difference in temp,

[cantix]

i get at normal, 49 and 56 in Distance to TJ Max, with temps at 51 and 44 respectively...
Calibrated i still get 49 and 56 at Distance to TJ Max with temps of 47 on both cores....
That is CPU running at 3Ghz with C1E disabled and Speed Step Disabled too.

 

[unclewebb]

Calibration doesn't change Distance to TjMax. That is a direct reading from the Intel CPU. Calibration simply changes the interpretation of that raw data into an absolute temperature number that users are more comfortable viewing. Did you do a full calibration as outlined in the RealTemp documentation?

 

[cantix]

yeah but i didn't pay attention to the TjMax, now that i do i see there is no change, so I'm going to be concerend about TjMax instead of Current Temp.

 

[wolf2009]

I get the picture, but in my case

Wouldn't it mean that I'm 50C away from tj max at idle ,and at 100% I am less than this, so the load temp is more accurate ? That would make my CPU be at 67 C ?

 

[cantix]

you guys think i should enable C1E? and Speed Step?

 

[burebista]

Why not? Do you need full power when you browsing/watch movies/listen music/other light stuff?
This is my idle frequency/VCore and CPU power consumption.
And I'm perfect fine browsing/whatever light stuff.

 

[unclewebb]

Because these sensors do not move at the exact same rate that the temperature changes (linearly) then technically you can't say that you are 50C away from TjMax. That's why I call this value a Distance to TjMax. It shows you how many units you are away from TjMax but those units may or may not be exactly equal to how many degrees you are away from TjMax. I'm sure that sounds a little confusing or just plain dumb but that's the way it is.

In theory, these sensors are calibrated by Intel somewhere near TjMax. They also left out in their recent presentation the exact temperature where they calibrate at so I can only guess. As these sensors get closer in temperature to the calibration point, they tend to line up better with each other. On a Quad, core0 and core1 will become equal and core2,core3 will be equal. The 2 separate Core Duo chips within a Quad don't usually line up with each other even at TjMax.

The Intel presentation shows that the slope error starts as soon as you leave the calibration point and increases linearly as you move farther away. There is also some unknown error at the calibration point. This results in a floating temperature curve with neither end of the line perfectly defined so getting some honest temperatures out of these things in more like a guessing game.

If you do some calibration, I'd like to think that RealTemp is a reasonably accurate guess.

burebista: If your Asus software is accurate and your CPU is only using 4 watts, then my assumption that these CPUs aren't that hot at idle seems pretty accurate, especially when you are using a large air cooler or a good water cooling set up.

 

[cantix]

the lowest i get is 1.14v hehehe, guess i'll do it and try it out. let me see....

 

[cantix]

I have enabled Speed Step and C1E but my OS apparently doesn't lower rated BuS speed or FSB, is that something thats done by TM1 & TM2?

 

[unclewebb]

The front side bus shouldn't change but the multiplier should drop to 6.0 when you are idle which reduces the total MHz. Most bioses let you manually set and lock the multiplier at 6.0.

 

[cantix]

hhmmm, well i get a speed of 2ghz at idle with 1.12v

 

[burebista]

burebista: If your Asus software is accurate and your CPU is only using 4 watts, then my assumption that these CPUs aren't that hot at idle seems pretty accurate, especially when you are using a large air cooler or a good water cooling set up.

Uncle I guess it's pretty accurate. Why? Because those guys measured Wolfdales power consumption in idle and in full load and it is very low.

During our experiments we measured the current going though the processor voltage regulator circuitry. This way we could estimate the actual CPU power consumption disregarding the losses in the voltage regulator circuitry.

Hope it helps you.


cantix My MB have EPU software and I can lower FSB and Vcore more than VID+Vdroop that's why I have that frequency and VCore and that's why I choose this MB because I'm a silent freak and I love to play with frequencies, Vcore and fan speeds.

 

[cantix]

my config is the following
C2D 8400 with Thermanlright Black Ultra/Fan Antec TriCool @2krpm(only 3xdB)
4Gb(1gb x 4) Corsair XMS DDR2800
a 7950GT with 512MB(old card which has never died on me lol)
and a WD 750 Caviar SE Black Edition (the one with 32MB of Buffer)
Case is a Cosmos with exhaust fans S-Flex back and S-Flex at the top and front mount cage with another antec TriCool fan.

First Gen Cosmos are known for not having very good cooling so I'm struggling with getting the best fans in the market with high CFM and low dB.

 

[cantix]

oh and the MB is an XFX 780i 3-way SLi.

 

[burebista]

And your MB?
This is how it looks my config (in System Specs)

I'm pretty happy noise/performance wise.

Uncle this is my consumption in load with a decent OC.

 

[cantix]

i'll see if i can get some pics uploaded ^_^, i really liked that case u have, if they would've made it a little bigger I'd buy it. but i'm going for a LianLi one or the new one from NZXT

 

[cantix]

ok, here are some pics,

the whole Illuminating Dominator Ram Fan I did it hehe, fans are rated at 8krpm

 

[unclewebb]

hhmmm, well i get a speed of 2ghz at idle with 1.12v

At default settings, your FSB should be about 333.3 MHz and your multi when C1E is enabled will be 6.0 so your total should be about 2000 MHz. With my motherboard I can manually lower the FSB to 266 MHz so my total is about 1600 MHz. In terms of heat output, there is virtually no difference between the two settings.

Thanks for that power chart burebista. I think Tom's Hardware did a similar chart a long time ago where they bypassed everything so they could accurately measure just the CPU wattage. When you put a big air cooler on something like that it's no surprise that the core temperature is only about 5C above the room temperature.

I noticed that the new E0 processors have the ability to enable the C4 power state which previously was reserved for the mobile processors. This should lower power consumption another watt at idle if parts of the processor are going into a deeper snooze mode. The general 65W Intel spec for most Core 2 Duo processors looks pretty conservative.

The Asus software looks like it is showing some realistic power consumption numbers. I have my E8400 in today so I might have to set my MHz and core voltage similar to yours and see what my Kill-a-Watt meter shows for overall system power consumption. The total difference should be a few more watts since it will include extra watts for the chipset, etc.

Edit: My Kill-a-Watt meter shows 127 watts at idle and 165 watts at load running Prime95 small FFTs for a difference of 38 watts. Here are my settings:

LinPack creates more heat than Prime small FFTs so I guess that makes sense. My E8400 isn't a happy camper without some voltage but it ran long enough to get some power usage numbers. Time to go back to 4 GHz with some voltage to keep it

 

[burebista]

I noticed that the new E0 processors have the ability to enable the C4 power state which previously was reserved for the mobile processors.

Aaa, I guess my E8400 C0 did this too with Intel C-State Tech enabled in BIOS. Remember those odd frequencies I have?

LinPack creates more heat than Prime small FFTs so I guess that makes sense.

Yep, this is how it looks power consumption in Orthos

10W less than Linpack.

 

[unclewebb]

Why not switch Orthos to small FFTs. That creates more heat than blend so our watts should be about the same.

 

[burebista]

2W more. Still a long way till Linpack.

 

[unclewebb]

Here's some math burebista to make our results match.

Your CPU when running small FFTs consumes 34.50 watts and 4.25 watts at idle so the difference is 30.25 watts

My Kill-a-Watt meter measures total power usage at the plug and during small FFTs reports a difference of about 38 watts. My power supply is only about 80% efficient so 38 watts at the wall equals 38 * 0.80 or 30.40 watts is being consumed by the CPU. I'm not sure if my math is 100% accurate but it's close enough for me.

 

[burebista]

I'm not sure if my math is 100% accurate but it's close enough for me.

And for me too.

Anyway is not 100% accurate because my EPU software shows only CPU consumption and your kill-a-watt device shows entire PC consumption.
So in your case you should substract at least MB and memory consumption which interfere with your CPU consumption.

 

[unclewebb]

I thought about that too but in both situations, idle or running Prime, the motherboard, graphics card, memory, hard drives, fans, etc. are running. I think the difference the Kill-a-Watt shows represents overall CPU power consumption pretty well after you include the power supply efficiency factor of about 80%.