RealTemp General Discussion

[unclewebb]

Unfortunately, this list is not accurate, especially for the 65nm processors. Intel has already started making corrections to this list but sadly, even their corrections don't agree with what an IR thermometer says.

A good example is my original B2 stepping E6400. The chart at Tom's says a B2 is TJMax=70C. When DTS=0 for that processor the IHS is at 85C. That implies that the actual core temperature is approximately 90C. It's definitely not 70C and it's not Intel's corrected version of 80C either.

At the moment these numbers are meaningless. I'm obviously a little biased here but until Intel publishes some test data, which they don't plan to do, I'll be sticking with values that actually make sense.

 

[a_ump]

well said and i'll trust anyone that's actually tested it, especially u, over intel anyday.

 

[a_ump]

well this is what i got from the sensor test, i just read how to do it and didn't really know what to look for on whether it was good or bad

 

[unclewebb]

a_ump: The CPU Cool Down Test is brand new and I'm still learning to interpret this new data.

As the CPU Load decreases from 100% to Idle, you should see a continuous drop in core temperatures. Your Average Distance to TJMax should continue to increase at each step of this test. That's exactly what your sensors are doing. Some 45nm CPUs get about half way through this test and suddenly a sensor will stop moving and it will report the same thing 4 or 5 times in a row. That's a good sign that a sensor is stuck and not working properly. The Sensor Movement Test will also show a low number or zero if a sensor is stuck at low temperatures.

Your Q6600 is performing very normally. At idle the temperatures are very similar and at full load the difference grows to 2C or 3C between the two sets of cores. A Quad Core contains 2 Dual Core CPUs and your results make it pretty clear that core0/core1 are joined together and core2/core3 are joined together. In almost every Q6600 I've seen, the first two cores will be reported at a slightly higher temperature than core2/core3 at full load.

Your Q6600 looks like a good one to me and should give you fairly accurate core temperatures.

 

[a_ump]

thanks man, and it's nice to see a program author that actually talks and helps out his hmmm consumers? well keep it up man

 

[modder]

hi , an example with three Stuck DTS sensors on my Q9450

 

[unclewebb]

If I was handing out awards modder you would get it for the CPU with the worst sensors. Most stuck sensors show some movement when running Prime 95 Small FFTs but 3 out of 4 of yours sensors are completely dead. I guess there's a reason why Intel doesn't recommend we use these things to report core temperatures.

There's no use running any sort of calibration factor on core1, core2 or core3. Core0 moves freely from idle to TJMax so when calibrated you should have at least one sensor that gives you a reasonably accurate core temperature.

 

[modder]

If I was handing out awards modder you would get it for the CPU with the worst sensors. Most stuck sensors show some movement when running Prime 95 Small FFTs but 3 out of 4 of yours sensors are completely dead. I guess there's a reason why Intel doesn't recommend we use these things to report core temperatures.

There's no use running any sort of calibration factor on core1, core2 or core3. Core0 moves freely from idle to TJMax so when calibrated you should have at least one sensor that gives you a reasonably accurate core temperature.

thanks for the award prize .yes ,the core 0 work fine,other work from :core 1 41C° ,core 2, 54C° , core 3 46C°.realtemp with Tcase it would be nice to compare Tjunction and Tcase .

 

[burebista]

If you really want to see them moving, OC a little with bumping Vcore a notch or two and run LinX.
I bet you'll see all core sensors moving.

 

[modder]

If you really want to see them moving, OC a little with bumping Vcore a notch or two and run LinX.
I bet you'll see all core sensors moving.

thanks for advice , i especially no overclocked my Q9450 for showing my bad sensors.

 

[burebista]

You can try, I see that Core0 sensor shows a very good distance to TJMax, but IMO Core1 sensor is lost forever.
I must admit that those are the worst sensors that I saw on any forums that I read (and trust me, I read a lot of forums ).

 

[thesmotis]

hi,
will realtemp 2.70 run fine on a server with 2 xeon quad processors and windows xp professional x64?
also will it always try run on the background or will i be able to run it and stop it whenever i need to?
thank you for your time

 

[unclewebb]

RealTemp doesn't yet support Dual Quad processors.

I've been working on that and have an option available in the latest beta version available here:

http://www.fileden.com/files/2008/3/3/1794507/RealTempBeta.zip

If you go into the RealTemp.INI file and add:

Skull=1

to read CPU1

or you can use

Skull=2

to try and read CPU2.

I don't know if this works properly yet because I don't have a computer with Dual Xeons to test on. If you want to help get RealTemp working for you then contact me at:

Real_Temp@yahoo.ca

and I'll see what I can do.

In the meantime, I'm pretty sure that Core Temp will work with your system.
http://www.alcpu.com/CoreTemp/

 

[thesmotis]

...

i was afraid it wouldn't support 2 cpus, i'll check coretemp and see how it works
anyway, thank you very much for the detailed response, i wish you good luck with developing the program further

for mp cpu support, maybe you could check with these guys:
forums.2cpu.com
in the past the forum has worked for mp support with other developers as well

 

[mlee49]

UncleWebb,

I've got a question. I run this on my e8200 and my cores are off like 10+ degrees. Core 0 runs at 32(average) and Core 1 runs at 44(average). The difference is always like +9 degrees even at load for a good about of time.

Are my cores bad? Is one of my cores bad? Perhaps is the sensor bad?

I'm just curious since this is an ES and not a fully supported stepping. Thanks

 

[unclewebb]

mlee49: Sounds like you probably have a crappy sensor or two. That's pretty typical in the 45nm CPUs.

The latest beta of RealTemp has a new feature that makes it possible to understand what your sensors are really up to and can spot a sticking sensor a lot easier as well. Scroll back to post #88 where I posted a link to it.

 

[PsySc0rpi0n]

I have read some posts here and in ohter foruns and you say to set the idle calibraton factor to 5ºC above room temperature. But i have no way to measure my room's temperature...

I'm from Portugal and it's cold but i believe that my room's temperature is somewhere between 20ºC and 25ºC because i have temperature control system in my house that keeps temperatures "worm" but not to much...

So i have no idea on how to calibrate my RealTemp...

At the moment i have set vCore to 1.30v in Bios (1.27v in Windows) and i have FSB@400Mhz (3.6Ghz) in a E8400 E0 and when testing Sensors with RealTemp i get 0 movement in both cores. I have set -5ºC in Idle Calibration for testing only...


By PsySc0rpi0nBaY


Edit;

Well, after searching for a while i've found a thermometer (lousy one) and stick it in the wall in my bedroom... It's retrieving 17ºC but i don't know if i can trust on it so i'll make it for 20ºC (for safety)....

So as unclewebb said, i'll set my P5K Pro Bios to 6x on multi, 1.00v of vCore and FSB at 266Mhz then i'll do the math to calculate my possible idle temp...

So if RealTemp shows 40ºC in idle state with this settings i'll set -15ºC for both Core0 and Core1 in Idle Calibration! Is this correct?

And TJmax of 100ºC for E8400 is already correct in RealTemp 2.90, right?????


Edit1;

So i have set vCore to 1.0000v in BIOS but it is not the lower value possible, FSB at 267Mhz and multi 6x. Here is a screen of RealTemp, CPU-z



Temps for both cores are 40ºC. So the formula to calculate the closest cpu real temp is Room Temp + 5ºC, right? So my Room temp is about 18ºC~20ºC so my CPU Temp will be about ~24ºC, correct?

So in Idle Calibration i have to set it with -16ºC for both cores wich is the actual RealTemp temperature value (40ºC) minus estimated cpu real temp (24ºC), am i thinkig correct?


Edit3;

Well, i think i'm not correct... If i set -16ºC in Idle Calibration, RealTemp will retrieve 16ºC of Temperature and i thought it would retrieve 24ºC...

To retrieve 24ºC i have to set idle calibration to -11ºC but i don't know the meaning of this... I don't know if i can trust my thought...

Please give a help on how to set RealTemp correctly...

 

[unclewebb]

Unfortunately both your cores have sticking sensors. They are both stuck at a Distance to TJMax of 60 so it is impossible with these sensors to get accurate idle temperatures. With TJMax set to 100C, you will have reasonably accurate core temperatures when your CPU is at 40C and above but below that there is nothing you can do with a stuck sensor. Calibration for your CPU is not possible. Not all sensors are capable of working at low temperatures. Intel says we shouldn't expect anything below about 50C.

 

[PsySc0rpi0n]

Unfortunately both your cores have sticking sensors. They are both stuck at a Distance to TJMax of 60 so it is impossible with these sensors to get accurate idle temperatures. With TJMax set to 100C, you will have reasonably accurate core temperatures when your CPU is at 40C and above but below that there is nothing you can do with a stuck sensor. Calibration for your CPU is not possible. Not all sensors are capable of working at low temperatures. Intel says we shouldn't expect anything below about 50C.

So if TJ Max is set at 100ºC and i set no Idle calibration and temps goes over 40ºC, in this case RealTemp is able to retrieve correct temps, right? At this time RealTemp is retrieving 54ºC for me with 1.33v at 4.0ghz... Is it too much? I think i can't go under this vCore to set the cpu at 4.0Ghz and FSB at 500Mhz and multi at 8...

 

[rampage]

if you temps are under 40c you wont get accurate temps, but if its over you will get accurate temps, so yes the 54c will be correct, 54c is on the hot side for me, buts its nothing that will destroy a cpu

 

[PsySc0rpi0n]

And the boards are nothing to do with this issue, right? Changing the board won't change anything, right?

 

[unclewebb]

People worry too much about temperatures.

Here's how my E8400 looks when overclocked with the CPU fan turned off for 3 hours while running Prime Small FFTs to create some heat:

Intel hasn't been specific but admits that the temperature sensors have errors where they calibrate them so TJMax is not an exact number. There can also be as much as plus or minus 10% slope error. If you can't do something to check the calibration of these sensors, then even at 54C, there's no guarantee that number is 100% accurate. It's usually reasonably close though.

I've done a lot of testing and I've learned one important thing. As long as your CPU is Prime stable and you are below the thermal throttling point (Distance to TJMax = 2) then you can completely ignore the temperature your CPU is running at. It will run fine. Intel has built a lot of technology into these things and they do a great job of looking after themselves even when grossly overclocked with additional core voltage.

Changing the motherboard or the bios shouldn't change anything because these sensors are within the CPU.

 

[burebista]

Intel has built a lot of technology into these things and they do a great job of looking after themselves even when grossly overclocked with additional core voltage.

Agree. I have a guy in our forums with a bad mount of stock cooler. He have something like this when playing games.



and the only downside was stuttering when CPU is throttling.

As long as your CPU is Prime stable and you are below the thermal throttling point (Distance to TJMax = 2) then you can completely ignore the temperature your CPU is running at.

Man you can put above quote in your signature.

 

[unclewebb]

Man you can put above quote in your signature.

Shhhh! That's a secret.
I can't let too many people know just how unimportant core temperatures are or else I'll be out of business.

 

[PsySc0rpi0n]

So i guess, i don't know why, my DTS are live...

After increasing a little the vCore 'till arround 1.44v i have some movement for both sensors...


By PsySc0rpi0nBaY

But what does means this movement values??? Higher values are better or worst?

And please help me understanding the idle calibration proccess in RealTemp....

You say:
1-Downclowck cpu 'till it's minimum (FSB-266; Muli-6; vCore - 1.00000v~1.10000v)
2-Watch reported temps - if they are lower than room temp then insert positive value, if not, insert negative values in Idle Calibration...

So in my situation, my room temperature is 20ºC (~25ºC for apprx real idle temp in cpu), RealTemp (without any calibration) report 40ºC in idle state, waht is the factor that i sould use to set the Idle Calibration correctly???