Baffling AMD X2 Temp issue

[Slinky]

Hello Everyone,

 

I'm a long time OCC reader, but first time poster.

 

Anyway, I recently bought an AMD X2 5200+ processor for my server, and it's been going great, except there is an odd temperature issue I'm coming across.

 

Here is the server temps under no load:

 

 

As you can see, Cool'n'Quiet is working well, but there is a massive 10 degree difference between the cores. Now, I know what you're thinking, heatsink isn't seated properly, let the thermal paste set, etc, etc (BTW, I have let the paste set, haven't gotten round to reseating the heatsink). I thought the same things, until I put the server under load:

 

 

Now, the cores are much closer in temp, but the previously cooler core is now hotter by 3 degrees.

 

I was a bit perplexed at these results, as I kinda expected that they wouldn't:

a) have a smaller difference in T between the cores, and

b) the cooler core (by 10 degrees!) ending up 3 degrees hotter

 

I thought the difference between the cool core and the hot core would have gotten even bigger, but.... obviously not.

 

Any ideas why this is happening, or how I can sort this out?

[damian]

That's normal. Sometimes the temperature difference across all cores get smaller when the cpu builds up more load.

Edited May 3, 2009 by damian

[Slinky]

But why the original 10 degree difference? I'm used to a couple of degree digressions between cores, but this one just seems excessive.

[nVidia_Freak]

Nevermind.

 

Reading fail on my part.

Edited May 3, 2009 by nVidia_Freak

[cold_snipe]

I know when i was reading up on intel's tjmax it was common for there to be a larger error when temperatures were low as opposed to load temps. just because of the design, which aims to be more accurate at higher temps. (i provide no evidence, but that what i read somewhere)

[upok]

Pretty common actually. Could have something to do with load dispersion. At idle, or near idle, processor load is left on one core because there is no need to disperse the load. This would lead to one core being always somewhat hotter than the other. When load is applied, dispersion kicks in and core temps get much closer because load levels are closer.

 

Could also be that the IHS is making better contact with one core than the other. The IHS may not be perfectly flat, it may have some consistency irregularities that cause it to conduct heat slightly better on one side than the other, etc. These kinds of things have been out of our control since the days of "Naked" cpus!

 

I wouldn't worry about it. All of the temps are acceptable for a stock cooler. I've had a couple of chips like that. One got closer and closer in temp as time went on. The rest just always did that. My current rig has 2 cores that run within 3 degrees of each other but 5-8 degrees hotter than the other 2 cores(which run within a couple degrees of each other).