pbo etc, that's not how they're used.

All benching should be how they're used, All , everywhere (I'll adopt a say no more on this here in in Tpu)

I would however like. This information put into reviews too though, as a heads up so to speak.
Posted on Reply
#32
thesmokingman
As I wrote the problem is the lack of disclosure. If you wanna bury your head in the sand about this that's your prerogative. The fact is MB makers grease their boards all the time, running stuff out of spec just to look good on benchmarks. That is a LIE and it does the consumer a disservice.
Posted on Reply
Add your own comment
Nov 26th, 2024 06:33 EST change timezone

New Forum Posts

Popular Reviews

Controversial News Posts