Some cores are hotter than the others - reasons?

[metadist]

Greetings all,
I have noticed that on my i7-8700k soe cores are noticably and consistently hotter. For example, below, cores 1,3,5 are hottern that 2,4,6

        Core [core-id]  :Actual Freq (Mult.)      C0%   Halt(C1)%  C3 %   C6 %  Temp      VCore
        Core 1 [0]:       4692.69 (46.98x)      99.5       0       0       0    79      1.3440
        Core 2 [1]:       4692.68 (46.98x)      99.5       0       0       0    71      1.3440
        Core 3 [2]:       4692.69 (46.98x)      99.5       0       0       0    80      1.3440
        Core 4 [3]:       4692.70 (46.98x)      99.5       0       0       0    75      1.3440
        Core 5 [4]:       4692.72 (46.98x)      99.5       0       0       0    79      1.3440
        Core 6 [5]:       4692.75 (46.98x)      99.5       0       0       0    69      1.3440

I wonder what can be the reason? Is it possibly because there are air gaps in some spots left after I applied liquid metal while delidding?

I also has an idea that this might be due to three cores are just next to the memory controller on the die, as seen on the picture. But I can not verify ifthey are indeed the same core I see running hot.



Any other ideas from the community?

Cheers!

[Rob_W_]

I had cores running hot, just two or three of them by about ten degrees above the others.
Same thing, thought it might be the delid so redid it and still same.
Seems it’s quite a common thing, it can be sorted by going the per core route if it is always the same cores getting hot, lowering the voltage slightly on the offending cores.
I think it’s down to that silicone lottery pain.:D*

[metadist]

Thanks for sharing your experience! Might save me re-lidding troubles them..
Other thing - I don't have an option to set different voltages to different cores on my Prime z370-A...

[XYchromosone]

metadist wrote:

Other thing - I don't have an option to set different voltages to different cores on my Prime z370-A...

I am not aware of any motherboard that allows per core voltage control if you use a manual voltage, possible the "Adaptive" or "Offset voltage" control would control a per core voltage, though.

However a possible option is to set a per core multiplier and have the core that run hotter set to a lower multiple. E.g., let's say if cores 4 and 5 are always showing that these 2 cores run much hotter than the rest of the cores when you have a voltage set manually, you can set a per core multiplier of :

Core 0 x 48
Core 1 x 48
Core 2 x 48
Core 3 x 45
Core 4 x 45
Core 5 x 48

This is just an example, but you might need to experiment with this including core voltage. You could see how a manual voltage of let's say v1.25 to v1.30 affects those 2 cores in a benchmark. Then try an Adaptive voltage setting etc. and see how which setting affects the cores temps in a benchmark under load.

As and end note, a build I did last year with a Maximus VIII Ranger mobo running an i7 6700 non K CPU one of the 4 cores ran about 20 deg hotter than the 3 other cores. This was pretty well at all times, especially when a little load was put on the CPU. I was tempted to do a delid just to satisfy my curiosity, but since it wasn't a K model CPU I decided to not go ahead with it. SO whether it was due to an uneven layer of heatsink compound on the CPU die or something else that was causing this heat differential with a single core, I will never know.

It goes to show that it can and it does happen. The thing that matters to me is does it affect the lifespan of that core, which effects the lifespan of the cpu, OR the work it can do under load when power throttling cuts in due to overheating that core, lowering the overall score in a benchmark test etc.

[geneo]

Variations in the accuracy of the heat sensors among other things.

[Silent_Scone]

As has been said already, core temperature variance is normal, on some platforms up to 20c is not uncommon.

[jologskyblues]

In my experience with the 8700K I got, the temperature variance between cores was +-10C before delidding. After delidding, the variance went down to about 4~6C.

[XYchromosone]

@jologskyblues I take it from what you say is that there was/is an AVERAGE temp variance between all 6 cores, which is normal. There will [always] be variation in temperature between all the cores in a multi-core CPU because the load the OS puts on the CPU varies between each core. I'm guessing you already know that.

However, this thread is about just one or two cores always running hotter by a good margin, than the rest of the cores, e.g. in my case with an i7 6700 one core ran hotter even at idle under no load by a good margin. This is definitely an indication that something is not 100% right with CPU either at the silicon level or something to do with the thermal paste intel uses or its application [air pockets left between the die and CPU lid e.g.].

But I do understand how the delid helped to reduce the temperature of the CPU as you explained, which would effect the temp variation % between all the cores to be less, I'd think.

[jologskyblues]

@XYchromosome

To clarify my statement, the temperature variance I was mentioning is specifically the difference in temperature between the hottest core and the coolest core when fully stressed in Realbench, x264, Prime95, etc.