Temp variations (11 degress) across cores during stress testing i4770k

[flexy]

Looking around I didn't find a conclusive answer.

The obvious one would be it's a problem with the heatsink seating or thermalpaste application, say if the heatsink would tilt downwards from the weight one could actually expect that temps would show this pattern.

(I used MX-4 paste, a small strip about 1" long down the IHS because this is the shape of the die. I am not a noob when it comes to this stuff

But some others say that those variations up to 10 deg are normal.

Checking my sink (Thermalright Ultra 120 Extreme), wiggling it etc. doesn't show any obvious problem, it seems fine to me.

So right now I am not sure what's going on.

(Temps at idle are equal across cores, the variations show only up wduring stress-testing. There is a *clear* pattern with the cores appearing cooler with the last one showing a variation 11deg to the first one.)

Edit: Those are overclocked temps of course

Edit2: Another explanation would be that this problem (if it is one) is because of bad TIM under the IHS. In other words: This CPU would need delidding and some coolliquid. (Not that I would like to delid if I really don't have to)

Edit3: Ok so doing some more research it looks that variations 8-10 degrees are indeed rather common, especially in the high temps range. I also see a pattern with many benchmarkers showing that the last core is normally always the coolest one.

 

[ClockHound]

Doesn't sound terribly abnormal...there's the dark silicon effect.

And the esteemed IDC mentioned this asymmetric behavior before.

Of course, he also advocated taking a sharp razor, a hammer and a vice to the delicate chip and cutting the heat spreader off, in the interests of 'science'...possibly to shuffle the cores around or something.

 

[Lepton87]

Actually I think that you have very little variation in the temperature of the cores. My CPU has a soldered IHS and I still have more way more variation. My coldest core is 63 while hottest is 83.

 

[KaRLiToS]

This is completely normal.

 

[rgallant]

I have one sensor at idle shows 17c , 25-32-17-26 water is around 20c
it catches up under load so it's not defective.
I would not take any of the temp readings as real unless their at the shut down temp. that they were designed for.

intel sensor offsets in software have been used for years to get a ball park/best guess then dislay that into numbers.

I would say just add them up divide by 4 and be happy.