Quote Originally Posted by aoch88 View Post
Are we suppose to set it so that it shows the temp we desire to see? LOL :d
Whatever makes users happy!

Personally, I don't think anyone needs to be touching the new TjMax adjustment feature but some of the 45nm sensors are so far gone that this might just help someone. This also saves me arguing with people over what TjMax is for their processor. If they don't like my choice, they can use their own.

NaeKuh: I had a look at your data before it mysteriously disappeared and even though it looked like a hopeless mess, I think RealTemp can be used with your cpu to get some reasonably accurate core temps out of it.

Most users get pretty suspicious the moment reported core temps go below ambient but the amount your readings were below your water temp / room temp was completely normal. The theory behind RealTemp is that the colder your processor gets and the farther you move away from TjMax, the larger the error is going to become in your readings. This error can go either way and without a user doing some testing there's no way I can look at a model number and predict what's going to happen.

With my E8400, when the Distance to TjMax (DTS) is at 55, there is already a 5C error between what an IR thermometer says and what RealTemp would be telling me with TjMax=95C and with no idle calibration. With CoreTemp using TjMax=105C, the reported error is 15C and growing as the temperature drops further.

When the DTS is reading 80 like on your processor, the size of this error can approach 10C. I'll admit that this isn't an exact science and I haven't tested enough processors to know exactly the average size or range of this error but at least I know it exists. Other core temp programs have been pretending that the DTS data is 100% linear but that's definitely not true at idle. I've included some code in RealTemp to compensate for this which will improve the accuracy of your idle core temperatures even if they'll never be 100% perfect. These sensors just weren't designed or calibrated for that.

From what I've seen so far, if TjMax is chosen correctly, low idle temps are only accurate +/- 10C and if TjMax is wrong, the error might approach 20C. Back up a page to yesterday and have a look at Dom7184's readings. His idle temps on a 45nm Quad Xeon are being reported at 50C by the competition's software which is completely wrong and he knows it.

When a processor is reading 10C too low at idle, some users will use CoreTemp which assumes TjMax=105C instead of 95C like RealTemp is using. That might cover up the problem with the DTS at idle and make the numbers look believable but it's not solving the problem. All you end up with are load temps that will be reported 10C too high. If a processor is already reading 10C too high at idle and you switch to CoreTemp and it adds on another 10C of error, your temps will be sky high and you'll probably start thinking that you must have got a bad cpu and Intel forgot to solder the IHS to the cores.

When you get your block attached correctly with enough thermal paste then do a 5 minute run of 1 minute idle, 3 minutes of Prime95 small FFTs on all cores and then back to 1 minute of idle so I can see what kind of data your DTS is giving out. Include that and your beginning water temp and room temp and PM it to me. Go into the RealTemp.ini file before you start and set LogInterval=1 so I'll have plenty of numbers to look at. I'm pretty sure that a slight calibration of RealTemp will get you some fairly accurate and believable temperatures from Idle to TjMax.