How room temp. affects processor temp. (graphs)

Discussion in 'Overclocking & Cooling' started by graysky, May 13, 2007.

  1. graysky

    graysky Gawd

    Messages:
    620
    Joined:
    May 6, 2007
    I was curious just how much room temperature can affect the temperature of the PC. I chose to look at the CPU since Speedfan can monitor and log temps for each of the 4 cores in my Q6600, and since of all the components, the CPU is probably the most sensitive to changes in air temp. since mine is air cooled.

    Anyway, I did the same x264.exe encode that I have been doing for all my temperature comparisons and monitored room temp. with a calibrated digital thermometer at several points during each encode. These were averaged and graphed against the averaged reported core temp* values from Speedfan for the entire second pass of a 2-pass x264 encode of the same video file. I was happy to see that for the different room temps used, the increases were pretty linear (certainly within error).

    Result:
    You can see by the slopes of the regression line that every delta °F of room temp. affected the average core temp by about 0.8 °C and for your Celsius folks, every delta °C of room temp. affected the average core temp by about 1-1/2 °C.

    So what does this mean and why do you care? Well, using these rules of thumb, if it's currently 70 °F in your room, and your average load core temp is 65 °C, you can expect that to change by roughly 0.8 °C for every single °F your room temp. change. Say your room hits 80 °F. Your load core temp should increase from 65 to 73 °C which may be unacceptable to you and you might want to adjust your o/c accordingly. This is just an approximation based on my system. Your mileage may vary...

    *The numbers I used are equivalent to those collected by TAT or RMClock: these temps are core temps. As I understand it, TJunction never changes and is a fixed value for a given chip. The Quads get a values of 100 °C and the duals get 85 °C. The core temp is defined as:
    Code:
    Core temp = TJunction - DTS
    
    Example, DTS reads 62.  You take 100-62=38 and your core temp is 38 °C.
    DTS (Digital Thermal Sensor) can be read directly for each core. See this thread for more on reading your DTS directly without software like TAT or RMClock paying attention to uncleweb's posts using crystalCPUID to read the DTS directly. When I tried this method, I was able to get the same values for the core temps on my Q6600 as TAT and RMClock gave me. For some reason, Speedfan always shows cooler core temps for my chip which I corrected by adding 15 °C to each temp (the table is CORRECTED temps). Read more about that in the caption under the graph.

    Raw data and graphs
    Hardware specs: Q6600 (lapped) @ 9x333, Ultra-120 Extreme (lapped), P5B-Del., P182 case w/ 4 fans on low, Corsair 620HX, Ballistix DDR2-800 @ 4-4-4-12 (1:1 Mem:CPU).

    [​IMG]

    Raw data table in case you want the individual points:

    [​IMG]

    Downloads and References
    To download crystalCPUID: http://crystalmark.info/software/CrystalCPUID/index-e.html
    To download rmclock: http://cpu.rightmark.org/products/rmclock.shtml
    To download speedfan: http://www.almico.com/speedfan.php
    To download TAT: http://www.techpowerup.com/downloads/392/mirrors.php

    To read more about TJunction and for a discussion about other things relating to thermal output fromC2D chips, see: http://forumz.tomshardware.com/hardware/modules.php?name=Forums&file=viewtopic&t=221745
     
  2. ghost6303

    ghost6303 2[H]4U

    Messages:
    2,291
    Joined:
    Jul 24, 2004
    very interesting. how did you change room temps between tests?
     
  3. graysky

    graysky Gawd

    Messages:
    620
    Joined:
    May 6, 2007
    I'm glad that someone else is finding it useful -- always glad to contribute.

    It did it on its own... it's been dropping down below 60 °F these past few nights and I've left the windows open. The room slowly warms up over the mornings... I got the 3 low temps this morning and the higher temp last week.

    To do this experiment under truly controlled conditions would require some sort of large enclosure with a ton of air at a given temp. These experiments are sort of a rough estimate since there can be fluctuations in room temp at different points in the room. Close enough for government work though :p
     
  4. ghost6303

    ghost6303 2[H]4U

    Messages:
    2,291
    Joined:
    Jul 24, 2004
    The US Government has been holding strict tollerances of + or - 99% for years. (when they even read the data before making decisions)
     
  5. Arcygenical

    Arcygenical Will Watercool for Crack

    Messages:
    24,689
    Joined:
    Jun 10, 2005
    :D


    OP: Don't I know what you mean!!! On thursday it went to 34c (PS: wtf) in my room, and my computers were loading in the mid 70s :eek:

    As air gets heatsoaked, it's ability to conduct heat is reduced (due to the energy level of the molecules, and the general expansion of the gas) so it's a lose-lose situation. A 6c increase in room temperature hit you with a 10c jump in temperatures. Following this trend, I bet a 11c jump in temps (what I had) could almost double that temperature delta!
     
  6. graysky

    graysky Gawd

    Messages:
    620
    Joined:
    May 6, 2007
    I'll update the thread in the coming weeks as it gets hotter.
     
  7. graysky

    graysky Gawd

    Messages:
    620
    Joined:
    May 6, 2007
    Updated 15-May with two higher temp points. It was like 85 °F outside today and I left the A/C off. Room got hot so I ran two additional points at about 76 and 77. Anyway, as you can see, it's all linear up to 77.5 °F and I suspect it'll continue that way if it gets hotter.
     
  8. graysky

    graysky Gawd

    Messages:
    620
    Joined:
    May 6, 2007
    Updated 19-May: Fixed a small error and put formulas for both °F and °C (of your room temp) on the graph. The relationship is: for every °F change in room temp, the load TJunction changes by 1-1/2 °C. Well, this is true for my system, YMMV...
     
  9. TigerLord

    TigerLord [H]ard|Gawd

    Messages:
    1,085
    Joined:
    Mar 11, 2007
    I would have been very suprised if that graph would have turned to be an exponential function :p