3 Replies Latest reply on Sep 5, 2010 1:53 PM by one

    I7 9xx TJmax DTS and Tcase story

    one

      trying to find out why Smart Fan is operating while the cores are at 75C degrees when it is calibrated at the BIOS to engage when the CPU temperature is above 60.

       

      going through some papers, it seems there isn't any IHS probe seated in where the Tcase is actually located,

      at some, Tcase delta to TJmax is ~5 degrees while at others it's 15 or even 1.5,

       

      some refer to Tcase as actually located at the socket while it is known to be located at the top middle side of the IHS.

       

      so basically, is there any information on the Tcase delta to TJmax?

      measuring with an IR thermometer shows only 1-1.5 degrees change,

       

      2nd,

      why is it that the Smart Fan function at the BIOS operates at 60 degrees while the DTS is showing the cores to be at 75 degrees while TJmax is set to 100?

       

      does it mean that if a 920 CPU has Tcase of 67.9, the cores can actually operate at 69.4 (+1.5), 82.9 (+15) or is the TJmax actually should be set to 85?

      nowhere it seems to be fully documented or explicitely writen.

      regards .

        • 1. Re: I7 9xx TJmax DTS and Tcase story
          Terabytes

          AFAIK the DTS are not intended as an absolute temperature measurement but as a relative temperature from maximum in specification core temperature. However it seems most people would prefer to see an absolute figure so the relative reading is converted to an 'approximate' temperature which usually degrades the further you get from the relative DTS reading of 0.

           

          To try and simplify things regarding Tjunc to Tcase delta. Heat is generated from the cores and dissipated through the package and IHS. There exists a thermal resistance between the cores and IHS so the more power you are trying to dissipate the higher the delta will be between core and IHS.

           

          For example, if the core to IHS thermal resistance were 0.3°C/W then at 10W (low load) the delta would be 3°C while at 100W (high load) it would be 30°C.

           

          To try and properly measure the IHS with an IR thermometer would probably require removing the main cooling heatsink which would result in having to run at low powers and therefore the core/IHS delta would be low also.

           

          Your BIOS could be using a different measurement and/or calibration factor and so could read differently from your core reading. You could possibly try and ask technical support of your mainboard to clarify this for you.

           

          For the i7 920 and other Nehalem CPUs you can read the minimum temperature (temperature target) at which DTS = 0 from MSRs 1A2H bits 23:16. AFAIK this should be the minimum for Tjmax but Tjmax will likely be a bit higher than this. For instance if Tjmax has an accuracy of ±5% at DTS = 0 and MSR 1A2H shows a temperature target of 95°C then Tjmax could vary from 95°C to 105°C with an average of 100°C.

          1 of 1 people found this helpful
          • 2. Re: I7 9xx TJmax DTS and Tcase story
            one
            To try and properly measure the IHS with an IR thermometer would probably require removing the main cooling heatsink which would result in having to run at low powers and therefore the core/IHS delta would be low also.

            o.k, that's an idea, it's reasonable that the delta becomes less when the power dissipation is lower,

            it's really a shame no one has done any detailed research about that, even not the companies,

            it's all sums up to assumptions and approximate calculations,

            nothing too accurate..

            thanks for your sharing though :).

             

             

            you can have a look here for some further info:

            http://www.xtremesystems.org/forums/showthread.php?t=179044&page=169

            • 3. Re: I7 9xx TJmax DTS and Tcase story
              one

              actually CrystalCPUID MSR walker shows an output of 64 at address 1A2h bits 23:16.

              so if were talking about a deviation of 5 degrees from TJmax it means the throtteling  goes at 69 degrees?

              it isn't so..., even not relative to Tcase at 84 degrees.

              E: 64h is 100!

              so that fits right in!

              case closed !

               

              http://lenry.atw.hu/tjmax.pdf

              that's the PDF where it specifies the MSR reading address at page 24,

              as a side note, at page 30 there seems to be a graph showing the thermal compound efficiency at different ambient temperatures, as lower the Tambient goes better is C/W coefficient.