Conroe CPU and CPU core temp comparison thread

[hags2k]

Originally posted by: cmdrdredd

Originally posted by: hags2k
I originally posted this in another thread on a similar topic, but that thread seems to be dead.

So, from what I've read (http://www.silentpcreview.com/article191-page1.html a good article for example) the readings your motherboard gives can be pretty inaccurate.

Mine are

Idle: 43C
Load: 63C
System/Motherboard: 43C

E6600 stock speeds, stock voltage, stock HSF, 120mm case fan, 21C ambient (according to the thermometer in my room).

I know something is up, because even at full load I can touch the heatsink and it's barely warm to the touch. I also know it's seated well and I'm using Artic Silver 5 thermal compound. I reseated the fan several times, and one time I did NOT get it reseated well and my temps spiked about 15 degrees higher idle. Also, when I ramp up the cpu usage while touching the heatsink, I can feel it getting warmer, but at idle it's practically cold to the touch, and simply gets warm under load, not hot by any means.

The northbridge chipset heatsink next to the processor, in contrast, is VERY hot to the touch. I'm not sure if that's the other temp my bios (and speedfan, which always matches the bios) is giving me, but I'm just guessing.

Also, my system, even with a bit of overcocking (from 2.4 to 2.53 Ghz) continues to be rock-solid stable, using prime95, cpustress test, folding@home, and also when running seti@home and einstein@home simultaneously (which actually gets my cpu hotter than any of the other programs I've used)

So, what do you guys think? Inaccuracies, or perhaps some conroes just like to run hotter than others?

I know the P5B Deluxe reads wrong, and I believe other apps are still wrong too. With that beuing said I've noticed that some people have a concave IHS on their C2D which prevents a good physical contact between the Heatsink and CPU. The only way to fix it is send it in for RMA and hope the next one is good, or lap it and make it flat yourself.

That's a really interesting point. Would that be rectified to a certain extent by putting more thermal compount on the heat spreader (to fill in the concave hole, so to speak), or are you saying that the IHS is DOMED so that the cpu core underneath isn't contacting the spreader well?

 

[dopee123]

I'm sending in my 6300 for RMA. Both my stock heatsink and Ultra-120 are not making contact right dab in the middle of the cpu. (light leakage and the razor blade test confirmed it. I'm sending it back to Intel. Has anyone had experience RMA'ing stuff from Intel? Please do share.

Oh and those with idle temps of 50+, take off your heatsink and look at the thermal paste distrubition, and post back here. I have a strong hunch it's not core temp or other software reporting programs, but a crappy concave IHS.

 

[Skott]

Well, just for comparison sakes I'll add mine. No OCing yet. Everything in sig set at stock.

Idle: 35C

Load: 47C (flucuates between 46C-48C so i'll call it 47C)

 

[aggressor]

Originally posted by: dopee123
I'm sending in my 6300 for RMA. Both my stock heatsink and Ultra-120 are not making contact right dab in the middle of the cpu. (light leakage and the razor blade test confirmed it. I'm sending it back to Intel. Has anyone had experience RMA'ing stuff from Intel? Please do share.

Oh and those with idle temps of 50+, take off your heatsink and look at the thermal paste distrubition, and post back here. I have a strong hunch it's not core temp or other software reporting programs, but a crappy concave IHS.

Have you already started the process of RMAing it? I'm extremely interesting in what Intel has to say and if they'll even RMA a CPU because of this.

I have the same expierences as hags2k with my E6400 and Ultra-120. The heatsink, even at the base, is warm to the touch at best.
Also, here is the test I did with my Ultra-120: http://img526.imageshack.us/my.php?image=img1521rn5.jpg

 

[n7]

Originally posted by: dopee123
I'm sending in my 6300 for RMA. Both my stock heatsink and Ultra-120 are not making contact right dab in the middle of the cpu. (light leakage and the razor blade test confirmed it. I'm sending it back to Intel. Has anyone had experience RMA'ing stuff from Intel? Please do share.

Oh and those with idle temps of 50+, take off your heatsink and look at the thermal paste distrubition, and post back here. I have a strong hunch it's not core temp or other software reporting programs, but a crappy concave IHS.

Several people have already confirmed this over at XS a while ago.

Mine has the same issue.

Only solution is to lap the CPU, which is simply retarded.

 

[orangat]

How rough is the top of the heatspreader anyway?

 

[Xvys]

I get real wacky temp. readings. My bios always reads CPU 62C. SpeedFan and Everest can read CPU anywhere from 19C to 62C, for no apparant reason. After strenuous gaming the CPU may read 21C, then sitting idle it will read 55C. The CPU fan speed does not seem to follow the temp readings at all. It will be at 1700 at idle, then go up to 2200 after stress, regardless of the supposed temp readings. I think the fan is correct, speeding up after heavier use. My feeling may have a bad sensor on the MB? I am using the stock Intel fan/hs with the 3 factory paste strips. The MB & HD temp readings are always in the 42C range, and the "Aux" always reads 62C on both programs?
----------------------------

AsRock 775i65G / E6300 @ 2.1
Samsung DDR400 512x2 @ 2.5-3-3-7
Geforce 5900XT
Seagate 80G sata HD
Soundblaster Audigy
Leadtek TV/DVD/Captue card

 

[kpamir]

With the Asus probe program and with Speedfan my temp readouts are as follows

29C idle
38C with both cores under 100% load

With Core Temp

id get an idle of 52 or so
and i never checked out what my load temps were, ill do that when i get my p5b.

 

[cmdrdredd]

Originally posted by: n7

Originally posted by: dopee123
I'm sending in my 6300 for RMA. Both my stock heatsink and Ultra-120 are not making contact right dab in the middle of the cpu. (light leakage and the razor blade test confirmed it. I'm sending it back to Intel. Has anyone had experience RMA'ing stuff from Intel? Please do share.

Oh and those with idle temps of 50+, take off your heatsink and look at the thermal paste distrubition, and post back here. I have a strong hunch it's not core temp or other software reporting programs, but a crappy concave IHS.

Several people have already confirmed this over at XS a while ago.

Mine has the same issue.

Only solution is to lap the CPU, which is simply retarded.

Maybe retarded for you, but not to me when I have a 100% working CPU and only about 30minutes of work smoothes it out and gives 5-10c lower temps.

 

[cmdrdredd]

Originally posted by: hags2k

Originally posted by: cmdrdredd

Originally posted by: hags2k
I originally posted this in another thread on a similar topic, but that thread seems to be dead.

So, from what I've read (http://www.silentpcreview.com/article191-page1.html a good article for example) the readings your motherboard gives can be pretty inaccurate.

Mine are

Idle: 43C
Load: 63C
System/Motherboard: 43C

E6600 stock speeds, stock voltage, stock HSF, 120mm case fan, 21C ambient (according to the thermometer in my room).

I know something is up, because even at full load I can touch the heatsink and it's barely warm to the touch. I also know it's seated well and I'm using Artic Silver 5 thermal compound. I reseated the fan several times, and one time I did NOT get it reseated well and my temps spiked about 15 degrees higher idle. Also, when I ramp up the cpu usage while touching the heatsink, I can feel it getting warmer, but at idle it's practically cold to the touch, and simply gets warm under load, not hot by any means.

The northbridge chipset heatsink next to the processor, in contrast, is VERY hot to the touch. I'm not sure if that's the other temp my bios (and speedfan, which always matches the bios) is giving me, but I'm just guessing.

Also, my system, even with a bit of overcocking (from 2.4 to 2.53 Ghz) continues to be rock-solid stable, using prime95, cpustress test, folding@home, and also when running seti@home and einstein@home simultaneously (which actually gets my cpu hotter than any of the other programs I've used)

So, what do you guys think? Inaccuracies, or perhaps some conroes just like to run hotter than others?

I know the P5B Deluxe reads wrong, and I believe other apps are still wrong too. With that beuing said I've noticed that some people have a concave IHS on their C2D which prevents a good physical contact between the Heatsink and CPU. The only way to fix it is send it in for RMA and hope the next one is good, or lap it and make it flat yourself.

That's a really interesting point. Would that be rectified to a certain extent by putting more thermal compount on the heat spreader (to fill in the concave hole, so to speak), or are you saying that the IHS is DOMED so that the cpu core underneath isn't contacting the spreader well?

The IHS is just not making contact to the bottom of the HSF. Nothing to do with the 2 cores underneath.

 

[hags2k]

Thanks for taking the time to answer my questions, cmdrdredd.

Anyway, quick update (also posted in the thread where they're talking about RMAing CPUs with concave IHS):

I don't know about any of your motherboards, but I was amazed to discover that even at stock speeds my motherboard had taken it upon itself to up the cpu voltage to about 1.41 V according to CPU-Z. I manually set the voltage to the reccomended 1.325 V that intel reccomends and my motherboard reports as the normal voltage, and much to my surprise I just shaved 15 DEGREES off of my temps. It's now running around 44C at full load, as opposed to 61C before. CPU-Z is also now reporting correct voltages (slightly under, at 1.312 actually).

I have a Gigabyte 965P-DS3 mobo, and i don't recall the voltages being out of whack until after updating my bios to the latest revision. If anyone has a similar board or hasn't checked thier voltage settings, I'd look into that. 15 degrees is freakin unbelievable, but I don't think I'll be needing to RMA my CPU after all

Anyone seen anything simliar happening?

 

[Compellor]

Originally posted by: hags2k
Thanks for taking the time to answer my questions, cmdrdredd.

Anyway, quick update (also posted in the thread where they're talking about RMAing CPUs with concave IHS):

I don't know about any of your motherboards, but I was amazed to discover that even at stock speeds my motherboard had taken it upon itself to up the cpu voltage to about 1.41 V according to CPU-Z. I manually set the voltage to the reccomended 1.325 V that intel reccomends and my motherboard reports as the normal voltage, and much to my surprise I just shaved 15 DEGREES off of my temps. It's now running around 44C at full load, as opposed to 61C before. CPU-Z is also now reporting correct voltages (slightly under, at 1.312 actually).

I have a Gigabyte 965P-DS3 mobo, and i don't recall the voltages being out of whack until after updating my bios to the latest revision. If anyone has a similar board or hasn't checked thier voltage settings, I'd look into that. 15 degrees is freakin unbelievable, but I don't think I'll be needing to RMA my CPU after all

Anyone seen anything simliar happening?

I undervolted my E6400 to 1.24V and it helped drop my core temps by about 10 degrees. Mine is overclocked to just 2.6 GHz. At that voltage, it passes Orthos Beta and Super PI 32 MB. The ASUS P5B vanilla board shows it running at 1.2v under heavy load using CPU-Z. Those running at 3 GHz+ overclocks with 1.4v or higher can probably be stable at lower voltages.

 

[Termie]

Based on Everest readings:
Idle:
Motherboard: 40C
CPU1: 40C
CPU2: 39C
Northbridge: 55C

Load (gaming):
Motherboard: 44C
CPUs: ~50C
Northbridge: 68-69C!!!

I'm pretty sure my IHS is flat, as I noticed my AS5 evenly spread out when I was installing my Scythe Mine, taking it on and off to struggle with the snap-in pins.

Only a few of you have mentioned the northbridge temps. Should I be concerned about these? And if nothing is wrong with my IHS, could those core temps be accurate, problematic, or both?

 

[Janne2]

I bought an E6400 to replace my E6300, going to a "HTPC". 6300 did 3220Mhz at 1.318v. 6400 does 3400 at 1.375v. Not tested any more. 6300 week 24, 6400 week 27.
The whole system stayed tha same, and E6400 runs cooler...!? lol

Stock speeds
E6300 45 idle, 55 load
E6400 40 idle, 50 load


6300@3220 48 idle, 57 full load at 1.318v
6400@3400 44 idle, 56 full load at 1.375v

6300 is lapped, temps were worse before.
All temps from core temp/TAT

Oddly sometimes TAT shows 3-4c lower than coretemp with 6400
with 6300 it was allways same, coretemp=TAT

 

[phile]

Originally posted by: Janne2
I bought an E6400 to replace my E6300, going to a "HTPC". 6300 did 3220Mhz at 1.318v. 6400 does 3400 at 1.375v. Not tested any more. 6300 week 24, 6400 week 27.
The whole system stayed tha same, and E6400 runs cooler...!? lol

Stock speeds
E6300 45 idle, 55 load
E6400 40 idle, 50 load


6300@3220 48 idle, 57 full load at 1.318v
6400@3400 44 idle, 56 full load at 1.375v

6300 is lapped, temps were worse before.
All temps from core temp/TAT

Oddly sometimes TAT shows 3-4c lower than coretemp with 6400
with 6300 it was allways same, coretemp=TAT

Jebus!!! It's starting to look s though later week chips are much, much better overclockers. I needed 1.35 to get my E6600 stable at 3GHz. You're getting better performace out a cheaper chip. Everytime I read a post like yours, I'm moved one step closer to ordering another one, and selling mine.

Regarding CoreTemp vs. TAT, they differ by a good 3C-5C on my rig. I'm left wondering if readings from the core digital thermal sensors are reliable.

-phil

 

[Janne2]

Sure theyre reliable, but is coretemp or Tat reliable to read them? coretemp a beta prog, and TAT made for yonah.
That E6300 surely would go much more, but those temps get out of hand. Not that comfortable if they exceed Intels 61.1c mark

 

[Ruhnie]

Is there a way to read northbridge temp other than using Everest? Or is the free edition enough?

 

[hags2k]

Originally posted by: Janne2
Sure theyre reliable, but is coretemp or Tat reliable to read them? coretemp a beta prog, and TAT made for yonah.
That E6300 surely would go much more, but those temps get out of hand. Not that comfortable if they exceed Intels 61.1c mark

TAT and CoreTemp seem to use exactly the same process and reference temp (TCaseMax 85C) to read the temps. I think both are incorrect, as according to those programs the cpu won't throttle back until it hits 85C, and this has been tested by at least one guy in this forum already. I just don't think the temperature is actually 85C when that happens. I think it's actually a lower temp, but since all the readings are made relative to that reference temp, that means all the readings coretemp/tat are giving are potentially bogus, possibly by as much as 10C.

 

[Janne2]

Where in TAT you get TcaseMax 85c? I noticed no mention. In XS forums the guy "the coolest" was not even sure what that 85c was!?
Od thing being with 6300 they showed exactly same, but with 6400 there is a 3-4 decree difference.

 

[hags2k]

Originally posted by: Janne2
Where in TAT you get TcaseMax 85c? I noticed no mention. In XS forums the guy "the coolest" was not even sure what that 85c was!?
Od thing being with 6300 they showed exactly same, but with 6400 there is a 3-4 decree difference.

Sorry, may have been a little deductive reasoning there. The guy who posted about coretemp (http://www.xtremesystems.org/forums/showthread.php?t=103638) aka "The Coolest" pointed out that the 85C or 100C values for Yonah (also the most recent processors that TAT is valid for, apparently) are the TCaseMax values used to calculate the temps of Conroe, Merom, and Yonah chips when using CoreTemp. The values I've gotten for coretemp EXACTLY match the values given by TAT, and given that according to "The Coolest" all temps given by the digital diodes are relative to the TCaseMax value, I assumed that TAT and CoreTemp were using the same starting point to measure the temps.

The 85C, by the way, is apparently given by the CPU itself, according to the author! However, there's no way to verify that this is true or that the 85C is an accurate temperature. It does seem to be something programmed into the CPU, though, as when you hit 85C according to TAT (and coretemp, by extension) the CPU throttles back. A guy in one of these threads did that and posted his results/screenshots.

I do believe that what we can say for sure is that TAT and almost certainly coretemp (at least on my system and those of many other users) measures the same thing on the chip, and that the digital data given by those diodes is relative to some reference temp. The problem we're facing now is whether or not that reference temperature is accurate. Either way, that reference temp is the one where the CPU starts to throttle back, which back in the P4 days was also the thermal junction temp, and was around 67C IIRC. I don't know if that means anything for us now, but that's the reason I also assumed that the 85C temp was not accurate.

 

[Baked]

Damn, just replaced the Artic Cooling Freezer Pro 7 w/ the Thermalright Ultra-120. At stock speed, I got a 10C drop. Holy crap. I was gonna use the Yate Loon 120mm, but found out it's one of those fans w/ pillar on all 4 sides, I can't use Thermalright's wire fan clip. Gonna do some dremel work and cut those out this weekend.

Did I mention how massive this Thermalright Ultra-120 is? You can kill somebody w/ this thing.

 

[Janne2]

Originally posted by: the coolest
*EDIT*: I haven't read any updated docs in a while, I might be confused. I will have to recheck and see if the 85C value is indeed the TCaseMax or Tjunction

http://www.xtremesystems.org/forums/showthread.php?t=112980&page=5

 

[hags2k]

Originally posted by: Janne2

Originally posted by: the coolest
*EDIT*: I haven't read any updated docs in a while, I might be confused. I will have to recheck and see if the 85C value is indeed the TCaseMax or Tjunction

http://www.xtremesystems.org/forums/showthread.php?t=112980&page=5

Excellent info. I had read the intel spec sheet and was aware of the 60.1c number. However, I'm still not totally convinced that the 85C reference that coretemp uses (and TAT, also designed for yonah) is actually TCaseMax and not a different temp point. I think it is TJunction, in fact, but Tjunction seems to be different as well. The guy who pointed out the discrepancy mentioned a temp relative to 60.1, but by my calculation that would put my current temp at about 12C, which as much as I'd like to dream just ain't gonna happen. Not with my AC unit, and not on air

His conclusion that on his board (and most boards that give accurate T-deltas through prope2/speedfan) is a more accurate read and the one we should be concerned about I think is accurate. It seems to me that as long as my speedfan reading is under 60C at load (which it definitely is) I'm more or less good. At 60C via probe2/speedfan, that puts me at about 68-70C via coretemp/TAT, but that's 15 degree below the temp where it will throttle my CPU down, so that seems like a pretty conservative reference point to me.

At this point, I'm thinking a good rule of thumb is subtract 10C from your coretemp/TAT readings if you want a reasonable guess as to your actual CPU temps we need to be concerned about.

 

[slatr]

TT big typhoon
DS3 F4e
6400 CPU
2 gb gskill 800
80 mm fan aimed at northbridge
2 intake , 2 exhaust fans i case

7X429 = 3ghz

1.31v

core temp

idle - 45c
load - 56c (orthos)

**********

will do 8 x 400 at 3.2 but has to be a 1.325 v to be stable, loads there at 60c.

I also have raised the fsb with the X7 multiplier some. Temps stay exactly the same so far in 10 mhz increments tested to about 3.1


edit speed fan shows cpu idle of 82f and 108f load