Please report all spam threads, posts and suspicious members. We receive spam notifications and will take immediate action!
Results 1 to 10 of 10

Thread: Digital Thermal Sensors (DTS) - seeking themal peace of mind




  1. #1
    Chike is offline Senior Member
    Join Date
    Jun 2009
    Location
    Israel
    Posts
    6,531

    Default Digital Thermal Sensors (DTS) - seeking themal peace of mind

    There is more unknown than known, or not enough known, about the DTS, and everyone has a different theory about them.
    How they work, how accurate are they, should er use them, and how? We will try to understand together in this thread.

    An excelent source would be the A direct Tc measurement experiment on a G0 Q6600, experiment conducted by gampuo. He is an active member of the form and may be able to add more information. There are also some other valuable links in his thread.
    Some of the results(post #3) are puzzling, but looking at the IBT test there is ~10c difference between CPU temp and DTS for all 4 of them.
    According to the controversial article(not realy j/k) at tom's hardware, "Tcase" is equal to (X + Y) / 2 + ambiant, at idle, where X is Idle power and Y is cooler efficiency. For the G0 Q6600 with stock cooler X = 4 and Y = 6, Tcase = 5 + ambiant. It would be good to know how far were the CPU temperarures readout from this value, at the time of the experiment. A -5c offset in CPU temperature may support the 5c DTS to diode gradient theory(for now).

    Any more input or data provided would help to expand ower knowledge, and get a better understanding of what actions must be taken to make a more reliable use of the DTS sensors.

  2. #2
    Join Date
    Apr 2009
    Location
    J'Habite En Angleterre
    Posts
    2,431

    Default Re: Digital Thermal Sensors (DTS) - seeking themal peace of mind

    I will be trying load calibration later today, when I have a spare few hours.

    I thought of including the following:

    Non-calibrated readings at stock levels with auto Vcore
    Non-calibrated readings at an easy to achieve at rated voltage clock of 3.6GHz
    Non-calibrated readings of my current overclock with a moderately high over volt (for air cooling)
    Calibrated readings from all three of the above after following the calibration instructions in the link provided by Chike.

    As you are all aware, my DTS sensors are stuck. They do move uniformly after the following values are reached:
    Core 0 = 34c (66 to TJmax)
    Core 1 = 39c (61 to TJMax)

    If there is anything else you would like including, feel free to ask.

    It would be most useful if someone is able to do similar before and after calibration tests using both idle calibration and load calibration (on the same CPU) to see what the difference is in values reached.

    As you have a similar CPU to me Chike, our results may be comparable. However please note that E7600, E7500 and E7400 processors are all "R0" stepping and all E7200, E7300 are "M0" stepping for some reason. It may be that some higher numbered E7xxx have an M0 stepping and visa versa due to Intel binning policy, but I have yet to see that. Possible though, similar to how AMD used cut down "Toledo" cores in some of their X 3800+, which were designed on the "Manchester" core.

    I also have access to a Q8200 and and the aforementioned X2 3800+, of any data from those would be of use.
    Coolermaster CM 690 II advance Case
    Corsair HX750 (CWT, 91%(80+ Gold rated @230V) single 62A 12V rail
    P55A-UD4 v2.0 @ F14
    Core i5 760 @ 20 x 201, 4.02GHz
    TRUE Black with a single Noctua NF-P12 pumping out 55 CFM @ 19db .
    2 x 2GB Mushkin Ridgeback (996902), @ 7-10-8-27, 2010-DDR, 1.66v
    2 x Gigabyte GTX 460 1024MB in SLI (Pre OC'd to 715MHz core and 1800MHz VRAM) @ 850 Core / 4100 Mem.
    Intel X25-M Boot Drive (OS and Programs) 200MB/s Read & 90MB/s Write
    Corsair X32 200MB/s Read & 100MB/s Write
    WD Caviar Blue 640GB C (Steam, Games, Storage, Temp Files & Folders, etc)
    Samsung F3 500GB Backup/Images
    Noctua 1300RPM 19dB case fan (rear extraction)
    3 x 140 MM Coolermaster LED fans (one front intake, one top extraction, one side intake)
    Dell Ultra Sharp 2209WAf E-IPS @ 1680x1050

  3. #3
    Chike is offline Senior Member
    Join Date
    Jun 2009
    Location
    Israel
    Posts
    6,531

    Default Re: Digital Thermal Sensors (DTS) - seeking themal peace of mind

    I didn't do that many tests. I think just calibrating under load would be enough unless you want to satisfy your curiosity, than compare to current readout.
    I've just idle calibrated the CPU temperature at idle, with stock vcore, internet disconnected, memory timings set to default JEDEC 400MHz.
    The case was open, and the mechanical thermometer was in front of the CPU HS intake.
    CPU temperature was 2c lower than calculated. DTS temperature was still ~10c greater than CPU at idle, after calibration.
    I then raised Vcore to 1.25 as my CPU would not exceed 55c at defalt bus speed and Vcore. DTS was ~6c above CPU temp @ ~65c so I ajusted it to be 5c using the idle calibration in RealTemp, since the TJmax 100 seems to be right and I didn't want to change it, and there is still a small slope error at these temps.

    Basically I didn't have to make the change, the DTS sensors are pretty close to the expected value at ~70c wothout calibration.
    The CPU temp readout at idle is the same(2c above ambiant) with my current settings.

  4. #4
    kick's Avatar
    kick is offline Senior Member
    Join Date
    Feb 2009
    Location
    Europe,the wild wild west.
    Posts
    1,223

    Default Re: Digital Thermal Sensors (DTS) - seeking themal peace of mind

    DTS

    try something like

    Slimline In Out Thermometer Digital Clock | Slimline Thermometer Digital Clock | Thermometer Digital Clock

    or similar for semi permanent installs .Guess what the supply voltage matches unless yer a trucker and theyre accurate enough.
    We can theorise a lot but it seems to me it comes down to the best bang for yer buck in the end.
    Current Systems:

    Asrock p67 Extreme6.............. Gigabyte EP-45 UD3 ...................... Gigabyte 73 PVM S2
    Intel i5 2500k 4.8ghz................ Intel Q8400 3.8ghz......................... Intel D820 2.8ghz
    Zalman 10x cooler.................... Coolermaster V8............................ HP cooler
    8GB Gskill ripjaw ddr3.............. 4GB Gskill PI ddr2.......................... 4GB samsung ddr2
    60GB ssd/500GB HDD .............. WD 1TB hdd.................................... Seagate 160GB hdd
    GTX 460 1GB x2 SLI ................. Msi 9600GT 512MB(died) ........... Onboard gx
    Win7 64 ,750w psu(ocz)............ Win7 64 ,520w psu,seasonic...... Win XP pro ,400w psu

    HEC 6A34 case . ....................... Jeantec R2 case............................ Packard Bell case

    hoping to upgrade to http://www.thinkgeek.com/stuff/41/habicase.shtml
    http://www.flixya.com/video/140325/Animal-launching

  5. #5
    Chike is offline Senior Member
    Join Date
    Jun 2009
    Location
    Israel
    Posts
    6,531

    Default Re: Digital Thermal Sensors (DTS) - seeking themal peace of mind

    Quote Originally Posted by kick View Post
    DTS

    try something like

    Slimline In Out Thermometer Digital Clock | Slimline Thermometer Digital Clock | Thermometer Digital Clock

    or similar for semi permanent installs .Guess what the supply voltage matches unless yer a trucker and theyre accurate enough.
    We can theorise a lot but it seems to me it comes down to the best bang for yer buck in the end.
    Nah, this isn't gonna be accurate, gampuo did it according to Intel's guidelines, they are accurate as can be, with my two left hands I'll never be able to reproduce his results.

  6. #6

    Default Re: Digital Thermal Sensors (DTS) - seeking themal peace of mind

    The first thing that should be noted when discussing the DTS based temperatures is fact the DTS on the CD2 were designed and intended only for the purpose of CPU thermal management only, i.e., the PROCHOT# and THERMTRIP# assertions (thermal throttling and shutdown triggering). The DTS were never designed to be used for determining absolute temperatures of the CPU cores, and problems are encountered when attempting to use them in such way.

    The core temperature reporting aspect of nearly all utilities out there are pretty much all the same in that they all use the following very simple formula: Absolute Core Temp = TJmax - DTS

    Clearly, the correct TJmax value is the key to the formula. But the TJmax is not a model/stepping specific number and there lies the first problem. According to Intel, the PROCHOT# activation point (TJmax) of each chip is calibrated individually. Unfortunately, determining the TJmax of a chip is not easily done and is well beyand the scope of most computer enthusists.

    The slope errors and sticking sensors are some of other problems encountered when using the DTS for temperature monitoring, but those problems are irrevent when the acccuate chip specfic TJmax is not known.

    IMHO, not knowing the accurate core temperatures is not really a problem. Besides, how would you interpret and use accurate core temperatures if you had them anyway?

    IMHO, relying on the DTS data directly instead of software guessed absolute temps is the best way to go. Anything above DTC=20 is absolutely safe.
    QX9650 batch L739A761/ GA-EP45-UD3P/ Kingston KHX9200 4x1G

  7. #7
    Chike is offline Senior Member
    Join Date
    Jun 2009
    Location
    Israel
    Posts
    6,531

    Default Re: Digital Thermal Sensors (DTS) - seeking themal peace of mind

    Quote Originally Posted by gampamu View Post
    IMHO, relying on the DTS data directly instead of software guessed absolute temps is the best way to go. Anything above DTC=20 is absolutely safe.
    This is a very good point, however I would like to go by something more reliable that just DTS=20, perheps Tcontrol with safety margins, unless Tcontrol is always higher than DTS=20. How do I find it?

  8. #8

    Default Re: Digital Thermal Sensors (DTS) - seeking themal peace of mind

    Pardon the late response. I was on the road for a week or so and my access to the internet was quite limited.

    The Tcontrol is a PECI (Platform Environment Control Interface) fan control parameter and it the minimum temperature at which the CPU fan speed is increased to the maximum full speed when PECI is enabled. The PECI lelels are pretty much the same as the DTS levels except the PECI levels are time averaged whereas the DTS levels are instantionous. That being the case the PECI levles are always going to be lower than the DTS when a system is under load. The Tcontrol is typically PECI=10, so if you want keep the CPU temperature to below the Tcontrol, limiting the DTC to 10 would be a good way to accomplish this.

    I prefer DTC<20 for 24x7 as it provides additional margin of safety.
    QX9650 batch L739A761/ GA-EP45-UD3P/ Kingston KHX9200 4x1G

  9. #9
    Chike is offline Senior Member
    Join Date
    Jun 2009
    Location
    Israel
    Posts
    6,531

    Default Re: Digital Thermal Sensors (DTS) - seeking themal peace of mind

    I thought they were greater than 20. I guess everyone should define safety for themselves. That article defines 25 as hot and 35 as safe.
    I think I'll go with 30 max for 24/7 use.

  10. #10

    Default Re: Digital Thermal Sensors (DTS) - seeking themal peace of mind

    CompuTronix's article is severly flawed in that it assumes the temperature derived from CPU diode ("CPU Temp") is the Tcase.

    Intel is quite specific regarding the measurement location of the Tcase: the temperature measured at the geometric center of the package on the surface of the IHS. Intel describes in detail how to install a thermocouple to obtain accurate Tcase on pp78-96 of this documentation. Clearly, it is unlikely Intel would suggest such a rigorous methodology for the Tcase measurement if the the temperature derived from the CPU diode is within a close proximity of the Tcase. The CPU diode temperature must not be anywhere close to the Tcase.

    According to my experiment and similar experiments conducted by others, at idle, the CPU diode temperature is around 7C lower than the measured Tcase and at full load the CPU diode temp can be up to 18C higher then the actual Tcase.

    CompuTronix's article is excellent otherwise. Just add 10 to 20 degrees to the recommended temps.
    QX9650 batch L739A761/ GA-EP45-UD3P/ Kingston KHX9200 4x1G

Thread Information

Users Browsing this Thread

There are currently 4 users browsing this thread. (0 members and 4 guests)

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •