Sign in with
Sign up | Sign in
Your question

C2D CPU FULL LOAD Temps - Why worry So Much?

Tags:
  • Cooling
  • CPUs
  • Temperature
  • Overclocking
Last response: in Overclocking
Share
January 5, 2007 2:43:06 PM

:?: I've been reading a lot of posts about what is the right temperature for this or that c2d cpu run at this or that speed (especially at full load) with this or that cooler measured with this or that program, etc.

First, I agree that cpu temps should NOT be totally discounted. They will tell you if you got your cooler mounted correctly and whether it continues to work correctly. (For example fan failure.) They also provide general info about the cpu voltage and speed settings. And yes, they will affect your machine's computing capability, and ultimately the cpu life.

Second, If you want/need to compare cpu cooler solutions to determine which one to purchase...OK. The specific full load temperature differentials provide the info to evaluate cooler efficiency and, given cost, cooler value.

But, cpu specific-temperature concern based on Full load max allowable temps doesn't otherwise seem to make sense.

The Intel C2D cpus are self-limiting based on internal temp sensor diodes. That is, if they get too hot the cores throttle back (lower speed) to a "safe" Intel determined temperature. So it would seem that once you are assured that the cpu/cooler is OK and cpu voltage is within limits, that you set your cpu speed based on full load stability testing and monitoring of the core usage for heat related throttling.

In other words the question is, as far as getting the max speed out of your cpu, you shouldn't worry so much about the specific full load temps? Or stated another way, if it's stable and not throttling, why worry about your specific full load temperature? If not, change your system setup and/or look for a better cooler. Please note, emotional outbursts and adjectives shall be binned.

More about : c2d cpu full load temps worry

a b à CPUs
a b K Overclocking
January 5, 2007 3:34:48 PM

I'm currently researching this topic on 3 other threads:

Quote:
I've been reading a lot of posts about what is the right temperature for this or that c2d cpu run at this or that speed (especially at full load) with this or that cooler measured with this or that program, etc...

But, cpu specific-temperature concern based on Full load max allowable temps doesn't otherwise seem to make sense...

So it would seem that once you are assured that the cpu/cooler is OK and cpu voltage is within limits, that you set your cpu speed based on full load stability testing and monitoring of the core usage for heat related throttling.

In other words the question is, as far as getting the max speed out of your cpu, you shouldn't worry so much about the specific full load temps? Or stated another way, if it's stable and not throttling, why worry about your specific full load temperature?


Let's take a look at Intel's Thermal Specification for the E6600 which is 60.1c as per the following link:

http://processorfinder.intel.com/details.aspx?sSpec=SL9...

Quote:
Thermal Specification: The thermal specification shown is the maximum case temperature at the maximum Thermal Design Power (TDP) value for that processor. It is measured at the geometric center on the topside of the processor integrated heat spreader. For processors without integrated heat spreaders such as mobile processors, the thermal specification is referred to as the junction temperature (Tj). The maximum junction temperature is defined by an activation of the processor Intel® Thermal Monitor. The Intel Thermal Monitor’s automatic mode is used to indicate that the maximum TJ has been reached.


The first part of the 60c spec implies a single measuring point, which would be in contact with the CPU cooler, via the heat sink compound. Since there is no sensor at this measuring point, then BIOS temps are indicated from CPU look-up tables, referencing the motherboard's CPU socket temp sensor, to calculate this single measuring point. Additionally, one core always runs hotter than the other, so this calculated measurement is at best, an average value.

Wusy's Overclocking Guide suggests using Core Temp, which displays slightly higher temps than Intel's Thermal Analysis Tool, (TAT). The second part of the spec refers to mobile CPU's measured at junction temperature (Tj) with Intel's Thermal Monitor (TAT), which reaches thermal red-line at 80c. Core Temp also measures internal CPU core (Tj) sensors, which correctly displays the (Tj max) spec as 85c.

Obviously, there are some calibration and specification ambiguities here. Since Core Temp displays higher load temps than TAT, which is a Notebook tool, then TAT's values are inacurate for PC's. It's vague as to how the ~ 15c discrepancies between BIOS, motherboard utilities and SpeedFan, versus TAT and Core Temp should be interpreted. The difference seems to approximate BIOS + ~ 15c = Core Temp.

Thermal Design Power (TDP) is exceeded by a considerable margin when vCore and CPU clock are increased over spec's at 100% load, so 50c BIOS and 65c Core Temp should be safe and reasonable thermal values for heavily OC'd worst-case gaming loads. :) 
January 5, 2007 3:53:30 PM

Quote:
:?: I've been reading a lot of posts about what is the right temperature for this or that c2d cpu run at this or that speed (especially at full load) with this or that cooler measured with this or that program, etc.

First, I agree that cpu temps should NOT be totally discounted. They will tell you if you got your cooler mounted correctly and whether it continues to work correctly. (For example fan failure.) They also provide general info about the cpu voltage and speed settings. And yes, they will affect your machine's computing capability, and ultimately the cpu life.

Second, If you want/need to compare cpu cooler solutions to determine which one to purchase...OK. The specific full load temperature differentials provide the info to evaluate cooler efficiency and, given cost, cooler value.

But, cpu specific-temperature concern based on Full load max allowable temps doesn't otherwise seem to make sense.

The Intel C2D cpus are self-limiting based on internal temp sensor diodes. That is, if they get too hot the cores throttle back (lower speed) to a "safe" Intel determined temperature. So it would seem that once you are assured that the cpu/cooler is OK and cpu voltage is within limits, that you set your cpu speed based on full load stability testing and monitoring of the core usage for heat related throttling.

In other words the question is, as far as getting the max speed out of your cpu, you shouldn't worry so much about the specific full load temps? Or stated another way, if it's stable and not throttling, why worry about your specific full load temperature? If not, change your system setup and/or look for a better cooler. Please note, emotional outbursts and adjectives shall be binned.
Cpu’s now don’t get near as hot as lets say an xp3200 however if dude's wants to push the OC'ing to the extreme than cooling is critical to sustain a on going O,C without damaging the system or burning out the address lines in the CPU, When the busses or address pipes get darkened from heat there are problems and next is CPU errors or registers in applications. Most cases your right but when ocing is concerned cooling is the real truth about the, Myth ocing shortens a CPU life cycle.
Related resources
January 6, 2007 12:42:41 AM

Exhibit A:

Quote:
First, I agree that cpu temps should NOT be totally discounted.


You're right... :)  Temperatures can't be discounted because they influence the operational life of our investment. We have two easy to obtain metrics: idle and load temperature. We use those to determine if our overclock is reasonable given our cooling.

Quote:
The Intel C2D cpus are self-limiting based on internal temp sensor diodes. That is, if they get too hot the cores throttle back (lower speed) to a "safe" Intel determined temperature.


I could be wrong about this, but my understanding is "step" throttling is only related to CPU load, not temperature. The cpu multiplier and voltage are scaled back when the cpu load is low. This can actually get in the way of OC stability and is disabled by OCers.

Quote:
But, cpu specific-temperature concern based on Full load max allowable temps doesn't otherwise seem to make sense.


Even if you discount my interpretation of stepping (which could be incorrect), ignoring the load temperature is a silly notion. The load temperature isn't a theoretical temperature. It is a temperature we witness when we stress our CPU. Any algorithmic throttling it part of the benchmark.
January 6, 2007 1:19:39 AM

Hate to throw another thing into the mix, but certain motherboards also allow you to DISABLE temp throttling for the Core2, namely, the 680i based current boards. Not sure about the Asus Striker and their other board, but any one based on the Nvidia reference (eVGA, BFG, etc).

Another thought when worrying about overclocking, full load temps. :p 
a c 140 à CPUs
a b K Overclocking
January 6, 2007 1:42:22 AM

Quote:
Hate to throw another thing into the mix, but certain motherboards also allow you to DISABLE temp throttling for the Core2, namely, the 680i based current boards. Not sure about the Asus Striker and their other board, but any one based on the Nvidia reference (eVGA, BFG, etc).

Maybe you can shut off Speed Step(the cool n quiet of intel). As far as i know Intels thermal throttle is built into the cpu....

Quote:
I could be wrong about this, but my understanding is "step" throttling is only related to CPU load, not temperature. The cpu multiplier and voltage are scaled back when the cpu load is low. This can actually get in the way of OC stability and is disabled by OCers.

I assume you mean speed step here as well....its a great feature but it does not seem to react as quick as cnq

Quote:
Cpu’s now don’t get near as hot as lets say an xp3200 however if dude's wants to push the OC'ing to the extreme than cooling is critical to sustain a on going O,C without damaging the system or burning out the address lines in the CPU, When the busses or address pipes get darkened from heat there are problems and next is CPU errors or registers in applications. Most cases your right but when ocing is concerned cooling is the real truth about the, Myth ocing shortens a CPU life cycle.

True. Another thing no ones takes into consideration now days is that most temperature sensing just years ago was socket based....so once 50-60 seemed HOT....but now that its in the core(A64 and Core2) 60 is still rather cool....at least in comparison to older cpu's.

All in all, I think we are going the right way.
CPU's DO run cooler...but seem to run hotter as a result of the new in core sensors....i ran my A64 around 52c/load bios(mbm5) and thats just fine....got up to 60 on a hot day(with the fans on low)....Get out my core 2 and get like 39-41 idle....WTF i had like 32-35 idle on my A64.... Why? simple the MBM readings where not in the core....core temp shows a higher temp.....I have not loaded it to check....but i say it runs as hot as my core2 @ load.....but cooler @ idle...
This is something that has to get out....lots of people think they are idling @ 25-30 when in fact they are @ 35-40. This comes into play with the fanboys too....ohhh my a64 runs cooler....ohh my core 2 runs soo cool....they are all close....

Back on topic... YES the core 2 should throttle to save it self....but i have not seem any confirmation of it....
a b à CPUs
a b K Overclocking
January 6, 2007 1:47:51 AM

CPU Temps *C.

BIOS/CoreTemp
-60--/----75-----
-55--/----70-----
-50--/----65-----
-45--/----60-----
-40--/----55-----
-35--/----50-----
-30--/----45-----

:D  :D  :D  :D  :D 
January 6, 2007 1:51:36 AM

Quote:
Hate to throw another thing into the mix, but certain motherboards also allow you to DISABLE temp throttling for the Core2, namely, the 680i based current boards. Not sure about the Asus Striker and their other board, but any one based on the Nvidia reference (eVGA, BFG, etc).

Maybe you can shut off Speed Step(the cool n quiet of intel). As far as i know Intels thermal throttle is built into the cpu....



Direct from the 'CPU Configuration' in the eVGA 680i board's bios.

>>>>>CPU THERMAL CONTROL (Disabled/TM1/TM2/TM1 and TM2) <<<<<<<<<

One of the settings throttles the speed back, and the other halts the processor completely when it overheats. I'd have to reboot to tell you which is which, but I'm too lazy to do that right now.

There's a seperate setting for Speedstep, has nothing to do with it. Speedtep is usage based, not heat based.

Using revision P23 bios, but it was there before I updated it as well.
January 6, 2007 3:52:46 AM

Well, my experience has shown that dropping the CPU core temps by 8C can mean the difference between a stable system and one that re-boots at random. If you're pushing the limits of the pieces parts, controlling tempurature is a key factor. If you're running stock, then I wouldn't even load a temp monitor and be happy.

Tom
a c 140 à CPUs
a b K Overclocking
January 7, 2007 5:53:53 AM

I stand corrected....Damn Asus and there bad description...

Well off to turn that back on.... :) 
January 7, 2007 6:05:37 AM

Quote:
I stand corrected....Damn Asus and there bad description...

Well off to turn that back on.... :) 


I've seen alot of people reccomend leaving it off if you're overclocking on the 680i boards, no clue if it's neccessary.

I left the one that HALTS the proccessor if it overheats on (TM2), but left the throttling on turned off (TM1). Makes more sense to me that way, i'd rather it be protected if it really reaches max temp.
a b à CPUs
a b K Overclocking
January 7, 2007 6:17:16 AM

Quote:
I left the one that HALTS the proccessor if it overheats on (TM2), but left the throttling on turned off (TM1). Makes more sense to me that way, i'd rather it be protected if it really reaches max temp.


Intel's spec for Thermal Case Maximum (Tc max) on the TM2 sensor is 60c. Temps displayed in BIOS, motherboard utilities and SpeedFan. 60c hot, 65c overtemp.

Intel's spec for Thermal Junction Maximum (Tj max) on the TM1 sensors is 85c. Temps displayed in Thermal Analysis Tool (TAT) and Core Temp. 75c hot, 80c throttle, 85c shutdown.

The scale illustrates TM2 BIOS and TM1 Core Temp typical offsets: BIOS + ~ 15c = Core Temp.

CPU Temps *C.

BIOS/CoreTemp
-60--/----75-----
-55--/----70-----
-50--/----65-----
-45--/----60-----
-40--/----55-----
-35--/----50-----
-30--/----45-----

I hope this helps to make sense of CPU temp alphabet soup. 8)
a c 140 à CPUs
a b K Overclocking
January 7, 2007 5:21:58 PM

that clears things up
January 7, 2007 6:43:54 PM

Ok, from what I gather from this thread, I'm totall fine on temps. But just so someone can have the pleasure of saying, "For God's sake, YES!" I have to ask the question for sheer humbling, yet, better sleeping purposes.

Here goes:

System with PECI off in the BIOS = true temps (speedfan and coretemp read the same):

Core 2 Duo 6400 @ 3.2Ghz
Asus P5B Deluxe
Corasair DDR 800
Thermalright XP-120 with AS5

Temps:

CPU: 41°-45°c idle, 55°-59°c load (with Orthos)
Mobo: 28°c idle, 33°c load
Tjunction 85°


So I'm totally fine, right?

Also, in Orhtos, is it normal for one core to be a little faster than the other? I compare their progress and one seems to always be a little ahead. Thanks!
January 7, 2007 7:16:46 PM

During stress testing your looking for errata not detected by those thermal diodes.
Keeping the max temps below 60.1C is the ideal.
True the CD2 takes care of itself but no one wants the CPU to throttle during heavy apps requiring maximum efficiency.

It doesn’t make sense to ignore temps higher than its designed thermal limitations
a b à CPUs
a b K Overclocking
January 7, 2007 8:44:34 PM

Quote:
CPU: 41°-45°c idle, 55°-59°c load (with Orthos)
Mobo: 28°c idle, 33°c load
Tjunction 85°

So I'm totally fine, right?

Also, in Orhtos, is it normal for one core to be a little faster than the other? I compare their progress and one seems to always be a little ahead. Thanks!


The best way to load test you CPU cores is to run Thermal Analysis Tool (TAT) which will simultaneously run both cores at 100% load, and observe the temps on Core Temp, which typically reads a few degrees higher than TAT. It's normal for one core to read higher than the other. 8)
January 8, 2007 12:12:18 AM

Ok, what the hell does TAT do? I ran it and my load temps jumped up to 65°c!

I stopped it, and ran Orthos, CPU Burn-in, AND Prime 95 simultaneously (while I type this). My CPU usage (in speedfan) is obviously 100% but my load temp tops out around 55°c.

How the HELL does TAT run my temps up an entire 10°c over 3 stress tests combined? Also, why does it see my Core 2 Duo as a Pentium M?

Thanks
a b à CPUs
a b K Overclocking
January 8, 2007 12:28:31 AM

Quote:
Ok, what the hell does TAT do? I ran it and my load temps jumped up to 65°c!


With TAT running and Core Temp monitoring, I get 67c, so your 65c is right. Remember that 65c Tjunction is still cool. 70c warm, 75c hot, 80c throttle, 85c shutdown. Refer to my scale above.

Quote:
I stopped it, and ran Orthos, CPU Burn-in, AND Prime 95 simultaneously (while I type this). My CPU usage (in speedfan) is obviously 100% but my load temp tops out around 55°c.

How the HELL does TAT run my temps up an entire 10°c over 3 stress tests combined?


Orthos, CPU Burn-in and Prime 95 simultaneously still represent a composite of intermittent loads. TAT runs continuous 100% loads.

Quote:
Also, why does it see my Core 2 Duo as a Pentium M?


TAT was first written for mobile processors, so disregard the Pentium M ID. This is why the temps are typically a few degrees lower than Core Temp. 8)
January 8, 2007 12:31:10 AM

Well, you say my temps are accurate. The $217 question (amount I paid for cpu - ha ha) is, are they safe?
January 8, 2007 12:34:01 AM

Yer safe. I wouldn't go any hotter.

Tom
a b à CPUs
a b K Overclocking
January 8, 2007 12:37:57 AM

Yes, you're operating within Intel's specs, and recommended overclocking margins.
January 8, 2007 12:39:35 AM

Hey CT, what are your voltages? Especially your ram voltage?
a b à CPUs
a b K Overclocking
January 8, 2007 12:49:46 AM

Quote:
Hey CT, what are your voltages? Especially your ram voltage?


3.7Ghz CPU
1.5125 vCore
411Mhz FSB
1.65 vMch
1.3 vTerm
822Mhz DDR2
2.05 vDimm
36c Idle, 52c Load - BIOS
51c Idle, 67c Load - Core Temp

8)
January 8, 2007 12:56:45 AM

Quote:
Hey CT, what are your voltages? Especially your ram voltage?


3.7Ghz CPU
1.5125 vCore
411Mhz FSB
1.65 vMch
1.3 vTerm
822Mhz DDR2
2.05 vDimm
36c Idle, 52c Load - BIOS
51c Idle, 67c Load - Core Temp

8)

Here's what I get with an e6700:

3.6Ghz CPU
1.4375 vCore
400Mhz FSB
1.45 Vfsb
1.45 Vnorthbridge
800Mhz DDR2
2.25 vDimm
35c Idle, 58c Load - Core Temp

Tom
January 8, 2007 1:08:02 AM

tom, do you have PECI disabled in your bios? It's under CPU config
a b à CPUs
a b K Overclocking
January 8, 2007 1:15:17 AM

Quote:
Here's what I get with an e6700:

3.6Ghz CPU
1.4375 vCore
400Mhz FSB
1.45 Vfsb
1.45 Vnorthbridge
800Mhz DDR2
2.25 vDimm
35c Idle, 58c Load - Core Temp


Remember to qualify "load" by using TAT to run both cores simultaneously at 100%, and Core Temp to monitor highest Tjunction temp. If you've used this method to find "load" temp, then you're in excellent shape. If not, then you may find that your true "load" temp is actually higher than 58c. 8)
January 8, 2007 1:19:17 AM

PECI is disabled. Even so, the BIOS temps are about 3 deg cooler than either core temp or TAT. I trust Core Temp and TAT. They agree and are the hottest. I also tested Cre Temp against throttling. When the CPU hit 83C on Core Temp, it started throttling. I'm sick ;) 

Tom
January 8, 2007 1:22:28 AM

Quote:
Here's what I get with an e6700:

3.6Ghz CPU
1.4375 vCore
400Mhz FSB
1.45 Vfsb
1.45 Vnorthbridge
800Mhz DDR2
2.25 vDimm
35c Idle, 58c Load - Core Temp


Remember to qualify "load" by using TAT to run both cores simultaneously at 100%, and Core Temp to monitor highest Tjunction temp. If you've used this method to find "load" temp, then you're in excellent shape. If not, then you may find that your true "load" temp is actually higher than 58c. 8)

Excellent point. The numbers I gave are COre Temp and Orthos data. TAT data is 38C idle and 66C TAT load. I agree that one should always qualify their data. Sorry about that.

Tom
a b à CPUs
a b K Overclocking
January 8, 2007 1:26:53 AM

Quote:
Excellent point. The numbers I gave are COre Temp and Orthos data. TAT data is 38C idle and 66C TAT load.


Thanks for the info. Your 66c is consistent with my 67c, so we've standardized apples to apples. 66c still a cool enough Tjunction temp, so you're in good shape. 8)
January 8, 2007 1:39:01 AM

and I'm right there between you guys at 65c. Makes sense. Don't worry, I won't dream that I'm skiing :wink:

Also, thanks for the voltages. Actually, all my voltages are lower than both of yours. So shouldn't I have lower temps? Then again, I'm using an XP-120 vs your Zalmans.

Here they are my voltages:

Mem Voltage: 2.1v (corsair spec)

Vcore: 1.3375v
FSB Termination: 1.3v
NB Core: 1.25
SB Core: 1.5
ICH Chipset: 1.057v

If I'm totally stable, should I change any of those? Maybe make my SB lower?
January 8, 2007 1:51:21 AM

If you're stable, I wouldn't worry about those voltages. Nothing high. If you want to play, work them down a little at a time until something doesn't work right.

Tom
a b à CPUs
a b K Overclocking
January 8, 2007 1:54:12 AM

Quote:
I'm right there between you guys at 65c. Makes sense...

Also, thanks for the voltages. Actually, all my voltages are lower than both of yours. So shouldn't I have lower temps? Then again, I'm using an XP-120 vs your Zalmans.


It probably relates to differences in case air flow between our rigs.

Quote:
Here they are my voltages:

Mem Voltage: 2.1v (corsair spec)

Vcore: 1.3375v
FSB Termination: 1.3v
NB Core: 1.25
SB Core: 1.5
ICH Chipset: Auto

If I'm totally stable, should I change any of those? Maybe make my SB lower?


I think you're right. 8)
January 8, 2007 2:07:32 AM

well, my sb won't go any lower than the 1.5 unless I want to do "auto".

I figure my airflow must be pretty good in my case. I have a Coolermaster Centurion 5 with the mesh front and side window. I have a 120mm and an 80mm fans in the front, a 120mm fan on my CPU, a 120mm fan in my PSU (Antec NeoPower 480) and a 120mm fan blowing out the back.

I like silence AND OCing so Q-fan is disabled on my CPU fan but my case fans are on Q-fan enabled in "performance mode" under my bios. My rig's pretty silent and cool. My mobo temp goes between 28-33c and my hdd temps hover around 30-35c. I never feel hot air coming out the back unless it's coming from my PSU. Then again, my XP-120 never feels hot or even warm for that matter.
a b à CPUs
a b K Overclocking
January 8, 2007 2:28:35 AM

Yeah, I'm probably the most anal man you could meet when it comes to neatness under the hood. With the cover off, the wiring is routed and zip tied so as to not present any obstruction to air flow.

I have a small tight case out of neccesity, because I live aboard my sailboat, so it has to fit into a confined space. Every possible air leak is duct taped from the inside, and plenums channel air flow for efficient cooling. There's a single 120mm intake, and twin 60mm exaust, with PSU and GPU for the remaing outflow.

I still haven't taken any images of the interior, so I can't post the proof, but I can say that the case cools with the best of them. 8)
January 8, 2007 2:33:42 AM

my cables aren't zip tied, but none of them obstruct airflow. They're wrapped and tucked neatly away so you can see everything in all it's glory through my window, nicely lit with blue LED's that are also tucked away so you can't tell where the light is coming from. Maybe I'll post on the forumz' photo album.
!