Sign in with
Sign up | Sign in
Your question

Unsure about C2D temperatures

Last response: in Overclocking
Share
March 22, 2007 7:47:22 PM

Sorry about this, I know there is a temperature guide, I read it plus a few threads, but I am still confused... Here are the details of my setup:

Gigabyte GA-965P-DS3 BIOS F9
C2D E4300
2 Gig G-Skill DDR2-800 (2x1GB) @ 5-5-5-15
Thermalright XP-120 heat sink with silent 120mm fan (sunon I think) intalled with Artic Silver 5
Antec Sonata 2 with the rear 120mm fan set to lowest speed (the sound is annoying at higher speed, but if I have no choice.... but REALLY, if I have no choice)
BFG Geforce 7600GT OC with a Zalman VF700-Cu

Now, the BIOS/OC settings. FSB is at 320, multiplier x9, my C2D E4300 is at 2.88Ghz, all voltages stock except Vcore at 1.3375 abd FSB set to +0.1V. CE1, CIA2, EIST, spread spectrum and "limit CPUID max 3" are disabled, CPU fan is connected directly to PSU, so always 100%.

The system has been tested stable with 10h30 Ortho. It does not even boot at 230 Mhz FSB without the +0.1V FSB. It does boot with Vcore at 1.32500 or 1.33125 and is stable with almost anything but eventually reboot when running Ortho.

Now, here are my reading:

CPU-Z:

Core speed 2880 Mhz
Rated FSB: 1280 Mhz

IDLE Readings:

SpeedFan:

Vcore1: 1.31V (which would mean my CPU is actually undervolted... This seems to be due to my PSU giving lower voltages than supposed to (Antec 450 Watts), my %v is at 4.81V)
Ok, temperatures now, tell me if I am wrong at some point:
Temp1: 34C
Temp2: 26C (this is supposed to be Tcase right?)
Temp3: -2C (this would actually mean an ambient temp of 21C)
Core0: 22C
Core1: 31C (those 2 are suposed to be Tcase for each core)

TAT:

CPU 0: 38
CPU 1: 45

Core Temp:

Core 0: 22C
Core 1: 29C

Now, under load (Ortho for 10h30 min)

SpeedFan:

Temp1: 34C
Temp2: 50C (24C higher than Idle, do I have a problem??)
Temp3: -2C
Core0: 46C (Those are Tcase right?, the C2D Temp Guide says 50C is safe, so I am safe right? But delta between Idle and Load is higher than 15.... what's that supposed to mean?)
Core1: 46C

TAT:

CPU0: 59C
CPU1: 61C (this is Tjunction, where 65C is safe, am I right, but still more than 15C delta between Idle and Load)

Core Temp:

Core 0: 44C
Core 1: 45C

If I understand all this correctly, my temps are safe, but still somehow high for a e4300 @ 2.88 Ghz with almost no Vcore boost and a thermalright XP-120 and a 120mm case fan.... And I have a big temperature difference between Idle and Load and it would be supposed to be much smaller since I have a great heat sink.

I would really really appreciate to hear what you guys think about this.
Thanks
Mike
March 22, 2007 9:51:27 PM

Allow me to try to answer a bit. You are a bit confused.

You have 2 Tjunctions (1 in each core) and 1 Tcase. Looks like TAT is getting your Tjunctions correct. Core 0 and Core 1 in Speedfan are Tjunctions. Your Speedfan and CoreTemp readings are clearly offset. You can adjust Speedfan with offsets, but I personally don't see the need. All that I am worried about at this point are my TAT readings of Tjunction.

15-degrees is the difference (roughly) you should see between Tcase and Tjunction. One of those temps in Speedfan is your Tcase...but again, I'm not sure that it's really necessary to worry about Tcase. 25-degrees is the max difference you want to see between Tjuncction at idle and Tjunction at 100% TAT load. Your load temps are from Orthos, which are not as high as you will see with TAT loading.

Also I think it's very strange how big of a temperature difference there is between your two cores at idle.
March 23, 2007 2:06:31 PM

what should I do about this big temperature difference between idle and load???
Related resources
March 23, 2007 2:50:31 PM

I dont know if any of you is familiar with the thermalright socket 775 adapter for socket 478 CPU cooler... It's an "X" that is screwed on the backplate of the motherboard. It's hard to explain (especially for a frech guy who has to write in english).

You can see a big picture of what this adapter is there: http://www.pc-look.com/boutik/product_plus2imagesold.ph...

Now, my questions is, do you guys think my high temperatures difference between idle and load are because the screws that attach the backplate to the 2 plastic retainers are not tight enough? I'm not sure how to say this in english, but, does the CPU-Heatsink contact need to have a high pressure? I am worried about tighting this too much, so I would like to be a bit more sure before I do this.

Thanks guys!
March 23, 2007 3:21:04 PM

Howdy ThereMike,

Quote:
I dont know if any of you is familiar with the thermalright socket 775 adapter for socket 478 CPU cooler... It's an "X" that is screwed on the backplate of the motherboard.


I had a Zalman 7700(?) that required the adapter as well. If you mean the two screws that affix the heatsink/fan to the adapter top plate, then yes, they require tightening sufficient enough to bend the retainer "ears" down toward the cpu heatspreader. Test by pushing down on the heatsink (down=toward cpu) while the computer is running and warmed up. If your Coretemp readings decrease within a minute,you need to reset your heatsink. That would also explain the difference between cores; which I would be more concerned with.

TonyB.

MSI P6NSLI Platinum, ver, 1.0
E6400 @2.14; idle=36c load=46-50c NOW=42/43c
(free air)
SJN1100b?, SytheNinja rev.B w/120mm sythe fan
2x1GB PDC6400LLK @800MHz
Antec TP430
XFX Fatal1ty 7600GT
March 23, 2007 6:52:54 PM

Applying a manual pressure to the heat sink dosent seem to change anything to temperatures.

About the difference in temperatures between the 2 cores, I find it strange too, but it's only different when idle, at load, temps are the same (+/- 1C).

Today, the difference between the 2 cores is "only" 5-6C. Maybe it's the artic silver 5 that wasent spead evenly and it starting spead better now.... I really have no clue what's going on. you guys would be worried with these temps or .... ?

The 4 major questions I have are:

What do these temps tell you (is it too hot, is it safe, is it more than safe and I have some more room for OC while keeping CPU to a safe temp)?

Do you think the difference between idle and load is too much and what could cause that?

Same question but for the diff in temp between Core 0 and 1 at idle.

Do these temps looks normal to you for a e4300 at 2.88 Ghz 1.3375V Vcore with a thermalright XP120 (with a 120mm fan at about 2000 RPM) in a case with a 120mm case fan (Antec Sonata 2 case fan set at low speed)?
March 23, 2007 10:16:45 PM

I think you are about maxed out on the OC imho. Some places you see people with the 'if it doesn't reboot it's OK' attitude. Around here everybody seems to want to keep their TAT 100% load temps under about 70 (some 65, some 75, you get the idea).

BUT, we don't know what your TAT 100% temps are. Run TAT at 100% for 10 minutes and report the temepratures read on the cores via TAT.

You temp between idle and load is not big (see the Core 2 Duo temp guide sticky) but that is because you are not reporting load temp from TAT 100% load.

I have never seem a temp difference between 2 cores as large as your (mine is uaully 2-3 degrees at tidle) but then I haven't been around here for very long.
March 24, 2007 5:58:46 AM

Ok, I've tested with TAT both cores at 100% for 10 minutes and my MAX temps was 68 for CPU 1 and 71 for CPU 2.
March 24, 2007 7:05:15 AM

I'm at about the same place as you. I'm scared to turn up the speed any higher. In fact I think I will be lapping my cpu this weekened...damn concave e4300's.
March 24, 2007 2:58:14 PM

Quote:
I'm at about the same place as you. I'm scared to turn up the speed any higher. In fact I think I will be lapping my cpu this weekened...damn concave e4300's.


Please, keep me posted about the result of this!
March 25, 2007 6:02:32 AM

Before you lap, it appears that TAT and the latest Core Temp build (0.95) assume a tjunction max of 100c on this chip. Intel has not made it clear whether the actual tjunction max is 85c or 100c or something else. IF the tjmax of the e4300 is 85c then TAT is overreporting temps on this chip by an average of +15c. Until Intel gives us more (oh I gotta :D  ) intel, hold on to your warranty. Your chip is probably fine.
a b à CPUs
a b K Overclocking
March 25, 2007 8:26:33 PM

From the Core 2 Duo Temperature Guide:

Quote:
Specifications

Intel's Thermal Specification: http://processorfinder.intel.com/details.aspx?sSpec=SL9...

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.

X6800 = 60.4c, Vcore max = 1.3525, TDP = 75w
E6700 = 60.1c, Vcore max = 1.3525, TDP = 65w
E6600 = 60.1c, Vcore max = 1.3525, TDP = 65w
E6400 = 61.4c, Vcore max = 1.3525, TDP = 65w
E6300 = 61.4c, Vcore max = 1.3525, TDP = 65w
E4300 = 61.4c, Vcore max = 1.3250, TDP = 65w
** 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.

Desktop C2D's = 85c
Ambient Temperature = 22c
Idle to Load Delta Max = 25c
Tcase + 15c = Tjunction
Thermal Sensor Accuracy = +/-1c


At 200 Mhz FSB, Core Temp may detect an E4300 as a Mobile C2D, which is Tjiunction 100c, as are the Quad Core's.

Hope this helps,

Comp
March 26, 2007 12:26:21 AM

Thanks compu :D  . This means that TAT and Core Temp 0.95 do read high on the e4300 by an average of +15c because they assign the tjunction max of 100c to this chip. 8)
(heres the actual e4300 link)
http://processorfinder.intel.com/details.aspx?sSpec=SL9...
Check out the Vcore numbers. I think I'm going to do some more undervolting :mrgreen:
March 26, 2007 1:28:03 AM

CoreTemp 0.94 for my e4300 listed Tjunction as 85 and core temps were 15-deg below TAT. CoreTemp 0.95 lists Tjunction as 100 and temps agree with TAT.

I though the consesus was that TAT was reading e4300s correctly?
March 26, 2007 1:44:45 AM

:D 
March 26, 2007 4:27:42 AM

Right now my e4300 is at 2.7-GHz and 1.325 volts.

It is idling at ~44 according to TAT.

If your 15-deg offset were true...then my core would be idling at 30-deg C. According to Compu, the Tcore to Tcase temp difference is roughly 15..which means my Tcase would be 15 deg...maybe 20 deg. But that's less than the ambiant of 23-deg in the room.
March 26, 2007 4:48:39 AM

What are your actual readings according to Core Temp 0.94?
March 26, 2007 5:05:43 AM

right now CoreTemp 0.94 says 25 and TAT says 42
March 26, 2007 5:13:56 AM

What is your system temp and what are your load temps? Setup details?

edit: Ouch...I see you already lapped
a b à CPUs
a b K Overclocking
March 26, 2007 2:50:28 PM

TheMike, mrknowitall, flyer335 and Pax2All, it seems that far too many users are becoming fixated on Core Temp's "Tjunction 85c" or "Tjunction 100c", and are also becoming confused by this field. I want to make it clearly understood that regardless of version, Core Temp's Tjunction field simply shows Intel's Tjunction max specification, and is INFORMATION ONLY!

Regardless of what Core Temp reports, Intel's spec's for Desktop C2D's (E4300 included) is 85c, while Mobile C2D's and Quad Cores are 100c. All too often, Core Temp does not report this Tjunction max spec correctly, and can misreport it at different FSB frequencies. This seems to confuse many users, so for this very reason, I no longer recommend Core Temp, since the release of SpeedFan 4.32, which can correct for offsets, while Core Temp can not.

From the Guide:

Quote:
Troubleshooting

(I) Core Temp shows Tjunction 85c or 100c. This is Intel's Tj max spec, and is information only.


Core Temp causes a tremendous amount of confusion with this Tjunction field, so my recommendation is to disregard Core Temp's Tjunction max spec. I hope this helps to clear up this issue, and put it to bed.

Comp 8)
March 26, 2007 4:53:37 PM

Computronix thats the crux of the issue.TAT and Core Temp 0.95 misreport the tjunction max of the e4300 and assign it at 100c instead of 85c. In fact the new 0.95 Core Temp build had the tjunction upped to 100c for the E4300 in its code so load temps would agree with TAT. That means the figure has real impact and is not just for information. As you have said the E4300 has a tjunction of 85c not 100c.

*****Update******
http://softwarecommunity.intel.com/isn/Community/en-US/...
Link credit to Billyc
Edit straight from Intels mouth-
Quote:
Some steppings of the mobile Intel® Core™2 processor do indicate Tj to be approximately 85 or 100 via a single bit in the EXT_CONFIG register (msr 0EEh) but desktop, workstation and server processors do not. Nor is there a register implemented in those processors that software can read to get the Tj value for either the Pentium® 4 processor, Intel® Xeon® processors or Intel® Core™2 processors.

Folks until we actually figure out the Tjunction of our chips we are all up a creek when it comes to accurate temp measurement. TAT is only guessware.
March 26, 2007 9:52:51 PM

See, that's why I'm so confused about my temperatures, everywhere I read in this forum, we never know if temperature reported is Tcase or Tjunction, or we dont know if there is an offset or not. It all seems to depend on the motherboard, the C2D model and now it may even depend on the stepping!

How am I supposed to know??? With all these "may be offset of +/- 15C", it's not even useful to monitor temperatures.

Please, if anyone know any way of being sure of any temperature at all (no matter if it's Tcase or Tjunction or even Tjunction of 1 specific core), please help us! I dont care if speedfan or if CoreTemp report bad temperatures if I can know 1 software that reports the correct one, I'll use whatever software I have to use to finally know how hot or cold my CPU is running.
a b à CPUs
a b K Overclocking
March 26, 2007 9:56:30 PM

Pax2All, thanks for the link. It's been my experience that trying to get a straight answer from Intel, unless they're inclined to give one, is like trying to get an elaphant to tapdance. If we simply apply some logic, we can deduce that since the E4300 is not a Mobile or a Quad, it's maximum junction temperature is 85c, regardless of steppings or readable register bits. If anyone cares to prove it, unplug the cooling fan and watch it shutdown at 85c.

Comp 8)
a b à CPUs
a b K Overclocking
March 26, 2007 10:09:57 PM

TheMike, I've given everyone the information and methods to work out their temps correctly, and it DOES work. C2D temperatures aren't rocket science, so don't give up. It's all in the Guide, and I just simply don't have enough hours in the day and night to help everyone individually. I realize there's alot of information to digest in the Guide, so keep re-reading it until you understand. If you pay too much attention to everyones thoughts and comments on temperatures, then you invite confusion. Focus on the Guide. It works.

Comp 8)
March 26, 2007 11:58:38 PM

What data do you base your Tjunction formula off of? I hope it is 85c because that would mean CoreTemp 0.94 was right all along.
March 27, 2007 12:48:09 AM

Has anyone run TAT and CoreTemp 0.95 on a Quad and compared results? In theory, the results should be skewed because of the 266fsb. Anyone out there able to do this for us? I suspect CoreTemp 0.95 will read higher than TAT.
March 27, 2007 12:52:24 AM

Where does the fsb come into play?
March 27, 2007 3:52:02 AM

Glad you asked. I suspect when TAT comes upon a 266fsb Core2 CPU it assumes a Tjunction of 85c, and when TAT comes upon a 200fsb Core2 CPU it assumes a Tjunction of 100c. The only 266fsb Core2 desktop CPU with an estimated 100c Tjunction is the Core2Quad. If I am correct and CoreTemp 0.95 assigns a 100c Tjunction to the core2Quad then it will read higher than TAT under load temps. If CoreTemp 0.95 assigns a Tjunction of 85c to the Core2Quad then I'm willing to bet it and TAT will report the same temps under load.
Now we just need the kind assistance of some Core2Quad owners with Intel chipsets :D 
March 27, 2007 4:21:00 AM

You think this is what TAT always does or you think this is what TAT does when presented with a desktop core 2 processor? Because it seems that for the laptop processors, there is a CPU register that indicated the approximate Tjmax:

"Some steppings of the mobile Intel® Core™2 processor do indicate Tj to be approximately 85 or 100 via a single bit in the EXT_CONFIG register (msr 0EEh) but desktop, workstation and server processors do not. Nor is there a register implemented in those processors that software can read to get the Tj value for either the Pentium® 4 processor, Intel® Xeon® processors or Intel® Core™2 processors."

--from http://softwarecommunity.intel.com/isn/Community/en-US/...
March 27, 2007 4:54:35 AM

I think this is what TAT does with desktop Core2 CPUs.
March 27, 2007 6:43:04 AM

Remember that these monitors are all software. Any good software developer will stay away from constant definitions if at all possible. One thing I got from the Intel thread is that computing the core temps is definitely tied to knowing the tjmax (85c or 100c).

I'm just going off of what I read, but there is a bit to be checked and if on the tjmax is 85c and if off the tjmax is 100c (I think that's the way I remember it). If that information is no longer supplied by the chip (say Intel), the bit to be checked is probably 0. If core temp is displaying 100c as the tjmax, I have to assume it is used in the temp formula in displaying the cores temp.

In my rig, Core Temp .095 and SpeedFan 4.32 are equal (even though speedFan is arbitrarily modified +15c). TAT is +2c - +3c higher. IMO TAT, being an Intel software utility, has the upper hand in being correct. Also consider if you go to the problem tracking area of both SpeedFan and Core Temp, you will see several entries related to E4300, C2D and super io chip IT8712F. This does not bode well for these utilities at the moment.

If any of my logic makes sense above, I'm surprised that TAT is higher than the other 2 rather than lower. I am in no way challenging Computronix on this subject - he is the man. This is just what I've picked up over a lot of research and tried to make sense out of.

For most, like me, making modest oc's, the temps are close enough now that I can make my decisions pretty confident using all 3 utilities, but I will use TAT as my lilmiting factor, just for safety sake.

Pax2All - I wouldn't get too hungry for the extra 15c. Even if it was 100c, it would probably only make a percentage difference based on the formula.

Please poke holes in my logic where you can.The end result is, listen to Computronix if you want to stay out of trouble.

Another thing that confused me was that in the Intel forum, they were using Tjunction synonymous with Tjmax and refering to what we have been calling Tjunction as core temps. I know the difference, just seems everyone should use the same terminology when it's so confusing in the first place.
March 27, 2007 6:58:32 AM

Would I like an extra 15c cushion? You bet :) , but I'd rather have the truth. CoreTemp 0.95 upped the E4300 Tjunction from 85c to 100c to match TAT more closely under load temps. The estimated Tjunction of the E4300 according to Computronix is 85c. Something just doesn't jive here and it smells like TAT.
You said: One thing I got from the Intel thread is that computing the core temps is definitely tied to knowing the tjmax (85c or 100c).
I completely agree.
March 27, 2007 8:59:32 AM

Yea, everyones definitely not on the same page. I have a question into Arthur Liberman, author of core temp to see if he will give input to us. I got the feeling he is good at responding, I just don't know how soon. If I get anything, I'll post. I'm still leaning toward TAT for the following reason.

To get tjunction core temp I had to adjust +15c. That is the guides specified value in difference between Tcase and Tjunction. But it is not an exact value, as specified ~+15. Since core temp matches SpeedFan exactly on my unit, I wonder if that's what core temp did also, add 15 to get the approx tjunction temp. If it's a fixed constant value to get Tjunction, why even bother with two temps? All dual cores would have to be manufactured exactly alike for the temp difference to be the same for all builds, and we know that's not the case.

Further evidence is that at all times, the tjunction reported by core temp is exactly 15c over the reported tcase value reported by speedfan, both idle and load. Maybe their both right and TAT's off a couple, but it seems kind of suspicious.
a b à CPUs
a b K Overclocking
March 27, 2007 2:33:12 PM

Billyc, Pax2All and mrknowitall, although I appreciate your interest in the technical details behind Tjunction, it's a moot point. I've had these conversations with Arthur Liberman and Intel, I've studied the available reference materials, and I've tested the popular monitoring programs on Desktop and Mobile C2D's, so here's the bottom line:

There are multiple hardware, firmware, and software platform combinations. Lower and mid-level variables involve C2D steppings, FSB frequencies, super I/O chips and chipsets, BIOS and drivers, all of which can introduce offsets. At the upper level, if TAT and/or Core Temp are offset, there is no recourse. Tcase and Tjunction sampled at Idle and Load remain relative to Ambient, Vcore, CPU coolers, etc, and are predictable values.

Since SpeedFan allows for offset corrections, the desired outcome to display accurate temperatures can be achieved, regardless of other factors. When SpeedFan 4.32 was released several weeks ago, I updated the Guide and replaced Core Temp as a tool, to eliminate confusion over Tjunction, and introduce SpeedFan as a means to apply offset corrections. My advice is to dismiss Core Temp, and to use SpeedFan for monitoring.

Hope this helps,

Comp 8)

Note: TAT will typically read ~2c low at 100% Load.
March 27, 2007 6:02:48 PM

Computronix, I understand what you are saying and have no doubt that what you are saying is correct. But you keep going back to the one thing that has kept me interested in this topic - SpeedFan. For me, this utility has a number of issues:

1. My +12V is reported as 1.34V - can't be correct.
2. CPU Fan: 0 RPM, sometimes 135000 RPM, divider does not correct these values, therefore no speed fan control - it's main function.
3. Core Speeds reported w/ Tcase temps - ability to modify these values is great, if you accept the fact that 15c is always the temp difference between Tcase and Tjunction.

These items end up showing a flaw in every block of reported values in the utility, yet it seems to work. Having a career as an Analyst, I have a strong urge to get the "why" answered. So I, for one, will continue to research and ask questions, but I will continue to use your guide as the bible for testing.
a b à CPUs
a b K Overclocking
March 27, 2007 6:32:43 PM

Yes, I agree that none of these utilities are perfect, but in my opinion, SpeedFan is the best compromise for the largest percentage of users. Considering what a mess temperatures are, in terms of hardware and software, until the OEM's group their poop and get on the same page, I think SpeedFan 4.32 is a great solution.

Comp 8)
a b à CPUs
a b K Overclocking
March 27, 2007 7:14:09 PM

TheMike, sorry your thread went off on a tangent. Let's get you back on track.

Quote:
Variables:

Ambient = 21c
Chipset = 965
C2D =E4300
CPU cooler = Thermalright XP-120
Frequency = 2.88 Ghz
Load = TAT 100% 10 Minutes
Motherboard = Gigabyte GA-965P-DS3 BIOS F9
Vcore at 1.3375

Results:

SpeedFan Idle:
Temp2: 26C
Core: 31C

TAT: 45
Core Temp: 29C

SpeedFan Orthos:
Temp2: 50C
Core: 46C

TAT: 61C
Core Temp: 45C

SpeedFan TAT 100%:
Temp2: ??C
Core: ??C

TAT: 71C
Core Temp: ??C

temps are safe, but high for a e4300 @ 2.88 Ghz with almost no Vcore boost and a thermalright XP-120 and a 120mm case fan.


I've reorganized your info as much as I can, but I need you to run TAT 100% load, then post Temp2 and the hottest core, so I can analyze your temps.

Thanks,

Comp 8)
March 29, 2007 6:10:27 AM

More research has shown the quadcore theory incorrect but I did find something that might finally explain things.
Computronix you might find this interesting. I can't say I'm too inclined to do this to my rig, but more power to the guy :wink:
http://www.xtremesystems.org/forums/showthread.php?p=20...

Quote: So I have just run a little test. I am on stock fan with stock TIM and stock settings in the gigabyte BIOS.

I loaded up TAT, throttle watch, core temp (0.95 and beta 0.94) and Speedfan. I loaded up TAT to 100% usage on both cores with CPU fan spinning away happily at 1400rpm (or whatever the stock does at full.).

After 5 minutes it had loaded to about:
64/66 on TAT/Coretemp (0.95)
49/51 on Speedfan and CT0.94
Touching the HSF it was cool right down to the base and so I assumed it was doing its job properly.

So I pulled the power to the CPU and exhaust fan and watched the temps rise . After another 4 minutes this happened:
94/95 on TAT/CT0.95
79/80 On Speedfan/CT0.94

Now the HSF was roasting hot almost too hot to touch and the motherboard started beeping away for overheating. No Throttling had taken place (even thought it was supposed to) at 94/95 on TAT as they start to throttle at 85c. Speedfan and CT0.94 both registered around 80c. As soon as those two hit 80c it was beep beep and I replugged the fan and switched off TAT.

Soooo which one is correct? Which one has the temps properly pegged? Currently sat idle with case side off in ambient of around 10-12c in my room (I like a cold room) and SF/CT0/94 has it at 18/19c compared to 33/34 of CT0.95.

That would put the 3.0GHz o/c of 77'c on the TTBT (using CT0.95) at around 52c which would seem more resonable.
March 29, 2007 6:20:11 AM

What I found interesting in the above quote was not that he began throttling at 80c. I found it interesting that TAT and Core Temp 0.95 had hit around 95c. He was 5c off the programmed Tjuntion of all three programs. If we treat our temps in TAT and CoreTemp 0.95 and 0.94 like our chip has a Tjunction each assigns it (100c for TAT and CT 0.95, and 85c for CT 0.94), I think any of the 3 programs would give useable load temp values.
What do you think?
March 29, 2007 6:50:54 AM

Quote:
Computronix thats the crux of the issue.TAT and Core Temp 0.95 misreport the tjunction max of the e4300 and assign it at 100c instead of 85c. In fact the new 0.95 Core Temp build had the tjunction upped to 100c for the E4300 in its code so load temps would agree with TAT. That means the figure has real impact and is not just for information. As you have said the E4300 has a tjunction of 85c not 100c.

*****Update******
http://softwarecommunity.intel.com/isn/Community/en-US/...
Link credit to Billyc
Edit straight from Intels mouth-
Quote:
Some steppings of the mobile Intel® Core™2 processor do indicate Tj to be approximately 85 or 100 via a single bit in the EXT_CONFIG register (msr 0EEh) but desktop, workstation and server processors do not. Nor is there a register implemented in those processors that software can read to get the Tj value for either the Pentium® 4 processor, Intel® Xeon® processors or Intel® Core™2 processors.

Folks until we actually figure out the Tjunction of our chips we are all up a creek when it comes to accurate temp measurement. TAT is only guessware.



This makes perfect sense for my problem..

I have an E6300 that is showing up in TAT as a Pentium-M.

CoreTemp 0.95 shows it as having a Tjunction of 100C


Today, I did a test.. I pushed my CPU to 86C Tjunction and it did NOT activate the thermal monitor.

My Tcase was at 60, but my Tjunction was at 86.. 26C difference? Doesn't fit the ~15C model as described in the C2D Temp Guide.

If I use CoreTemp 0.94, all of a sudden the temperatures make MUCH more sense.

It recognises my chip as an 85C Tjunction, and my Tjunction temps are ~15C lower..

Details here:

http://forumz.tomshardware.com/hardware/E6300-wild-temp...


After reading the C2D Temp. Guide, I modified SpeedFan to increase the Core temps by 15C, but perhaps SpeedFan was the only one that had it accurate in the first place?


My CPUID according to TAT is 6FXh.

CoreTemp shows it as a 0x6F2.


I'm hoping this is a new batch of the C2D CPU's that are incorrectly being identified as mobile processors and this is causing problems with the core temperature monitoring..
March 29, 2007 7:03:38 AM

Pax2all

I don't see a problem, outside this seems to indicate the 100c might be tjmax. CT094 reports tcase, not really full support for c2d. Speedfan is also reporting tcase and needs a bump of +15c to compensate for tjunction. With that, TAT, CT095, speedfan and really Ct094 all agree - you started to burn at 94/95c.
March 29, 2007 9:10:14 AM

TAT and Core Temp 0.95 apply a Tjunction of 100c to the E4300. For all we know, all the L2 chips now have a Tjunction of 100c. I'm saying that whichever program you use if you use that program's Tjunction as a guide for load temps you should be fine. Core Temp 0.94 needs no offset if you view load temps with an 85c Tjunction in mind. TAT and Core Temp 0.95 need no correction if you view load temps with an 100c Tjunction in mind. They all report correctly when viewed this way.
And before someone out there gets cheeky, yes I know Tjunction doesn't really exist in the CPU, but it is a figure that (call it what you want) is used by monitoring programs to calculate temps. Absolute temps are beyond our grasp at this point.
a b à CPUs
a b K Overclocking
March 29, 2007 5:34:07 PM

Pax2All, Billyc, auldar, mrknowitall, TheMike, and to whomever else it may concern, I read the link above in message 38, but let's apply some common sense here and look at a bigger picture. There are platforms out there which work correctly. For example, an E6600 on a 975 chipset accurately reports both Tcase (CPU) and / or Tjunction (Core 0 / Core 1) temperatures with motherboard utilities, Everest, SpeedFan 4.31 or 4.32, TAT, Core Temp 0.94 or 0.95 (85c) with Tjunction ~15c higher than Tcase, without offsets.

There is no logical reason to conclude that an E4300 is a higher performance / higher temperature tolerant C2D than an E6600, since Tj max 100c is obviously far too hot, as evidenced by the post in the ExtremeSystems thread. The cores simply DO NOT run 30c hotter than the die, as this would represent obscene power disipation converted to heat, and would exceed Thermal Design Power (TDP), making the E4300 hotter than a Pentium D, which was clearly not Intel's goal for the C2D. Desktop C2D Tjunction ~15c > Tcase > Ambient. Simple concept, and it works.

Most users just want to know what proper temps should look like, how to correctly offset inaccurate temps, and don't want to be confused by the technical fine points of how Tjunction is acquired. I've tried my best in the Guide to strike a balance between simplicity, complexity, readability and comprehension. I've received many dozens of PM's and posts from users who are so technically challenged they shouldn't own the power cord to a computer, as well as from users who "get it" the first time, and correctly apply it without doubt or the need for questions.

I wrote the Guide for everyone's benefit, to make sense out of apples and oranges temperature fruit salad, so even first build beginners can understand their temps. I don't want to complicate the Guide by delving into unnecessarily diverse or complex technical details, or by writing contigencies for every hardware variant. Therefore, to further simplify the Guide and eliminate confusion, I have updated it as of 3/29/07 and deleted from the affected sections, any references to Tjunction max, throttling or shutdown temperatures, including the Scale. Also, there is no longer reference to "Core Temp" software anywhere in the Guide, as I've reached the end of my patience with the confusion it causes.

I hope this latest update meets with everyone's approval. Since this "one-size-fits-all" solution can be difficult to wear accross all platform combinations, I suggest that if anyone can build a better mousetrap, then write it up. I'll delete the Guide and logout, so you can all stuff this OEM hardware / firmware / software mess back into the binary blender, unless, of course, you would all like to spend many dozens of hours reading and correlating all the reference materials, just to arrive at the same conclusions I have as an engineer, as well as many others, including some very sharp people who reviewed, nominated and approved the Guide as a "Sticky". I can understand why wusy rarely avails himself to answer questions regarding his Core2Duo Overclocking Guide.

Please feel free to reinvent the wheel, because I'm done with this Tjunction max debate.

Comp 8)
March 29, 2007 7:09:02 PM

This will be my last post on this topic in this forum.

Quote:
Pax2All, Billyc, auldar, mrknowitall, TheMike, and to whomever else it may concern, let's apply some common sense here and look at a bigger picture. There are platforms out there which work correctly. For example, an E6600 on a 975 chipset accurately reports both Tcase (CPU) and / or Tjunction (Core 0 / Core 1) temperatures with motherboard utilities, Everest, SpeedFan 4.31 or 4.32, TAT, Core Temp 0.94 or 0.95 with Tjunction ~15c higher than Tcase, without offsets. There is no logical reason to conclude that an E4300 is a higher performance / higher temperature tolerant C2D than an E6600, since Tj max 100c is obviously too hot.


All accepted. My question goes to Speedfan - why do we need an offset w/E4300? SpeedFan is where I need to go to get that answer.


Quote:
Most users simply want to know what proper temps should look like, how to correctly offset inaccurate temps, and don't want to be confused by the technical fine points of how Tjunction is acquired. Desktop C2D Tjunction ~15c > Tcase > Ambient. Simple concept, and it works. The cores DO NOT run 30c hotter than the die. I've tried my best in the Guide to strike a balance between simplicity, complexity, readability and comprehension. I've received many dozens of PM's and posts from users who are so technically challenged they shouldn't own the power cord to a computer, as well as from users who "get it" the first time, and correctly apply it without doubt or the need for questions.


I would have been completely lost without the guide to begin with. Not being very knowledgable in this area, I had to read it several times to get it in my head, but all the necessary info was there without being overly technical. I guess I just fall out of the "most users" category and want to learn more. I thought that these topics were to ask questions and resolve problems. My posts should probably have been in another forum category, such as software/other software, so as not to appear to be a challenge to the guide. I would like to know why I need a modifier in speedfan for the E4300? Why are my volts wrong in sf? Why does core temp label Tjmax as Tjunction? Why is TAT a little higher than other reported core temps? These questions I will have to go to the developers to get answers (which I have done, or at least requested).

I really appreciate all your hard work and time put into the guide and these forums. I'm sorry for your frustration but the guy with all the knowledge always gets hammered.
March 29, 2007 11:11:13 PM

Computronix no one is saying these Tjunction temps are right. 85c is far more reasonable than 100c. I am simply trying to get across that TAT and Core Temp 0.95 use the 100c figure. I am not trying to offend. But in the end I think the important thing is that all of these programs appear to be proportional in reported load temps when you look at the Tjunction they assign. Dont ask me why some use 100c as a base to calculate from. It could fall back to TAT originally being written for the mobile platform. Who knows. We really do appreciate all the work and time you have put into bringing the C2D temperatures to the forefront and providing a place where cognitive debate could happen.
Thanks :trophy:
a b à CPUs
a b K Overclocking
March 29, 2007 11:29:53 PM

Quote:
There are platforms out there which work correctly. For example, an E6600 on a 975 chipset accurately reports both Tcase (CPU) and / or Tjunction (Core 0 / Core 1) temperatures with motherboard utilities, Everest, SpeedFan 4.31 or 4.32, TAT, Core Temp 0.94 or 0.95 with Tjunction ~15c higher than Tcase, without offsets.


E6600 with 975 chipset shows Core Temp 0.94 and 0.95 Tjunction 85c.

I apologize for not making this clear.

Comp 8)
March 29, 2007 11:49:51 PM

To all those confused when I say Core Temp 0.94 and 0.95 and TAT all report proportionally correct temps according to their respective Tjunction values:
I put my comp to load by running TAT

Core temp 0.94 reads 65c (65/85=76.4% of 85c)

CT0.95/TAT read 78c (78/100= 78% of 100c)

a difference of only 1.6%

Do the math on your own machines and I think youll see what I mean.
March 30, 2007 12:12:53 AM

Quote:
To all those confused when I say Core Temp 0.94 and 0.95 and TAT all report proportionally correct temps according to their respective Tjunction values:
I put my comp to load by running TAT

Core temp 0.94 reads 65c (65/85=76.4% of 85c)

CT0.95/TAT read 78c (78/100= 78% of 100c)

a difference of only 1.6%

Do the math on your own machines and I think youll see what I mean.


This sums up my thoughts pretty well..

The core temperatures appear to be reported as basically a converted percentage of Tjunction max, so for any CPU that shows up in TAT as a Pentium-M (or in CoreTemp as Tjunction max of 100C), it would probably be best just to look at Tcase and add 15C to that temperature rather than using any application that reads the temperature of the cores.

It looks like the existing applications are not able to accurately represent the core temps of these CPUs that show a Tjunction max of 100C.

It's *possible* that SpeedFan without offsets, and CoreTemp 0.94 are accurate for these Tjunction temps, but given the proven thermal characteristics of these CPUs, adding 15C to Tcase is probably the most accurate since it is a linear representation (the converted percentage is not linear as it approaches Tjunction Max).


EDIT: Actually, I just looked at a screenshot of a workmates E6300 running TAT, and his CPU is also being recognised as a Pentium-M, and yet his core temperatures from TAT are being reported about equal to his Tcase in SpeedFan (SpeedFan is also showing the cores about the same as Tcase - he hasn't applied any offsets). While I can't explain that result, my point is that just because TAT shows it as a Pentium-M, it doesn't mean it's going to have a reported Tjunction max of 100C.

I think we're just going to have to take it with a grain of salt that there are some CPUs that appear to have a Tjunction max of 100C with current software, and their core temps are going to be reported as a converted percentage of the proximity to Tjunction max, which will be 15C out at Tjunction max, and more the further away you get from max.
March 30, 2007 12:24:43 AM

It's actually not the percentage that stays the same. It's the delta in degrees that is the same.
March 30, 2007 3:43:21 AM

Take your load temps across the different programs and divide them by the assigned Tjunction value of that program. The percentages should be nearly identical. Lets keep it simple folks.
Note: Core Temp 0.94 and 0.95 clearly show the Tjunction value being used
Note2: It appears the newer L2 stepping E6xxx chips are being read in TAT and Core Temp 0.95 as having a 100c Tjunction value
!