So I had a water cooled system built into an Antec 1200, absolutely cool and quite. Average core idle at 29 stress at 40ish with a mild overclock. I had not run speedfan for a long time but after reintsinstalling my system and reaplying thermal grease I ran it just to check, now its 40 isle and goes up to 85 under load using prime 95 CPU under speedfan goes up to 51 and that is with no overclock!!!
I checked the cooling system and even kept my hand on the tubes and copper cooler and they feel normal, so my initial thoughts was that it is malfunctioning sensors.
Sure enough running Real Temp's sensor test I get the following readings:
That means the sensors are busted right? Thing is my processor does not include a waranty or anything so I cant return it, should I just ignore the values? is there something I need to turn off in my BIOS to make sure it is not throttled down due to inaccurate readings?
Those RealTemp sensor test numbers look okay to me (my Q9450 has 2 mostly stuck sensors with test results of 7-5-0-0, though Intel says they will function properly the closer to the maximum temp that they get).
You have either an interface problem of some kind between the CPU and the cooling block, or the coolant is not properly flowing through the system (trapped air, low fluid level, debris blockage, pump motor pumping too slow, etc), or a problem with the radiator fan (if so equipped).
There's really only one thing that can go wrong in a water cooling unit and that's the pump. Very rarely will the waterblocks scale over or debris block a tube. If you want to get an idea if the unit is working put your hand on the cpu block, if it is feeling hot then there is a flow rate problem. Sometimes the binding agent or mechanism on the turbine shaft works itself loose.
Also sensors get stuck on the way down, that is if you load up your cpu and the temps climb then they will do what Intel intended, however if they don't all settle to around the minimum tempreature, then they are stuck on the way down. People have "unstucked" them by running prime or long renders over periods of 5+ hours, sometimes people have had systems running 48 hours and they sensors still remain stuck, good luck.
When you run RealTemp, and use that sensor test, did you read what it says at the very bottom?
The core temps are part of a process to activate throttling or thermal shutdown. Since there are other programs that can get core temps, all those programs out there share one common critical thing, which is known as the tjunction max. And since Intel doesn't provide what the max is, the people that made the software have to guess/assume what it is (95C/100C/105C) and that is why you have different readings from different programs. In order to get somewhat of a more accurate tjmax setting, you would need to calibrate it using an external temp monitor.
Once you done that, any program that allows an offset, or to change the tjmax, you can adjust.
Back to what RealTemp says at the very bottom after the test is done... It basically states that Intel does not recommend it to be used to report idle temps. And that would be the issue of the sensors being stuck.
I normally go by the Tcase temp, which is the sensor between the cores or IHS temp. That sensor is being used by the bios, since it reports a single temp, and also your MB utility app, and SpeedFan should report the Tcase sensor that the bios uses. Now Intel does publish the thermal spec, which would be the max temp on the Tcase sensor, that you should not exceed.
You don't want that pump the way it is, I suggest you find out why you need to whack it to get it to work and if it's unrepairable, get a new one. Knowing murphy's law, it will die right when you don't need it to.