C2D temps revisited....

Discussion in 'Overclocking & Cooling' started by TheRapture, Mar 31, 2007.

  1. cornelious0_0

    cornelious0_0 [H]ardForum Junkie

    Messages:
    12,783
    Joined:
    Apr 6, 2003
    It shouldn't matter TOO much, unless he's got one of the earliest released E6600's it should be a B2 revision with a tJunction of 85C, and typically TAT and Speedfan/Core Temp 0.94 all report about the same, it's only with most L2 revision chips that there's typically a big difference in temps reported by different programs. Most people I talk to with B2 revision chips don't see any difference between 0.94 and 0.95 Core Temp readings.
     
  2. Ste

    Ste Gawd

    Messages:
    574
    Joined:
    Nov 27, 2006
    Hrmm, like I said 54C max load.

    Program: CoreTemp

    The CPU is a revision B..
     
  3. cornelious0_0

    cornelious0_0 [H]ardForum Junkie

    Messages:
    12,783
    Joined:
    Apr 6, 2003
    Yeah, B2 revision C2D's typically report the same between TAT and Core Temp .94/.95...at least any that I've seen. Heck, it wouldn't matter, because TAT would report higher then Core Temp 0.94 or Speedfan and if TAT says 54C fully loaded then you've got nothing to worry about. :cool:
     
  4. Ste

    Ste Gawd

    Messages:
    574
    Joined:
    Nov 27, 2006
    Good to know.

    I was stable At 3.41 Ghz for a time (Everything else stable, testing, 36mark06, gamming, normal use while SMP F@H). However SMP Folding would be unstable at those settings( At night, when it was the only thing running). So i started over, and I am now seeing if SMP Folding will remain stable. If SMP causes restart or the core processes disappear in the middle of the night I up the vcore until it no longer happens.

    My temps are probably as good as they are because of how I positioned the Heatsink. I have it rotated so the built in fan on it blows out right at another case 120MM exhaust fan. Not only that I have more Inflow fans than I do out.
    I also have a tornado pointed right at my northbridge.

    One of these days I will get pics.
     
  5. cornelious0_0

    cornelious0_0 [H]ardForum Junkie

    Messages:
    12,783
    Joined:
    Apr 6, 2003
    Interesting sounding setup.

    I haven't had any issues folding, but my fiance had Sims 2 lock on her twice tonight after the system had passed 10 hours of Small FFTs the day before. Turns out I had only done about 6 hours of an Orthos Blend test for memory stability initially but upon running 2 instances of HCI Memtest each loading 750MB of memory, errors started popping up around 100% coverage. Upped the memory voltage to 2.2v or so and all's well so far, 820% coverage and counting, letting it hit 1000% out of habit.

    There also might be a chance I can clear 3.2GHz on stock cooling, as I've found a couple guys on XS that have run much higher voltages under the stock cooler with higher room temps then mine and not had the shutdown issue I've run into, so hopefully it's a NB/SB voltage and I can push for 3.3GHz or 3.4GHz on stock cooling fully stable. :cool:
     
  6. Ste

    Ste Gawd

    Messages:
    574
    Joined:
    Nov 27, 2006
    Have you been watching the temps? I know with stock cooling you can still get an adequete OC. But, its stock cooling, Lol, So Maybe it is the temps?

    Dunnue.
    Gl to ya.
     
  7. cornelious0_0

    cornelious0_0 [H]ardForum Junkie

    Messages:
    12,783
    Joined:
    Apr 6, 2003
    No, it's weird...when it shuts down it's only at 62-65 degrees.

    I'm almost wondering what normal behavior is when a CPU hits an FSB wall, because I have seen/heard of a few people hitting FSB walls between 350FSB and 400FSB with their E4300's. I sure hope that's not what it is, and I know that with decent room temps I should be capable of more then this even with the stock cooler...guys on XS have done 3.6GHz /1.525v or so with the stock cooler and been fine, so I think it's either a motherboard voltage/setting or possible NB temps, seeing as I was idling high from the start and this NB heatsink isn't flat at all.

    I dunno, I'll keep looking into it, gonna see if I can get info from people who have hit FSB walls to see if they've seen similar behavior or not.

    I'll report back with my findings, but most likely in my "Superchip" thread.
     
  8. ccityinstaller

    ccityinstaller 2[H]4U

    Messages:
    3,903
    Joined:
    Feb 23, 2007
    After seeing my cpu hit 70C under load using CoreTemp 0.95, I really started to wonder if my cpu has a tjunction of 85C instead of 100C. I down clocked my cpu to1.2Ghz ( 6 x 200), however I couldn't undervolt my cpu. At the stock Vcore and idling for 5 minutes, my temps according to CoreTemp 0.94 were 22C, which is 3F higher then my actual room temp. I have decided to trust 0.94 from now on. Here is a link displaying the differences.
     
  9. cornelious0_0

    cornelious0_0 [H]ardForum Junkie

    Messages:
    12,783
    Joined:
    Apr 6, 2003
    Tough to say using stock vcore...but it's a fair enough bet.

    I wouldn't be 100% satisfied with that though, try and find someone with similar room temps/cooling and see if you can compare temps at stock voltage.
     
  10. unclewebb

    unclewebb [H]ard|Gawd

    Messages:
    1,799
    Joined:
    Jun 21, 2006
    If you can't manually lower your core voltage when trying to do a low temp test you might try setting the FSB MHz to default in the bios as well as the multiplier and set the core voltage to default or AUTO.

    Enable C1E and SpeedStep and after your computer boots up and is idle the multiplier should drop to 6X and the core voltage should also drop as long as your motherboard supports C1E properly. My P5B Deluxe idles at 1.304 volts with C1E disabled and will idle at 1.144 volts automatically when C1E is enabled. You can use RMClock to toggle C1E on and off.

    [​IMG]

    If C1E works on your board you'll have meaningful results within a few minutes after booting up.

    With a room temp of 21C this evening it's pretty clear that SpeedFan and CoreTemp 0.95 which both assume a TjMax of 85C are correct.

    There are new E6600 cpus out with the same revision B2 as the original E6600 but some now have a TjMax of 100C instead of the original 85C. You really can't assume too much when it comes to guessing at TjMax.
     
  11. cornelious0_0

    cornelious0_0 [H]ardForum Junkie

    Messages:
    12,783
    Joined:
    Apr 6, 2003
    Good call unclewebb, I (as well as most other readers here) continue to value and appreciate your input in this thread and others.
     
  12. SuperKeijo

    SuperKeijo Gawd

    Messages:
    705
    Joined:
    Jan 18, 2007
    Unclewebb:

    I must say I think you are dead wrong with the assumption that your or anyone elses c2d for that matter has a tjunction of 85c. And that some c2ds have a tjunction of 85 and some 100. If that were true, cpus with a real tjunction of 100 should show similar temps at same speeds and voltages at idle with coretemp reporting tjunction 100 (0.95 for e4x00 series) than cpus with a real tjunction of 85 with coretemp reporting tjunction of 85. BUT that is not the case. My e4300 reports similar idle temps than your 6400 with 0.94 coretemp (tjunction =85, like yours) but with 0.95 which gives a tjunction of 100, I get ~15 degrees higher idle temps than you. And this is the same with all c2ds, they all give you about similar temps in 0.94 but with 0.95 the e4 series for some reason or another is assumed to have a tjunction of 100, WHICH I BELIEVE IS THE CORRECT VALUE FOR ALL C2Ds. Why? because if 85 was right, that would mean that the cpu were producing next to no heat at all. I'm talking like 1-2 watts in idle. 1-2c higher core temps than the temps OUTSIDE the case just sounds pretty rediculous. Compare that to previous gen cpus. No other cpu idles at 1-2c higher than the ambient temp outside the case. ~17 degrees higher than ambient just makes a whole lot more sense to me. AND a ~2GHz cpu overclocked at ~3-3.3GHz which is a HUGE overclock at full load reaching temps of 80+ does not seem unbelievably high to me at all. Old pentium 4s at 3,5ghz+ ran very hot too, and modern gpus run at over 80c easy too. Sinse intel hasn't shead any light on the tjunction values, how can we assume it's 85? And that test of yours where you underclock and volt the cpu to ~1,6 and see the idle temps is also bogus. So what if the result says 1-2c higher than ambient case temps? that doesn't say that 85c tjunction is right. Imo it just proves it's wrong, because I do not believe in idle temps (monitored INSIDE the cores themselves) of 1-3c higher than the surrounding air. ~16-18 makes a lot more sense, do you disagree?

    Based on what? the fact that it runs 1c higher than the surrounding air is proof to you?

    The testing I did here is pretty clear imo:

    http://hardforum.com/showpost.php?p=1030995819&postcount=98

    Even the heatsink itself showed 5-7c degrees higher temps than coretemp 0.94 reported with tjuncion of 85c, SO 85 CANNOT be right, and here with kind of cold ambient temps the ambient temp and the temp 0.94 reports were very close, just like with your cpu too (1-2c higher than ambient).

    Where as the 0.95 reporting ~9-10c higher than heatsink temps makes MUCH more sense to me. The temps inside cores are certain to be higher than heatsink temps because its the cores that heat the heatsink and heat is transfered quickly away from cores, so from cores to heatsink a temp drop of 9-10c makes sense to me. I know I had a higher cpu speed and higher voltages than you did but the outcome is the same regardless. If 0.94 ct reports lower temps than heatsink, then it's wrong regardless of cpu speed and voltages.

    So I do believe my testing (and common sense) proves tjunction of 85c to be wrong on not only allendale c2ds but all c2ds in general because all including the 6600-6800 show similar idle temps (22-27c, this ofcourse varies abit depending on ambient temps and overclocks) so I do believe they all have the same tjunction, not like you report that some might have 85c and some 100.

    Please, prove me wrong.
     
  13. cornelious0_0

    cornelious0_0 [H]ardForum Junkie

    Messages:
    12,783
    Joined:
    Apr 6, 2003
    There seems to be a growing debate on this topic, as I've also run into a couple guys on XS that believe this testing method to mean nothing. I'm fine "assuming" that I've got a 85C chip...as the temps make much more sense compared to other results I've seen...and it does not make sense that the CPU would run 15C above ambient @ 1.2GHz 1.1v...but that's just me.

    I do NOT want to start any kind of argument, but we each have our beliefs, and until Intel does release hard evidence on what the story is with these chips it's hard to say one way or another.
     
  14. SuperKeijo

    SuperKeijo Gawd

    Messages:
    705
    Joined:
    Jan 18, 2007
    The thing is, as I stated that you do not know the tjunction of the other cpus youve compared to, so how can you assume anything? And if youve had any other cpu besides c2d in the past few years you'll know that they do not idle at 1-3c above ambient temps, and c2ds have actually 2 cores side by side making it seem even more unbelievable that they idle so close to ambient temps. And with my testing at cold temps I actually got the other core to idle 1c colder than ambient!! I'll try some more testing trying to get the cpu to very slow speeds and low voltages.
     
  15. cornelious0_0

    cornelious0_0 [H]ardForum Junkie

    Messages:
    12,783
    Joined:
    Apr 6, 2003
    Alright, well, I won't completely disprove/ignore the fact that my CPU may be running as hot as Core Temp 0.95 displays...but it's one or the other, and other then unofficial testing and assumptions I guess we'll just have to wait until Intel has the final say and tells us what's what.
     
  16. unclewebb

    unclewebb [H]ard|Gawd

    Messages:
    1,799
    Joined:
    Jun 21, 2006
    You definitely could be correct that all Core 2 Duos have a TjMax of 100C but here's what I've observed that doesn't agree with that.

    I have access to a Dell laptop with a Core Duo processor which uses the same digital thermal sensor that the Core 2 Duo uses. These processors have an Intel documented TjMax of 100C. When coming out of Stand By mode first thing in the morning, they start up very quickly and within seconds you are at the Windows Desktop. CoreTemp 0.95 always reports an initial temperature within 1C or 2C of ambient temperature.

    When I start up my E6400 Core 2 Duo in the morning it also starts very rapidly from Stand By mode and using an assumed TjMax of 85C it also first reports a core temperature within a degree or two of room temperature. If TjMax was actually 100C for my C2D then the first reported temperature would be 16C to 17C higher than ambient temperature.

    That doesn't make any sense that two very similar processors would act completely different when first started. The heatsink and fan on a desktop processor can dissipate more heat and should heat up slower than a laptop processor, not way faster.

    I'll look at acquiring an IR temperature gauge like you have for some further testing. Maybe I'll do something really crazy like pop the heatsink and fan off my C2D while it's running so I can get in there for some "real" core temperatures! At idle it shouldn't get too hot. :D

    The idle temps of previous Intel processors like the Pentium 4 can't be compared to the Core series of processors. The P4 cores were heat monsters and didn't have the same technology as the core series.

    As for getting a reported core temperature 1C lower than ambient that is quite possible when you are dealing with integer numbers and not knowing how real numbers are being converted to integers displayed by the DTS. I think I posted about that in the other thread you mentioined.
     
  17. SuperKeijo

    SuperKeijo Gawd

    Messages:
    705
    Joined:
    Jan 18, 2007
    I took this thing to the "next level" so to speak :)

    This was my setup:

    [​IMG]

    Ambient temp recorded from the cardboard box next to cpu heatsink.

    So, lets get started

    [​IMG]

    As you can see, settings are 600MHz on the cpu and 0.675v on the core. Now to put this in perspective, thats about half of stock volts and a third of stock speeds (and just 1/5 of 3GHz).

    Ambient temp:

    [​IMG]

    heatsink temp:

    [​IMG]

    And finally coretemp 0.94 and 0.95 results:

    [​IMG]

    @ my usual 24/7 settings (3040MHz, 1.365v) taken right after previous testings:

    ambient:

    [​IMG]

    (hasn't changed)

    Heatsink (1.6 degrees higher allready):

    [​IMG]

    And coretemps:

    [​IMG]


    So as we can se here, temps drop a rather dramatic 5c on just idle from 3040MHz, 1.365v to 600MHz and 0.675v. That alone is proof that at ~3ghz speeds it is impossible to achieve 1-3c higher temps in cores compared to ambient.

    Previously I had taken ambient temps from the floor, but I discovered that just 30cm higher from the floor the ambient temp is ~1-1.5c higher! So now we're getting below ambient temps with 0.94 even with the 3GHz setting. And with 600MHz we have over 6c below ambient!!

    0.95 on the 600mhz gives ~8c higher temp compared to ambient and ~6c higher compared to the heatsink reading. Now this is something that's much easier to believe. If we could run the cpu at ~50MHz and 0.1v I'm sure we'll be within ~1c differences.

    3040MHz we have 0.95 at ~9-10c higher than heatsink and ~12c higher than ambient. Again, realistic results taken voltage and cpu speed into consideration. 0.94 ~2c below ambient.

    Just playing around and having the AC blowing from ~1.5m towards the computer. Ambient dropped to about 15-16c:

    [​IMG]
     
  18. unclewebb

    unclewebb [H]ard|Gawd

    Messages:
    1,799
    Joined:
    Jun 21, 2006
    Could you try running SpeedFan when you are at 600 MHz and low volts. CPUz doesn't report core voltages correctly when the volts are too high or too low.

    Your testing certainly backs up your opinion that all C2Ds have a 100C TjMax. Time for some more testing.
     
  19. SuperKeijo

    SuperKeijo Gawd

    Messages:
    705
    Joined:
    Jan 18, 2007
    Ok, here we go @ 600MHz and 0.675 in bios:

    [​IMG]

    Unclewebb, could you try something similar? I mean lower the cpu speed to insanely low frequencies and voltages and do some testing? something like 900MHz and 0.7v should work pretty easily atleast. Would be interesting to see if you can get coretemp 0.94 values below ambient when running very low voltages. Thank..

    edit. And the lower the ambient temps the better! (I think)
     
  20. jmackay

    jmackay 2[H]4U

    Messages:
    3,929
    Joined:
    Nov 23, 2006
    Couldn't temps actually go below ambient? Isn't that the point of fans in the first place? To cool areas that are warm?

    I mean I could turn on my desk fan and point it at my hand and my hand would get cold, even though the room is the same temp, and my hand is producing heat.

    So couldn't fans that blow down onto the cpu area read lower than ambient?
     
  21. unclewebb

    unclewebb [H]ard|Gawd

    Messages:
    1,799
    Joined:
    Jun 21, 2006
    Unfortunately the lowest voltage I can get to is 1.104 volts. I think the lowest I've been able to get the MHz to is around 950 MHz using ClockGen.

    The wattage or heat a processor puts out is proportional to the square of the voltage. Running your C2D at 0.69 volts is putting out less than 40% of the heat that my C2D does at 1.104 volts.

    I have a Tuniq Tower that will hopefully show up here tomorrow. Anandtech dropped his idle temps by 14C when he bolted one on. If I can lower my idle temps by even 1C or 2C during my test then they will go below ambient and my low temp test will have a big fat hole in it. I will definitely be acquiring an IR thermometer as well to back things up.

    Thanks SuperKeijo for all your testing. It's really going to help me out and get us to the bottom of this great debate.
     
  22. unclewebb

    unclewebb [H]ard|Gawd

    Messages:
    1,799
    Joined:
    Jun 21, 2006
    No. A heatsink and fan can remove heat but it's impossible to go below ambient.
     
  23. SuperKeijo

    SuperKeijo Gawd

    Messages:
    705
    Joined:
    Jan 18, 2007
    Cpu core temperatures cannot go below the temperature of the surrounding air using air or water cooling. The reason is that the cpu outputs energy (in the form of heat) at a certain rate (example 50 watts, means that 50J of heat energy is produced every second). Differences in cooling performance comes from the speed that heat is transfered away from the cpu. Watercooling is efficient because water has a very high heat capacity. It means that water can store high amounts of energy without warmimg up too much, thus transfering the heat energy quickly away from the cpu core.

    Yep. I'm pretty much certain that the tj cannot be 85c for c2ds. Ofcourse it isn't necessarily 100 either but that would be my best guess atm anyway. Maybe it is 95? maybe 105? who knows. But 85 it really cannot be. Maybe in the year 2010 Intel will actually tell us the truth LOL.
     
  24. Ste

    Ste Gawd

    Messages:
    574
    Joined:
    Nov 27, 2006
    Update:
    Vcore 1.53V
    Clock 3.357 Ghz
    StressCPU2 Temp == 58C

    So whats the word? Can I Go over 60C Load or not?
    I ask because I will eventually work back to 3.402 Ghz and the Vcore will probably be around 1.55 or 1.57, therefore temps would probably be above 60C max Load..
    (Others temps are
    Complete Idleness: 25C
    SMP Folding but otherwise Idle: 45C
    SMP Folding and Gamming: 50C
    )
     
  25. Evoic

    Evoic Gawd

    Messages:
    856
    Joined:
    Apr 30, 2007
    I just posted a question about C2D temps here: Any Help Is Appreciated.

    Is this in line with the conversation of this thread or am I completely off base?

    Thanks!
     
  26. BillParrish

    BillParrish [H]ardness Supreme

    Messages:
    7,520
    Joined:
    Aug 25, 2006
    Intel disagrees with you and publicly publishes the fact.
     
  27. Ste

    Ste Gawd

    Messages:
    574
    Joined:
    Nov 27, 2006

    How about answering Evoic and I?

    (Not trying to be rude, more of a Bump type Post)
     
  28. SuperKeijo

    SuperKeijo Gawd

    Messages:
    705
    Joined:
    Jan 18, 2007
    And where is this fact?
     
  29. cornelious0_0

    cornelious0_0 [H]ardForum Junkie

    Messages:
    12,783
    Joined:
    Apr 6, 2003
    I'm also curious to see a link for this, as it would completely demolish any/many arguements/debates on the topic.
     
  30. unclewebb

    unclewebb [H]ard|Gawd

    Messages:
    1,799
    Joined:
    Jun 21, 2006
    SuperKeijo: You've opened up a big can of worms but the more I look at your numbers and compare it to the data I've gathered, the more I have to agree with you.

    Bill, can you please post a source where Intel has published this fact.

    I was pouring through the C2D datasheets last night and I can't find out where this 85C number originated from. It's almost as if people have said it enough times that it has somehow become fact but after seeing SuperKeijo's results, I'm having serious doubts.

    I know in the mobile processors there is an Intel documented bit within that processor and if it was set then you had an 85C TjMax processor. The Intel rep Lexi S. has said that bit doesn't apply to the C2D desktop processors. That bit is set in my processor so maybe that's why the programmer of CoreTemp assumed that my original revision B2 Conroe core E6400 has a TjMax of 85C.

    I believe SpeedFan 4.32 with core temperature monitoring came out after CoreTemp so perhaps he made sure his results were the same as CoreTemp by choosing a TjMax of 85C as well.

    My E6400 is listed as having an Extended Halt ( C1E ) power of 22 watts. When I have C1E enabled my CPU idles at 6 x 266 = 1600 MHz and the core voltage drops to 1.144 volts as reported by CPUz and confirmed by SpeeedFan.

    When C1E is enabled my reported core temperature based on a TjMax of 85C drops down exactly equal to room temperature. How can that be true? It can't be. At this level it is still producing the exact same amount of heat as a 22 watt light bulb. It would have to be down to zero watts before the core wasn't producing at least some heat.

    My low MHz / voltage test produced some nice numbers that I used to assume a TjMax = 85C but logically, knowing how much power a C2D still puts out at this level, it just doesn't make any sense and isn't valid. I originally thought that getting down to 1200 MHz and 1.1 volts was enough to get all the heat out of the processor core but that's not true and is confirmed by the Intel C1E specs. At this level a C2D is still putting out 22 watts of heat energy.

    The E4300 has an Extended Halt ( C1E ) power of 12 watts. I believe Intel was able to lower the rated wattage by dropping the core voltage during C1E idle from approximately 1.15 volts to 0.85 volts. The 0.85 number comes from the Intel datasheet as the minimum voltage a C2D is designed to run at. It would be interesting if someone with an E4300 or one of the C1E 12 watt processors ( Allendale E6300 / E6400 ) could confirm that by enabling C1E and reporting their voltage and MHz at idle. I don't know if a 12 watt chip would be able to drop its voltage down that low on my board but SuperKeijo's Gigabyte board seems to properly support low voltages like that.

    Even when SuperKeijo dropped his voltage and MHz down as low as he could go, his E4300 would still be putting out approximately 8 watts of heat. That's not a huge amount but I would think that it's still enough to warm his core temperature up above ambient. A kid's night light that plugs into the wall and has a small automotive sized bulb in it is rated at approximately 10 watts and they definitely still generate a small amount of heat.

    I think in this great heat debate, way too much has been assumed since day 1. Time to go look for an IR Thermometer and prove or disprove some of these myths.
     
  31. SuperKeijo

    SuperKeijo Gawd

    Messages:
    705
    Joined:
    Jan 18, 2007
    Unclewebb, my thoughts exactly! I have no degree in physics or anything but I do know that if the cpu is generating heat at even 10w, in order to maintain a delta of 1-3c compared to ambient air temp, the heat would have to be transformed INSANELY fast from the core to air.

    I'll put the cpu at atock and see the idle voltage using c1e..

    EDIT. Ok, here I have everything at stock.. Load and idle voltages:

    [​IMG]
    [​IMG]

    I think it's trying to use 1.15v here (at idle), atleast with this mobo..

    BTW unclewebb, didn't you have an L2 (allendale) 6400? It's c1e power is 12w reported by intel.

    L2 e6400
    B2 e6400
     
  32. Warl0rd

    Warl0rd Limp Gawd

    Messages:
    170
    Joined:
    May 28, 2007
    [​IMG]

    I have EIST and C1E enabled, all voltages set to manual at normal/default.
     
  33. unclewebb

    unclewebb [H]ard|Gawd

    Messages:
    1,799
    Joined:
    Jun 21, 2006
    The funny thing is that I have a revision B2 E6400 with a C1E power rating of 22 watts at idle which everyone is 100% certain has a TjMax of 85C. At the moment I'm not convinced but I'm determined to find out.

    I just got home with my shiny new Fluke 62 Mini IR Thermometer and I'm already learning.

    With my computer in Stand By mode my Kill-a-Watt meter shows a draw of 6 watts and some of those watts are going to the northbridge and to the southbridge and even to the processor. The cpu heatsink is already holding steady at about 3C above ambient and I haven't even started my computer yet!

    With my computer at idle with C1E enabled, the reported CoreTemp is 4C or 5C less than the heatsink temp which of course is impossible.

    Other ICs on the board like on my wireless card are up over 40C while CoreTemp is reporting 22C for my processor. With a TjMax of 100C, my core temperature would be reported as 37C which starts to look a lot more believable.

    I still need to do a lot more testing but already the standard line of "it must be 85C" doesn't seem to be true.
     
  34. jmackay

    jmackay 2[H]4U

    Messages:
    3,929
    Joined:
    Nov 23, 2006
    So does that mean that my 6600 doesn't have a TjMax85 and is 100? So its not running @ 44c its running @ 69c?

    TAT, CoreTemp .94 and .95 all say its running @ 44c.
     
  35. Warl0rd

    Warl0rd Limp Gawd

    Messages:
    170
    Joined:
    May 28, 2007
    How much does it say when at idle? if its close to ambient temp then your load temp should be higher then you think...
     
  36. jmackay

    jmackay 2[H]4U

    Messages:
    3,929
    Joined:
    Nov 23, 2006
    30ish... and its not that hot in here but I don't have anything to test with.
     
  37. unclewebb

    unclewebb [H]ard|Gawd

    Messages:
    1,799
    Joined:
    Jun 21, 2006
    Same here. With revision B2 processors those 3 programs are all pretty much in agreement but that doesn't mean any of them are right.

    TAT was programmed to read temps from the core mobile chips. It has never been updated to handle the C2D desktop processors. Most users assume it is right because it has the Intel logo on it but it hasn't been proven for desktop processors.

    As I said before there is a single bit of data stored in the core mobile processors that determines whether a mobile processor has a TjMax of 85C or 100C. Software can read this bit in a C2D desktop processor as well but the interpretation of this information is meaningless. Intel has confirmed that this bit has no meaning in a desktop processor.

    If TAT uses this bit to interpret whether a C2D mobile processor is a TjMax 85C processor then it could easily be guessing wrong when it comes to a desktop processor. Using CrystalCPUID I was able to test this bit in my revision B2 E6400 CPU and it is set so I can understand why TAT would assume TjMax = 85C.

    If TAT is wrong, then the programmer for CoreTemp could do the same thing and make his program the same as TAT because TAT must be right and then SpeedFan could come along and do the same thing again.

    At this point, anything is possible. I'll try to stop guessing and get back to more testing.
     
  38. unclewebb

    unclewebb [H]ard|Gawd

    Messages:
    1,799
    Joined:
    Jun 21, 2006
    It's time for a timeout. Using an IR thermometer near the edges of a heatsink can produce some very inconsistent results.

    I pulled the plug on my computer and let it sit for over an hour with absolutely no power going to it so everything would return to the ambient temperature.

    On flat surfaces the reported IR temperatures were very consistent throughout the motherboard area. However, with the computer not turned on, the temperature at the edge of the northbridge heatsink could vary by up to 6C over ambient. That's not right. Pointing towards the round heatpipe also gave some inconsistent results.

    There's definitely more to using an IR gun than pointing and believing every number that pops up.

    I restarted the computer and set things on low volts and MHz. In order to get a clear shot at the core I removed and disconnected the cpu fan. It took a while but after about 10 minutes CoreTemp had stabilized at 63C with the OEM heatsink in fanless mode. :D

    The copper insert in the center of the Intel heatsink is shaped like a small vase.
    [​IMG]

    Pointing the IR thermometer in here and trying to get a reliable, repeatable reading was absolutely impossible. At a CoreTemp reported 63C, based on a TjMax of 85C, I ended up getting some IR readings of 28C when I tried to move in closer to the middle which is totally wrong. I couldn't hold my finger on this cup for very long so obviously it wasn't 28C.

    Near where the fins meet this copper plug I would get temperature readings of ~52C but moving a little closer to the center totally screwed up the readings. Even a slight angle towards the lower inside of this cup and the readings would be jumping around with only a slight movement of my hand. I don't think the copper material is causing the problem. The shape of this copper plug seems to result in meaningless results, for me anyhow.

    I'm still waiting on a new Tuniq Tower for some further testing. With the Tuniq fan removed, I should have a straight shot with the IR at the flat copper plate that is attached to the CPU. Hopefully I'll be able to get some more meaningful results then. In the meantime, ignore all my previous theories and brain storming because nothing has been proven yet for the revision B2 Conroes.
     
  39. Ste

    Ste Gawd

    Messages:
    574
    Joined:
    Nov 27, 2006
    Bah, this is all so confusing.

    All I know is that Im SMP Stable, and as for the temps, after reading that I honestly don't know what my coretemp reading actually Means.

    Says Tjunction max of 85C and shows 57C on max Load.

    w/e. Ill keep watching this thread I guess.

    I should probably get a temp probe..
     
  40. unclewebb

    unclewebb [H]ard|Gawd

    Messages:
    1,799
    Joined:
    Jun 21, 2006
    While waiting for my Tuniq, I think I'll strap on a 7000 rpm Pabst fan tomorrow. It's from the good old AMD Athlon days but it moves plenty of air. If I can push the reported core temp below ambient with air cooling, that would confirm that TjMax = 85C is wrong.
    You might want to hold off on that. I think if enough time and money is thrown at this problem that sooner or later we're going to learn something.