CORE 2 DUO RMA BECAUSE OF HEAT ISSUES?

[serializable]

My temps aren't that high but they aren't even. At idle core 1 is at 34C but core 2 is at 28C. At load core 1 goes to low 40s and core 2 goes to mid 30s. There is usually a 6 degree difference between the two. I'm not overclocking and Im using a zalman heatsink with as. Im using core temp to monitor the temperatures. Looking at other peoples temperatures it doesn't seem normal to me that there is such a difference between the two. Does anybody know why this is?

 

[augiem]

I'm not gettting all this noise about concave heatsinks and having to lap these cpu's. It's almost like Rollo was here defending his paid-for brand with his wacky brand of misinformation.

The absolute highest I have been able to bring my slightly overclocked E6400 under extreme load was 50C and even after that the temp dropped to around 48C. I guess some guys know how to install heatsink/fans and some don't?

Please don't assume that because you don't have a problem everyone else is imagining things. I don't have cancer, but that doesn't mean cancer doesn't exist.

I was very shocked to see that link to Intel's spec sheets on the C2D says 60.1°C is the highest reliable temp. It isn't really that hard to reach for some of us.


Augie

 

[phile]

Originally posted by: augiem
I was very shocked to see that link to Intel's spec sheets on the C2D says 60.1°C is the highest reliable temp. It isn't really that hard to reach for some of us.

The big question is this: are the readings from the core DTSs (Digital Thermal Sensors), as read by CoreTemp, Everest, and the Intel Thermal Analysis Tool, reliable? We're all trying to ensure that we do not exceed Intel's Thermal Specification of 60.1, but then we have this tool that ostensibly tells us the core temps, using some sort of deltaT calculation based on TCaseMax - whatever the hell that is. ^%$^#$!!! Other than that, we have the mobo's CPU temp sensor, which is located nowhere near the hottest part of the CPU core.

As you can tell, I even find it difficult to properly articulate what I think is an unneccesarily complicated and confusing situation, regarding conroe temperature measurement.

-phil

 

[HyperVista]

For what it's worth, I just built my C2D EE system this weekend. Once I got the bios on my P5W DH Deluxe upgraded, I got the system running. The thrill of seeing the system boot was short lived when I noticed temps pushing 80C!!!

To get the bios updated, I had to remove the C2D EE processor, insert an older Pentium 4 D, upgrad the bios, and reinsert the C2D EE. I think the issue was the multiple re-seating the HS and resulting "thinning" of the HS material (sorry, I'm new to cooling issues and don't know the correct terminology - that silver goo). Anyway, once I noticed that the old Pentium 4 I used to boot to get the updated bios had a new HS exactly like the one the new C2D EE had and the silver goo on that one was pristine, I carefully mounted the new HS and temps have been down in the mid-to-high 30s at idle.

I know you guys are much more experienced at these things than I am, but this was my experience. That HS and silver goo absolutely needs to be seated correctly (and I was convinced I had it seated properly, but evidently not).

 

[phile]

Originally posted by: HyperVista
For what it's worth, I just built my C2D EE system this weekend. Once I got the bios on my P5W DH Deluxe upgraded, I got the system running. The thrill of seeing the system boot was short lived when I noticed temps pushing 80C!!!

To get the bios updated, I had to remove the C2D EE processor, insert an older Pentium 4 D, upgrad the bios, and reinsert the C2D EE. I think the issue was the multiple re-seating the HS and resulting "thinning" of the HS material (sorry, I'm new to cooling issues and don't know the correct terminology - that silver goo). Anyway, once I noticed that the old Pentium 4 I used to boot to get the updated bios had a new HS exactly like the one the new C2D EE had and the silver goo on that one was pristine, I carefully mounted the new HS and temps have been down in the mid-to-high 30s at idle.

I know you guys are much more experienced at these things than I am, but this was my experience. That HS and silver goo absolutely needs to be seated correctly (and I was convinced I had it seated properly, but evidently not).

Yes, proper seating of the HSF is crucial, and the installation of these stock Intel LGA775 coolers is very easy to screw-up. In any case, it's a good idea to get yourself a good aftermarket cooler. There are plenty that will do a superior job to the stock cooler. I'm using the Zalman CNPS9500 LED, which has the added benefit of being quite sexy.

BTW, the proper term for that silver goo is "Stock Thermal Crap". "Stock Thermal Gunk" is also acceptable.

-phil

 

[hags2k]

Originally posted by: phile

Originally posted by: augiem
I was very shocked to see that link to Intel's spec sheets on the C2D says 60.1°C is the highest reliable temp. It isn't really that hard to reach for some of us.

The big question is this: are the readings from the core DTSs (Digital Thermal Sensors), as read by CoreTemp, Everest, and the Intel Thermal Analysis Tool, reliable? We're all trying to ensure that we do not exceed Intel's Thermal Specification of 60.1, but then we have this tool that ostensibly tells us the core temps, using some sort of deltaT calculation based on TCaseMax - whatever the hell that is. ^%$^#$!!! Other than that, we have the mobo's CPU temp sensor, which is located nowhere near the hottest part of the CPU core.

As you can tell, I even find it difficult to properly articulate what I think is an unneccesarily complicated and confusing situation, regarding conroe temperature measurement.

-phil

As I understand it, there is no "motherboard" sensor for the cpu (with most of these boards, at least), and the bios, speedfan, and coretemp/TAT all get thier info from the same place (on die thermal sensor). Some claim that no LGA775 boards have separate sensors (since there's no socket to stick it in). I read that the only big difference between bios/speedfan and coretemp/tat is that one only reports the temp reading for one core, whereas the other programs report temps for both cores. On my system, the readings were never mroe than 1 degree apart between the cores, so that probably doesn't make much difference.

So, if that's true, that means that they're calibrated very differently. Personally, I'm more inclined to go with the BIOS readings, but perhaps that's just because they're quite a bit lower than what TAT/coretemp report

Also, when going quickly from idle to load, the programs respond at almost identical speeds, maintaining a more or less constant temp difference between them. I've been told that if you're measuring the temp from outside the cpu, there should be more "lag" between increasing the load and seeing an increased temp reading.

So, does this make sense, or is there something big I'm missing or been misinformed about?

 

[phile]

Originally posted by: hags2k

Originally posted by: phile

Originally posted by: augiem
I was very shocked to see that link to Intel's spec sheets on the C2D says 60.1°C is the highest reliable temp. It isn't really that hard to reach for some of us.

The big question is this: are the readings from the core DTSs (Digital Thermal Sensors), as read by CoreTemp, Everest, and the Intel Thermal Analysis Tool, reliable? We're all trying to ensure that we do not exceed Intel's Thermal Specification of 60.1, but then we have this tool that ostensibly tells us the core temps, using some sort of deltaT calculation based on TCaseMax - whatever the hell that is. ^%$^#$!!! Other than that, we have the mobo's CPU temp sensor, which is located nowhere near the hottest part of the CPU core.

As you can tell, I even find it difficult to properly articulate what I think is an unneccesarily complicated and confusing situation, regarding conroe temperature measurement.

-phil

As I understand it, there is no "motherboard" sensor for the cpu (with most of these boards, at least), and the bios, speedfan, and coretemp/TAT all get thier info from the same place (on die thermal sensor). Some claim that no LGA775 boards have separate sensors (since there's no socket to stick it in). I read that the only big difference between bios/speedfan and coretemp/tat is that one only reports the temp reading for one core, whereas the other programs report temps for both cores. On my system, the readings were never mroe than 1 degree apart between the cores, so that probably doesn't make much difference.

So, if that's true, that means that they're calibrated very differently. Personally, I'm more inclined to go with the BIOS readings, but perhaps that's just because they're quite a bit lower than what TAT/coretemp report

Also, when going quickly from idle to load, the programs respond at almost identical speeds, maintaining a more or less constant temp difference between them. I've been told that if you're measuring the temp from outside the cpu, there should be more "lag" between increasing the load and seeing an increased temp reading.

So, does this make sense, or is there something big I'm missing or been misinformed about?

I'm convinced this is not the case. Probe 2, speedfan, and everest all give me the same reading as my BIOS, while CoreTemp and Intel's Thermal Analysis Tool give me core readings that are always 15C-20C higher. If what you're saying were the case, how do we explain such a discrepancy?

Here are my temps.

IDLE:
CPU (Probe 2, speedfan): 30C
CPU Core1 (CoreTemp, TAT): 48C
CPU Core2 (CoreTemp, TAT): 50C

LOAD:
CPU (Probe 2, speedfan): 39C
CPU Core1 (CoreTemp, TAT): 64C
CPU Core2 (CoreTemp, TAT): 65C

-phil

 

[hags2k]

Originally posted by: phile

I'm convinced this is not the case. Probe 2, speedfan, and everest all give me the same reading as my BIOS, while CoreTemp and Intel's Thermal Analysis Tool give me core readings that are always 15C-20C higher. If what you're saying were the case, how do we explain such a discrepancy?

Here are my temps.

IDLE:
CPU (Probe 2, speedfan): 30C
CPU Core1 (CoreTemp, TAT): 48C
CPU Core2 (CoreTemp, TAT): 50C

LOAD:
CPU (Probe 2, speedfan): 39C
CPU Core1 (CoreTemp, TAT): 64C
CPU Core2 (CoreTemp, TAT): 65C

-phil

I think perhaps this varies from board to board as well. First, I have to ask, are you sure you're using the right reading from speedfan? I know it wasn't immediately obvious which one i should use. I'm almost certain that you are (based on your level of expertise), but I wanted to mention it on the off chance.

Second, I think the chipset used might make a difference. Mine is 965P-based.

Third, on my board the readings are:

CPU idle:

Bios/Speedfan: 34C

CPU Load:

Bios/Speedfan: 50C

Which is a 16 degree deltaT, as opposed to your 9 degree deltaT from speedfan. That deltaT, though, is right in line with the deltaT you get from coretemp. The deltas are equal in both programs on my system. It's only the min/max values that are different. That's why I think that, on my system at least, bios/speedfan are getting data from the same place.

EDIT: by the way, the difference is always 8-9 degrees between speedfan/bios and coretemp/tat. If you want my coretemp/tat values, you can just add 8-9 degrees to the speedfan values. The fact that this difference is amost perfectly constant tends to suggest to me a calibration difference more than anything.

 

[phile]

Originally posted by: hags2k

Originally posted by: phile

I'm convinced this is not the case. Probe 2, speedfan, and everest all give me the same reading as my BIOS, while CoreTemp and Intel's Thermal Analysis Tool give me core readings that are always 15C-20C higher. If what you're saying were the case, how do we explain such a discrepancy?

Here are my temps.

IDLE:
CPU (Probe 2, speedfan): 30C
CPU Core1 (CoreTemp, TAT): 48C
CPU Core2 (CoreTemp, TAT): 50C

LOAD:
CPU (Probe 2, speedfan): 39C
CPU Core1 (CoreTemp, TAT): 64C
CPU Core2 (CoreTemp, TAT): 65C

-phil

I think perhaps this varies from board to board as well. First, I have to ask, are you sure you're using the right reading from speedfan? I know it wasn't immediately obvious which one i should use. I'm almost certain that you are (based on your level of expertise), but I wanted to mention it on the off chance.

Second, I think the chipset used might make a difference. Mine is 965P-based.

Third, on my board the readings are:

CPU idle:

Bios/Speedfan: 34C

CPU Load:

Bios/Speedfan: 50C

Which is a 16 degree deltaT, as opposed to your 6 degree deltaT from speedfan. That deltaT, though, is right in line with the deltaT you get from coretemp. The deltas are equal in both programs on my system. It's only the min/max values that are different. That's why I think that, on my system at least, bios/speedfan are getting data from the same place.

It certainly does appear that your mobo has the capacity to plug into the core temps, while mine has a separate sensor. That's the only explanation that I can think of, that explains our significantly different situations. What bothers me is that my mobo CPU temps strike me as low, while my core temps strike me as high. I'm now very tempted to buy a P5B Deluxe P965. Which board/conroe are you using? And, if you're overclocking, what's your FSB/VCORE?

-phil

 

[hags2k]

Originally posted by: phile

It certainly does appear that your mobo has the capacity to plug into the core temps, while mine has a separate sensor. That's the only explanation that I can think of, that explains our significantly different situations. What bothers me is that my mobo CPU temps strike me as low, while my core temps strike me as high. I'm now very tempted to buy a P5B Deluxe P965. Which board/conroe are you using? And, if you're overclocking, what's your FSB/VCORE?

-phil

I'm using a gigabyte P965-DS3 motherboard. My current settings are

FSB:285 (got crappy DDR2-533 RAM with locked 2x multiplier, so that's about the best I can do without buying better RAM)

Vcore set to 1.325 manually (the intel default setting) though it usually reads slightly below that

CPU: E6600 at 2.53 Ghz (based on the meager FSB settings, but it works for me)

Also, the readings I get at these settings are not measurably different from what I get at stock settings. That's to be expected, since it's only overclocked by about 5 percent :-D

 

[phile]

You know, at that FSB, you could probably drop your VCORE quite a bit. On my P5WDH/E6600, my rig is rock solid at 1.225v. All of my temps are approx. 10C lower dropping from 1.35v to 1.225v.

-phil

 

[StrangerGuy]

My 3.03GHz E6300 (everything at stock voltage) with Big Typhoon - 55C on idle, 68C on load. Last time with the stock HSF the load temps went all the way up to 85C, yet strangely enough it doesn't throttle or gave any stability issues.

 

[hags2k]

Originally posted by: phile
You know, at that FSB, you could probably drop your VCORE quite a bit. On my P5WDH/E6600, my rig is rock solid at 1.225v. All of my temps are approx. 10C lower dropping from 1.35v to 1.225v.

-phil

You know, I might give that a try. Couldn't hurt anything if I got it running even COOLER.

Also, from what I've been reading, the TCaseMax is the metric that coretemp/tat and some other programs use to figure out what the thermal diode is saying. All temps are relative to TCaseMax, so if CoreTemp says the temp is 45 degrees, it means that the diode is reading 40 degrees below the TCaseMax value. If 85C is NOT what that value actually is (and 85 seems rather high), then all these temps would actually be LOWER. I'm thinking that 75 or 70 is a much more reasonable value, as this is the point where you're approaching the throttling or maximum junction temperature.

The good news is, though, that if coretemp/tat is reading 65C , you're still 20C away from the throttling temperature, which should be (barring a major design flaw) the point where the chip becomes unstable, not the point where it will actually damage itself. It's just a matter of what that TCaseMax temperature ACTUALLY is.

It's possible my motherboard (which is reading the diode but calibrated to a different max temperature, I'm assuming) is actually calibrated to a more sane maximum, and therefore giving a more accurate read. As I said, the values and graphs of temperature are paralell between speedfan and coretemp/tat, just one is always 8-9 degrees higher than the other.

I think your temps, phile, are definitely somewhere between what coretemp says and that 39C that speedtemp gives you.

For comparison, the TCaseMax of many high-end AMD processors is between 60C and 75C, so 85C seems really high for a relatively high-efficency processor like the C2D.

 

[phile]

hags,

That makes a tremendoous amount of sense. I was wondering what exactly TCaseMax represented, as it's a good 25C above Intel's Thermal Specification for the C2D chips - 60.1C. Regardless, you're explanation is useful. I'm almost tempted to boost my vcore to observe the point where the CPU throttles, just to take all these measurement, and see how they compare.

On this topic, the following link was posted in a thread on alt.comp.periphs.mainboard.asus:
http://www.tomshardware.com/2006/07/14/core2_duo_knocks_out_athlon_64/page7.html

Core 2 processors carry four temperature sensors (two per core). These are digital now, which also means that they are considerably smaller than the previous, analog sensor generation. The DTS sensors thus are much closer to the hot spots now, and they can track temperature quicker and more accurate.

The same poster also claimed that the C2D chips have yet another thermal diode that is accessed by motherboards, and is the source of the BIOS and Probe II measurements. I'd like to get some sort of confirmation on that.

-phil

 

[augiem]

The temperatures given by bios and various programs vary wildly. That's given, and I think I can understand it. But it seems to me that Vcore voltages also vary from program to program, and are always different from the BIOS reading.

My DFI Infinity 975X/G bios reported voltage is almost always higher than software reported voltages. I have no idea which is accurate.

 

[serializable]

Heres a pic of my temps using core temp http://img469.imageshack.us/my.php?image=cpuvc1.jpg I've reseated the heatsink many times but Im not sure why there is a variance. The cpu looks flat and I put a small dab of artic silver right in the middle.

 

[hags2k]

Originally posted by: serializable
Heres a pic of my temps using core temp http://img469.imageshack.us/my.php?image=cpuvc1.jpg I've reseated the heatsink many times but Im not sure why there is a variance. The cpu looks flat and I put a small dab of artic silver right in the middle.

Anything that's running a single thread might conceivably make one core hotter than the other. This happens from time to time on my system. It's just that most progams people use to benchmark or stability test thier C2D systems are multi-threaded so that the temps are both very close to one another.

Anyway, a reading of 36C from coretemp is damn good, especially since it seems coretemp is seemingly giving higher than actual values. With temps like that, I wouldn't worry one bit.

 

[augiem]

My coretemp scores are also very similar to the scores from the mobo monitoring progs. I wonder if this vast discrepency is only with certain mobos?

BTW, my 2 cores vary are about the same as serializable's. I would guess it's normal.

 

[cmdrdredd]

Originally posted by: augiem
My coretemp scores are also very similar to the scores from the mobo monitoring progs. I wonder if this vast discrepency is only with certain mobos?

BTW, my 2 cores vary are about the same as serializable's. I would guess it's normal.

Someone I talked to locally says it's possible that the diode that the motherboard is reading is right, but the motherboard is interpreting the information wrong and thus giving bad info. It's possible that the mobo starts at say 10c instead of 0c and automatically reports everything 10c higher than it should.

Although I agree it's possible nobody knows for sure.

 

[hags2k]

Originally posted by: cmdrdredd

Originally posted by: augiem
My coretemp scores are also very similar to the scores from the mobo monitoring progs. I wonder if this vast discrepency is only with certain mobos?

BTW, my 2 cores vary are about the same as serializable's. I would guess it's normal.

Someone I talked to locally says it's possible that the diode that the motherboard is reading is right, but the motherboard is interpreting the information wrong and thus giving bad info. It's possible that the mobo starts at say 10c instead of 0c and automatically reports everything 10c higher than it should.

Although I agree it's possible nobody knows for sure.

I'm pretty sure that for those of us (like me) whose coretemp/tat readings and bios/mobo readings are always exactly a certain number of degrees apart (but always the same deltaT values) that this is the case. The description by the author of coretemp even specifies that all readings are relative to TCaseMax, and that the digital thermal diodes on the cpu only specify the DELTA between current temp and TCaseMax, so if those values are different, then you have a different calibration but still correct relative values.

Some motherboards, though, appear to read from a different sensor altogether.

 

[serializable]

Originally posted by: hags2k
Anything that's running a single thread might conceivably make one core hotter than the other. This happens from time to time on my system. It's just that most progams people use to benchmark or stability test thier C2D systems are multi-threaded so that the temps are both very close to one another.

Anyway, a reading of 36C from coretemp is damn good, especially since it seems coretemp is seemingly giving higher than actual values. With temps like that, I wouldn't worry one bit.

That was at idle. Even after prime testing for over an hour there is still a 4 degree difference. I'm not too worried though becuase like you said its still a low temp and that was with room temperature of 80F. http://img466.imageshack.us/img466/3949/cpurl2.jpg

 

[hags2k]

Originally posted by: serializable

Originally posted by: hags2k
Anything that's running a single thread might conceivably make one core hotter than the other. This happens from time to time on my system. It's just that most progams people use to benchmark or stability test thier C2D systems are multi-threaded so that the temps are both very close to one another.

Anyway, a reading of 36C from coretemp is damn good, especially since it seems coretemp is seemingly giving higher than actual values. With temps like that, I wouldn't worry one bit.

That was at idle. Even after prime testing for over an hour there is still a 4 degree difference. I'm not too worried though becuase like you said its still a low temp and that was with room temperature of 80F. http://img466.imageshack.us/img466/3949/cpurl2.jpg

Very true. Some of the people here I think would sell thier souls and/or various body parts for temp readings like that, especially the ones lapping thier cpus :-D