hello, I have an intel i5 2400 processor and I've noticed something that quite a few people
told me was a bit odd. While monitoring my core temperature using an application like Real Temp
I've noticed that my core 0,1, and 2 are close to the same temperature as one another except for
core 3 that is always about 10c or so cooler even during full load(most of the time core 0 is always
hotter, but it isn't while using prime95). I've used the sensor movement test
on Real temp and I notcied that core 3 is ALWAYS far apart from the other three cores. For
example, one movement test done while idle is core 0 = 13, core 1 = 14, core 2 =14, and core 3 = 7
Do you think this indicates a stuck sensor? I've already reapplied thermal paste and even replaced
my heatsink yet core 3 is still always about 10c cooler. Do you think i should be worried or need to
do an RMA for this processor?
Also note that what makes things even more weird is that when I do a movement sensor while running prime95 or under about 80% load, the movement sensor results for the cores go down to 2s or 3s or even all 0s on all 4 cores. Is this normal or should i worry? My temps while running prime 95 stay around 55-60c. The cpu was never even tampered with or removed since the day it was put in the socket so I have no clue why I might have a stuck sensor. Could there be a problem with the thermal paste dripping or something? because I can't really think of anything else that might have harmed it.
The thing is, I've replaced the thermal paste and even the heatsinks multiple times and already made sure of that yet I always get the same thing. At 100% load the first 3 cores are sort of equal, but core 3 is always way off like 10c cooler.
For example, Idle for me right now is 36-31-30-26. Under load the first 3 cores start to catch up to each other, but core 3 still decides to stay behind. I know most people would say that every cpu is different and some cores can be a lot hotter or cooler than others, but how the hell am I suppose to know whether my sensors are picking up the correct temperatures or not?
55-60C while running Prime95 is OK. The fact that one of the 4 cores reports a much lower temp is not a concern as the rest of the cores are adequately cooled.
You can always call Intel and get an official recommendation... they are surprisingly friendly most of the time. Pretty sure they will tell you not to worry, but they'll swap it if you like.
Somewhere I think I read that the maximum acceptable deviation in the sensors was 10C, but I see about 7C quite often and would not have freaked out at 10C as long as the other cores showed a good temp.
Last month i purchased an intel i3 3220 from amazon, and i notice a difference of 15-30c between both cores at idle, core0 doesn't go below 33c even on a very cold night with 3 case fans, hyper 212+ and arctic silver 5....and at the same time, core1 stays at 3c
I read that 5-10c of difference between both cores is normal....so for me, that indicates that something is wrong, so i decide to contact amazon today and ask for a replacement and in less than 6 hours i received an email with the tracking number of the replacement CPU and the instructions for me to return the other one.
I'm not saying you that your cpu is defective...i'm just sharing my experience so you can decide if your cpu is fine or not
Well, my cpu has been doing what I said it is doing for a long time now, but I haven't been experiencing any throttling or anything like that. To tell you the truth I wish it would even give me a sign that something could be wrong at all so I could finally rest my case and RMA the cpu instead of RMAing it and going through the pain of RMAing just to find out that nothing is wrong with it or if something bad happens to it during shipping. My cpu never really goes above 12c difference(talking about core 3 only since the other 3 are always near each other) so i'm still a little concerned. Btw I mentioned that the core movement for all my cores always stays at 0 while running prime95 at 100%, is that normal or could that be a bad sign? I'm also trying to develop an understanding of this random number generator that they call a "sensor test" on real temp so I would appreciate it if someone explains the mechanics behind it and how it works.
Intel engineered and calibrated these sensors to manage thermal throttling at approximately 100C and to control thermal shutdown at approximately 125C. At the thermal throttling point, these sensors are only accurate to +/- 5C. Why do users keep thinking that these sensors should be 100% accurate at reporting idle temperatures? They are not even close to that, especially at idle.
Any reported temperature less than about 40C should also be taken with a grain of salt. The 45nm Core 2 sensors used to be horrible and would often times get stuck at a fixed value. Core i sensors are much better but as accio_fawkes discovered, some of them have this same problem too. Big deal. As long as your CPU runs reliably, there is no need to be concerned about its core temperature.
The RealTemp sensor test is only supposed to be run when your CPU is completely idle. If you can sit back for 10 minutes and let the test run without using your computer, it will vary the load on each core so you should see the temperatures vary as well. On the 45nm Core 2 CPUs, this test made it very easy to see what the lower temperature was before a sensor would get stuck. It was also a good way to see the slope error these sensors have. Slope error is when the data coming from the sensors changes values at a different rate compared to the change in actual temperature. It's not that bad of a test if you know how to interpret the results. I am sure it helped encourage Intel to start using some better temperature sensors. After the 45nm Core 2 temperature sensor fiasco, the first generation Core i CPUs like the Core i7-920 had some of Intel's best sensors.
Hard to fault Intel for being cheap. When you make a product by the millions, if you can save a few pennies on each one by using a cheaper temperature sensor, why wouldn't you?