Power Values (Core i7-5930K + Asus Rampage V Extreme)

[tistou77]

Hello
 

With a 5930K and Rampage V Extreme, is it possible to have the "real" Power Value for CPU Package?
In load, this value is about 2W

 

 

Thanks

[Fiery]

Can you please check if you experience the same issue with HWMonitor?

BTW, on our 5820K + Gigabyte X99-UD4 system AIDA64 measures cca. 10W at idle and 100W under heavy load. So the problem may be related to 5930K or Rampage V Extreme motherboard.

[tistou77]

With HWInfo, it's indicate 0.58xW (idle) and with HWMonitor it's the same

 

Thanks Fiery

[Fiery]

Please right-click on the bottom status bar of AIDA64 main window --> CPU Debug --> CPUID & MSR Dump. Copy-paste the full results into this topic, or attach the results as a TXT file to your post. We'll have to check the registers, and compare them against our 5820K. Maybe there's a way to fix it up by applying a workaround.

Thanks,

Fiery

[tistou77]

The report

 

cpuidmsrdump.txt

 

Thanks for your great work

[Praz]

Hello Fiery

 

There is nothing wrong with AIDA64 CPU/DRAM power reporting. If SVID support is disable in the motherboard's UEFI power reporting is broken. If enabled the power/current is properly displayed.

[tistou77]

Yes, that's what I've seen on the ROG forum, I will test it while ago to see.

[tistou77]

It's ok with SVID Support enabled

Damages whatsoever related, it's better to disabled it for the OC
With SVID enabled, this indicated 117W in stress with AIDA64 (not the total power I guess)

[Fiery]

Thank you. I suppose when SVID support is disabled, the power circuitry doesn't go through the standard path where the processor could measure its own power draw. I'm afraid we have no idea how to get around that limitation Intel's CPU documents do not feature enough details to fully understand how the power circuitry and its various modes work exactly

[tistou77]

Too bad, it is not serious

 

Thanks Fiery

[tistou77]

Hello

 

With HWMonitor, there is the uncore in addition for the Power

 

 

With AIDA64, there are not that info. This is the same value as the Power Package or it's "missing"?

 

 

Thanks

[Fiery]

I'm pretty sure that value is just calculated as:

Uncore = Package - DIMM

To me that doesn't sound like a proper formula for your CPU...

[tistou77]

With DRAM SVID enabled

 

 

Thanks

 

 

[Fiery]

No, I don't think you understood what I meant What I meant to say is that Uncore is not Package minus DRAM. Your CPU package includes the following components:

- Uncore

- CPU cores (6 cores)

DRAM power is supposed to be the total power draw of the DRAM modules, but the DRAM modules are not part of the CPU package at all. But the six CPU cores are part of the CPU package, but those are not accounted there at all. So the formula used by HWMonitor is completely wrong and pointless.

[tistou77]

Ok, thanks so much for your explanation

[tistou77]

Hello

 

With my 5930K OC to 4600mhz and vcore to 1.25v, according to calculation, to consumption will be at ~ 260W
With Stress of AIDA64, it's indicate ~ 180W.
Because AIDA64 does not stress "fully" the CPU (AVX 2 instruction)?

 

Thanks

[Fiery]

With my 5930K OC to 4600mhz and vcore to 1.25v, according to calculation, to consumption will be at ~ 260W

What kind of calculation would lead to 260W in such circumstances? Certainly Intel never published any formula to do that. And if you're not going by Intel's own formula, then the result will not be comparable to the Intel CPU power measurement wattage.

 

Because AIDA64 does not stress "fully" the CPU (AVX 2 instruction)?

AIDA64 System Stability Test (SST) -- when you use only the FPU subtest -- puts an enermous stress on the CPU, and makes it draw very high power. I wouldn't state that no other software or trick could make your CPU draw even more power, but our solution (the AVX2 and FMA accelerated FPU subtest of the SST) is considered a "power virus" by Intel, so it's definitely good at driving your CPU to its limits

[tistou77]

What kind of calculation would lead to 260W in such circumstances? Certainly Intel never published any formula to do that. And if you're not going by Intel's own formula, then the result will not be comparable to the Intel CPU power measurement wattage.

 

Calculation is

 

Dissipation oc = Dissipation stock * (Fréquency oc / Fréquency stock) * (Vcore oc / Vcore stock)²

 

AIDA64 System Stability Test (SST) -- when you use only the FPU subtest -- puts an enermous stress on the CPU, and makes it draw very high power. I wouldn't state that no other software or trick could make your CPU draw even more power, but our solution (the AVX2 and FMA accelerated FPU subtest of the SST) is considered a "power virus" by Intel, so it's definitely good at driving your CPU to its limits

 

 

Thanks for the explanation

[Fiery]

Calculation is

 

Dissipation oc = Dissipation stock * (Fréquency oc / Fréquency stock) * (Vcore oc / Vcore stock)²

That formula isn't quite scientific For example, your CPU has multiple clock planes and multiple voltage planes. Not to mention Turbo Boost, throttling and other factors that could ruin the validity of that formula. And also, how do you know the stock dissipation of your CPU? The value Intel publishes as TDP has nothing to do with the stock power draw of the CPU. It is only an informational value for motherboard manufacturers to help them design their motherboard PCB. Also, TDP is an average value, under regular usage scenario. It's not a peak value at all, not measured under heavy load. A 130W TDP CPU could easily draw over 160W, even at stock settings, if you put it under heavy stress (FPU subtest of AIDA64 System Stability Test for example). Of course, in order to measure such relatively accurate values, you would need a lot more than relying on the dissipation that the CPU reports for itself

[tistou77]

Ok thanks for this explanation

 

And how Aida reads this value?
I get some information, and with ~180W of consumption, it would like to say that my vcore with a OC will be the same as vcore at "stock"

Even if the formula is "approximate" (different condition) 90W of difference, I find is a lot

 

Thanks

[Fiery]

AIDA64 uses the standard method Intel provides to measure energy consumption of the CPU. It is never intended to be used by 3rd party applications. It was designed for Turbo Boost originally, to make sure the CPU knows how much energy it draws, to make sure it stays between the pre-configured TDP and TDC levels. But just like with the CPU core diode temperature measurement (which was designed to provide protection against overheating, and not to measure absolute temperature), Intel CPU owners demanded various software like AIDA64 to measure and show those readings. Even though in many cases the readings will not reflect anything useful

[tistou77]

Ok thank you, I understand better.
And when Turbo Boost is disabled (for OC) is the same?

[Fiery]

Ok thank you, I understand better.

And when Turbo Boost is disabled (for OC) is the same?

Only Intel knows what changes about power measurement when you disable Turbo Boost. Even without Turbo Boost the CPU must stay inside the power envelope, so I suppose the power measurement should still work as before.