Is this power reading right (Sandy Bridge)?

[tokie]

I never noticed this before, but I am assuming this reading is being drawn from some sensor inside the CPU. Is this correct?

Is this fairly accurate for others here? It seems amazing that at full load my 2500k only uses ~65 watts.

 

[BoomerD]

That's definitely odd. Mine shows 108.33 all the way across.

 

[QuackPot]

Doesn't Sandy Bridge have some sort of power saving mode?

 

[PreferLinux]

Not ~65 W, ~90 W.

 

[Castiel]

When used with a P67 board the IGP is disabled bringing TDP down to around 80W. It's pretty normal

 

[tokie]

That's definitely odd. Mine shows 108.33 all the way across.

Well, since the columns are current, min, and max I would say something is definitely wrong with yours.

I just can't believe that a fully loaded 2500k only uses 65 watts... would this power figure be coming from some diode on the CPU itself?

 

[PreferLinux]

Well, since the columns are current, min, and max I would say something is definitely wrong with yours.

Ahh.. OK, I'm not familiar with that program, and I couldn't see the labels!

 

[Vinwiesel]

I was also seeing about 6W at idle with stock settings with the original bios on a p8p67 pro. After flashing 1253 my idle power is reported as 9.6W, but I believe this bios raised the idle voltage offset as well. What bios are you running?

 

[Dufus]

I never noticed this before, but I am assuming this reading is being drawn from some sensor inside the CPU. Is this correct?

Hard to say. With my P55-UD6 HWMonitor (1.17 IIRC) had 2 power readings. One was derived from the SIO and the other was calculated using the TDP. The first method said my i7 860 was pulling 1000W at load and the second method would be a bad guess if the TDP limit was changeable. With a i5 655k Gigabyte set TDP at 300W so HWMonitor would calculate I was using that much at 100% load. Likewise setting TDP at 1W it would calculate from 0.01W to a whopping 1W. lol

Best if you can physically verify. There are so many different configurations out there it's not surprising results aren't always right.

With SB there are energy registers that can be read from the CPU however AFAIK these seem tied to VID so if your overriding stock Vcore it might be off a fair bit. I posted some power levels here for stock settings using the CPU package energy sensor and measuring power into the VRM but really need to follow up with further testing some time.

 

[coffeejunkee]

Hard to say. With my P55-UD6 HWMonitor (1.17 IIRC) had 2 power readings. One was derived from the SIO and the other was calculated using the TDP. The first method said my i7 860 was pulling 1000W at load and the second method would be a bad guess if the TDP limit was changeable. With a i5 655k Gigabyte set TDP at 300W so HWMonitor would calculate I was using that much at 100% load. Likewise setting TDP at 1W it would calculate from 0.01W to a whopping 1W. lol

Best if you can physically verify. There are so many different configurations out there it's not surprising results aren't always right.

With SB there are energy registers that can be read from the CPU however AFAIK these seem tied to VID so if your overriding stock Vcore it might be off a fair bit. I posted some power levels here for stock settings using the CPU package energy sensor and measuring power into the VRM but really need to follow up with further testing some time.

I have an i5 750 + Gigabyte mobo and like you, in HWmonitor I have 2 values for cpu power consumption. One is under Gigabyte DES > powers > cpu and one is under the normal cpu entry (just like the screenshot in the opening post).

Unlike you, I have found the DES readout to appear quite believable. I know it's a very coarse method, but when I subtract idle power from load power as indicated by my voltcraft powerreader and factor in a bit of hypothetical vrm efficiency it seems to match up reasonably well. The values make senses as well (about 87W during Linx testing and about 60W gaming, stock) and also scale up correctly when oc-ing (higher oc means higher draw).

The value under the cpu entry indeed seems incorrect. It often stays at 95W or doesn't make sense at all (fex. lower draw with higher oc/vcore). So based on that, I would say TS readout shouldn't be trusted. But based on my DES readout I would say it might be correct. The numbers don't seem too unrealistic to me.

missed this one:

I was also seeing about 6W at idle with stock settings with the original bios on a p8p67 pro. After flashing 1253 my idle power is reported as 9.6W, but I believe this bios raised the idle voltage offset as well. What bios are you running?

9,6W idle is the same minimum value HWmonitor shows for my cpu. As stated, that readout is very likely to be incorrect, certainly when overclocking. Please try some different overclocks and see what the readout says.

My DES power sensor shows a lower value when idle: 2,5W. I know it sounds low but as I've understood, modern core chips almost completely turn themself off during idle.