Different temps on different cores

[Ken g6]

I just built my first Core 2 Quad machine, a Q9400, in a 6-year-old case. The case has good airflow, but doesn't have one of those ducts just for the CPU. I had a heck of a time getting the stock cooler on. (I'm sure someone will say using the stock cooler is wrong, but let's ignore that for now. ) I accidentally lost some of the thermal grease from the top of the cooler. And, despite my best efforts, the top 2 pins went in better than the bottom 2. I think I almost bruised my hand getting the last pin in.

But it's working now, underclocked to 2 GHz for now because I wasn't sure how well I got the cooler on, or how well it would work. I finally got lm-sensors running in Linux, started a Prime95 torture test, and here are my current CPU temps:

coretemp-isa-0000
Adapter: ISA adapter
Core 0: +55.0C (high = +72.0C, crit = +100.0C)

coretemp-isa-0001
Adapter: ISA adapter
Core 1: +51.0C (high = +72.0C, crit = +100.0C)

coretemp-isa-0002
Adapter: ISA adapter
Core 2: +53.0C (high = +72.0C, crit = +100.0C)

coretemp-isa-0003
Adapter: ISA adapter
Core 3: +45.0C (high = +72.0C, crit = +100.0C)

As you can see, there's some asymmetry. Is this normal? Is this caused by the fans' positions (a case fan is on one side, and the power supply fan is directly above it. Or did I put the heat sink on wrong? I'd really rather not re-seat it!

Thanks!

 

[TridenT]

It could be a misreading...

In Realtemp I get like 38C idle on Core 0 and 31C on Core 1. They both register like 65C during load though...

 

[PCTC2]

It could be that one of the pins is not actually in. That would lead to a differential in temperatures. Also, normal usage can cause different core temps. Right now, mine are 45,44,39,39 running iTunes and Chrome.

Why not pick up an aftermarket cooler that bolts down? Or use a bolt-down kit on your stock cooler?

 

[Ken g6]

I put the thing on before I put the motherboard in. All the pins were definitely in; although that last one was really hard to get in, all four black pins were down as far as the end of the white pins.

As for why not a better cooler or attachment...I'm cheap, I'm trying to be energy efficient so I thought the temps would be low enough, and I had no idea how bad those pins would be! :Q It looks like even mid 60's at load are OK for Core2s, and I'm in the 40s-50's (even after ramping the clock a little), so I think I'll stick with what I've got.

Thanks!

 

[Beanie46]

First, I continue to be amazed so many people have any sort of difficulty getting the push pins in with the Intel coolers....one of the easiest heatsink mounting sytems and so many have difficulties with them. Baffles me completely.

Anyway, you did have the black "pins" turned the correct way, right? I've seen too many people not understand which way they have to be oriented to get them to "click" into the motherboard correctly or easily. I'd try to explain how they go....and I'm sure you already know.....but I'm tired so here are some pics of how they should be positioned.....


Pic of Intel hs ready to be installed....notice how the black tops and their lines are oriented.

Close up of how the black pins/tops are supposed to be turned/oriented for installation.

Close up of black pins when they're to be uninstalled....the arrow denotes how to turn the black pin to release the heatsink's mounting pins from the motherboard.


If yours are OK, and they probably are, either you lost enough thermal paste in one spot to make a difference, or you may just have one sensor in your cpu gone bonkers, as some have been and are. Early in the C2D/C2Q production, sensors were known to be stuck at one temp....no fault in the cpu otherwise, just temp reading stuck at one temp. Guess that could effect overheat shutdown, but otherwise didn't impact the cpu's function at all.

Personally, if installed correctly, I'd not worry about it. I've had quads show up to 7-8C differences between the two cores while idle or loaded. Luckily, right now, my Q9550 is almost perfect with only a 1-2C difference between the two cores. Rare.

 

[God Mode]

Originally posted by: Beanie46
First, I continue to be amazed so many people have any sort of difficulty getting the push pins in with the Intel coolers....one of the easiest heatsink mounting sytems and so many have difficulties with them. Baffles me completely.

Anyway, you did have the black "pins" turned the correct way, right? I've seen too many people not understand which way they have to be oriented to get them to "click" into the motherboard correctly or easily. I'd try to explain how they go....and I'm sure you already know.....but I'm tired so here are some pics of how they should be positioned.....


Pic of Intel hs ready to be installed....notice how the black tops and their lines are oriented.

Close up of how the black pins/tops are supposed to be turned/oriented for installation.

Close up of black pins when they're to be uninstalled....the arrow denotes how to turn the black pin to release the heatsink's mounting pins from the motherboard.


If yours are OK, and they probably are, either you lost enough thermal paste in one spot to make a difference, or you may just have one sensor in your cpu gone bonkers, as some have been and are. Early in the C2D/C2Q production, sensors were known to be stuck at one temp....no fault in the cpu otherwise, just temp reading stuck at one temp. Guess that could effect overheat shutdown, but otherwise didn't impact the cpu's function at all.

Personally, if installed correctly, I'd not worry about it. I've had quads show up to 7-8C differences between the two cores while idle or loaded. Luckily, right now, my Q9550 is almost perfect with only a 1-2C difference between the two cores. Rare.

I think most of the difficulty or reluctance stems from hearing the dreaded cracking sound and seeing your motherboard bend like rubber. It's understandable to see people apprehensive on installing push pin heatsinks due to the amount of force necessary. I have no problem with them but its a very anti consumer design. IMO, AMD's clip mechanism superior and leads to less second guessing whether the heatsink is correctly installed.

 

[BTRY B 529th FA BN]

I had a Q9300 that did that.

 

[error8]

Originally posted by: God Mode
IMO, AMD's clip mechanism superior and leads to less second guessing whether the heatsink is correctly installed.

Yes AMD's design is better from this perspective, since you know it has been secured. The problem is that the clips from the stock cooler, are generating an impressive force on to the little two plastic teeth and they are prone to break. This has happened to me. But it seems that it' a very rare problem though.

 

[Ken g6]

Yes, the clips were oriented correctly. The part I lost some grease from was the top. Then I had trouble pushing the bottom-right pin in - nerds like us shouldn't be expected to produce that kind of force without a simple machine! Also, the bottom-left pin somehow got itself in the wrong orientation during installation and popped out. So I eventually got them all in right, but I think all the grease got oriented downward.

Anyway, it's still working fine. I'm getting 7-9C difference between the cores at load, less at idle; plus I've undervolted for now so that the overall ("thermal diode") temp is only 50C at 2.66 GHz.

Thanks!

 

[BonzaiDuck]

Originally posted by: Beanie46
First, I continue to be amazed so many people have any sort of difficulty getting the push pins in with the Intel coolers....one of the easiest heatsink mounting sytems and so many have difficulties with them. Baffles me completely.

Anyway, you did have the black "pins" turned the correct way, right? I've seen too many people not understand which way they have to be oriented to get them to "click" into the motherboard correctly or easily. I'd try to explain how they go....and I'm sure you already know.....but I'm tired so here are some pics of how they should be positioned.....


Pic of Intel hs ready to be installed....notice how the black tops and their lines are oriented.

Close up of how the black pins/tops are supposed to be turned/oriented for installation.

Close up of black pins when they're to be uninstalled....the arrow denotes how to turn the black pin to release the heatsink's mounting pins from the motherboard.


If yours are OK, and they probably are, either you lost enough thermal paste in one spot to make a difference, or you may just have one sensor in your cpu gone bonkers, as some have been and are. Early in the C2D/C2Q production, sensors were known to be stuck at one temp....no fault in the cpu otherwise, just temp reading stuck at one temp. Guess that could effect overheat shutdown, but otherwise didn't impact the cpu's function at all.

Personally, if installed correctly, I'd not worry about it. I've had quads show up to 7-8C differences between the two cores while idle or loaded. Luckily, right now, my Q9550 is almost perfect with only a 1-2C difference between the two cores. Rare.

Personally, even in the "long comet-tail of the B*** Economy," after working with coolers that have spring-loaded screws and backplates, I wouldn't hesitate to spend up to a half-Franklin-note and toss the stock cooler. I hate the push-pins. And for some few machines I've built for others over the last couple years for which I just used the stock-cooler, there were no mishaps. I just hate the pushpin setup.

There are a lot of reasons for discrepancies in temperatures on these Yorkfield cores. Many of them shipped with defective sensors, or the core sensors built into the cores were not intended (per Intel's disclaimer) to measure idle temperatures accurately. The legacy TCase sensor is probably more accurate than the core sensors. I've even discovered -- in my E8600 E0 CPU that TCase will exceed the core sensors when it is well established that TCase should lag behind the readings from the cores.

The other possibility is that the IHS is uneven. You can lap it flat and down to bare copper, although it voids the warranty, and it won't shorten the real expected lifespan of the processor.

But mostly . . . . .I HATE those PUSH-PIN coolers . . . .