Help!... 1 Hot Core!

[Gimli43Orcs]

I have a Q9450 with a Thermalright Ultra-120 extreme heatsink. As I type this, one of my cores is at 64 degrees with the other 3 at 49 degrees. I have re-seated the heatsink, change the orientation of the heatsink, change the thermal compound and finally lapped the heatsink, with no changes. The only thing I haven't done is lap the processor, but I am hesistant to do it now just in case the processor is actually defective.

The only thing that confuses me is that when I run Prime95, the hot core only goes up 5 degrees, while the others go up 10-12. How can this be possible if the processor heat spreader not flat is the cause of the high temp? Or is that cores thermal sensor just whacked??

 

[myocardia]

Who cares? Idle temps mean absolutely nothing. As long as your load temps aren't 10 or 15 degrees different from each other, you're fine. BTW, your temps are too high for your heatsink, unless you aren't at stock speed. Which software are you using for temps, BTW?

 

[Gimli43Orcs]

Im using Core-Temp. Im at 3.2GHz. Don't know what else I can do to lower the temps. Im in the testing stage with extra fans blasting it on a open case.

 

[ShadowFlareX]

His room temp should be around 30oC / 86oF to 32oC / 90oF for an idle temp of 49oC... because mine is. Did you apply your thermal paste This way?

 

[myocardia]

Originally posted by: Gimli43Orcs
Im using Core-Temp.

CoreTemp reads 10C too high, with the 45nm chips.

Im at 3.2GHz.

At what Vcore?

Don't know what else I can do to lower the temps.

Lower Vcore always helps, as does applying the right amount of thermal paste, as ShadowFlareX alluded to.

 

[Gimli43Orcs]

Im at 1.4 Vcore right now. But like I said, Im in the testing stage, so that can hopefully go down a bit. Ive tried different amounts of thermal paste and using the various methods (a line,credit card spread, a drop ) with same result.

 

[Aisengard]

1.4 is borderline too high for 45nm parts. Lower it.

 

[aigomorla]

Originally posted by: myocardia

Originally posted by: Gimli43Orcs
Im using Core-Temp.

CoreTemp reads 10C too high, with the 45nm chips.

no it just reads wrong temps period. LOL...

if you want to take that rule, then when my comp idles im sitting in sub ambient which is impossible.

My guess with a TRUE is a always first a bad base. :T

 

[bdubyah]

get the newest version of realtemp. go find unclewebbs' thread at xtremesytems.org and d/l the newest beta he has put out. works well for 45nm chips.

 

[myocardia]

Originally posted by: aigomorla
My guess with a TRUE is a always first a bad base.

Very true. Some of them seem about as flat as corduroy.

 

[Syzygies]

1. Water cooling folks use inserts to change the bowing in cpu water blocks. Has anyone ever tried "orthotics" for air? The Intel thermal manuals all describe how to groove the IHS to place a test sensor; what if one put a groove past the edge of the hot core and played with wire thickness, to slightly torque down the IHS over the errant core?

2. AMD people take off the IHS to run naked, for some processors; here's a thread asking how for a Q6600. Turns out it's nearly impossible because Intel uses solder, but has anyone played with reflowing the solder? I have a digital soldering iron that will hold any specified temperature, if one knew the magic temperature one could play "processor acupuncture" touching the troublesome area.

3. Alternatively, drill a pair of tiny holes, inject a nonconductive TIM to fill the airspace under the IHS, seal the holes.

I'm sure both that some people could have a field day pillorying me for these suggestions, and that some people could come up with three better suggestions, one of which might actually get adopted down the road. After all, we've all surely seen the look on some people's faces when we describe lapping? It seems odd that we all take extraordinary measures above a flattened IHS surface, and suffer the consequences of mild chaos between that surface and the processor cores, like a dog that's been trained to stay in its yard.

 

[rge]

You clearly have one stuck/bad sensor. If the die attachment was the problem on the one core, then when the other three cores went up 10-12C, the problem core would also go up at least 10-12C (and probably more if the IHS was not properly attached). But since it only goes up 5C, ie starts to read more accurately at load, it is just another of the many "stuck" sensors, ie, it is very nonlinear and inaccurate at low temps. Intel has confirmed that all DTS are inaccurate and nonlinear at low temps, but some are off a few degrees at idle, some off as much as 20 at idle, and some have been completely non-functional, or will not report below a certain temp even when exposed to liquid nitrogen.

But if yours is reading accurate at load and thus functional, it is just going to be an annoyance at idle.

And if you get realtemp as was suggested, all 4 temps will read 10C lower, and probably closer to the real temps (but that wont fix your bad sensor)