RealTemp plugin for RivaTuner

Discussion in 'Rivatuner Statistics Server (RTSS) Forum' started by burebista, Jan 19, 2009.

  1. Messerschmitt

    Messerschmitt Master Guru

    Messages:
    465
    Likes Received:
    0
    GPU:
    ASUS 7970 DCU II
    I have ran IntelBurnTest for 1-2 minutes and there is still a discripancy between RT monitoring and RealTemp. It was mostly accurate, as in RealTemp had the highest temp reported of 72 and RT 72 as well (with some small delays between them).
    There are however instances where Realtemp will be constant while RT plugin will jump a bit in temp (by 1 or 2 degrees), or vice versa.

    Anyway, they are very close together so I think the difference is insignificant. A bit strange tho that in your screenshots both RT plugin and RealTemp show the same values
     
  2. unclewebb

    unclewebb Guest

    Messages:
    71
    Likes Received:
    0
    GPU:
    VAPOR-X HD5770 1GB
    I prefer to use Prime95 Small FFTs for testing purposes because the CPU load and reported temperatures are very consistent.

    Core temperatures change instantaneously. I don't know enough about the inner workings of RivaTuner but with the plug-in, it might be reading the sensor immediately after having done something else that is causing a slight amount of extra heat to be reported when lightly loaded.

    I like using Process Explorer which shows that RealTemp is very efficient.

    When RivaTuner is drawing those good looking graphs, it's burning through more CPU cycles so it doesn't surprise me that it reports a degree or two hotter when the CPU is lightly loaded. At a full constant load running P95 Small FFTs, RealTemp and the RT plug-in should both be reporting the same, the majority of the time.

    My screen shot is of a Quad so maybe your Dual core is working a little harder.
     
  3. MrWoo

    MrWoo Active Member

    Messages:
    65
    Likes Received:
    0
    I have tried the RT plugin now, as well as the standalone RealTemp program. But I find discrepancy between readings. Let me explain.

    Start CoreTemp.
    Start Intel TAT.
    Start RealTempBeta.
    Start RivaTuner with c2d plugin & sf2rt plugin.
    Start PCWizard 2008.
    Start SpeedFan

    For instance now showing:
    Speedfan: 19 ; 20
    CoreTemp: 19 ; 20
    c2d plugin: 19 ; 21
    sf2rt plugin: 19 ; 20
    pcwizard: 19 ; 20

    but

    Intel TAT: 22 ; 22
    RealTempBeta: 24 ; 26 (RTPlugin most always shows same as Beta, so did not use in Riva as RTpluging and c2d plugin seem to conflict when ran together)

    So, there are no adjustments in any sensor program for offset.

    Which is to be trusted? This is on Intel 975xbx2 with e6700 c2duo cpu. Using Thermalright cooler, ambient room temp is approx 63F, cpu idling fan @ 30%. This is normal idle temp this time of year.

    Any thoughts?

    Thank you.
    MrWoo.
     
  4. burebista

    burebista Ancient Guru

    Messages:
    1,740
    Likes Received:
    36
    GPU:
    MSI GTX1060GAMING X
    At a first glance it's 5°C delta between RealTemp and others so I guess that it's another TJMax value set in RealTemp.
    Distance to TJMax from RealTemp match distance to TJMax in CoreTemp? I'm 100% for yes. :)
     

  5. MrWoo

    MrWoo Active Member

    Messages:
    65
    Likes Received:
    0
    Yes, an offset can be made. I was wondering if there was something I have not been aware of that would mean that RealTemp is somehow accurate and other sensors were not because of XXX reason.

    Yes, TJMax does match in RealTemp (distance + current) = coretemp max.

    So, one can trust sensors like cortemp or speedfan, where TAT, being intel made, is not correct without offset? I have read TAT was not accurate.

    MrWoo.
     
  6. burebista

    burebista Ancient Guru

    Messages:
    1,740
    Likes Received:
    36
    GPU:
    MSI GTX1060GAMING X
    TAT is outdated IMO. You can use RealTemp ore CoreTemp and my advise is not to rely on TJMax but on distance to TJMax in full-load. Keep it >20 and forget about temps. :)
     
  7. fgw

    fgw Member Guru

    Messages:
    115
    Likes Received:
    0
    GPU:
    MSI N680GTX Lightning
    every program that claims it is reading temperatures off the digital thermal sensor located within the cores is reading the very same value - assuming they are doing it right! there is just one place to read this dts values from, lets call this value dts.

    if you want to display an absoulte temperature, this reading needs to be converted using the formula: temp = TJmax - dts

    as you can see the only variable within this formula that could be changed is TJmax as dts is read directly from the sensor. thats the point where various programs differ. as intel so far did not release the exact TJmax values for their various processors AND to make this even worse this values might differ from part to part, every programmer needs to ASSUME the correct TJmax used in the calculation.
    so, for example, one program might use 95C while the other might use 90C and thus display temperatures which differ in exactly this 5C!
    some programs allow the user to change the TJmax value used in the calculation and thus adjust it to the processor used.

    as uncle already mentioned in one of his previous posts, this temps fluctuate rapidly and the variuos programs might read the sensor at slightly different times, thus reading slightly different values. this is also adding to the difference you see in various programs even if they are using the very same TJmax value!

    there is another feature incorporated into realtemp which differentiates it from all other programs: realtemp is able to compensate for nonlinearity in dts sensors! this feature is called calibration and can be adjusted seperately for each core. if this is changed from the default to anything else, realtemp is showing different temperatures as other programs. you need this set to 0 in order to compare realtemp readings to other programs.

    this all is valid if you are looking at one certain processor. use caution when you compare values from on processor to the other as the two processors most likely will have different TJmax values and also different nonlinearity in the sensors!

    additionally the sensors are calibrated to be most accurate at high temperatures (near TJmax) and might not even work at all at lower temps.
     
  8. Grendel_66

    Grendel_66 Guest

    Messages:
    791
    Likes Received:
    0
    GPU:
    MSI GTX 970 Gaming 4G
    The embedded sensors are not acurate at all, they where never ment to be used for temperature measurements, just to trigger the overheat actions. This has been discussed to death and then some over at xtremsystems and finally resulted in RealTemp allowing you to calibrate the sensors to a certain extend. Bottom line is that you can only trust that your CPU isn't burning up, anything else is just a ballpark.
     
    Last edited: Jan 29, 2009
  9. unclewebb

    unclewebb Guest

    Messages:
    71
    Likes Received:
    0
    GPU:
    VAPOR-X HD5770 1GB
    RealTemp is probably the only program that assumes TJMax = 90C for your E6700 Mr. Woo. That's why it reads 5C hotter than all of those other programs. rge over on XtremeSystems did some extensive testing and came up with these guidelines to go by:

    http://www.xtremesystems.org/forums/showpost.php?p=3384504&postcount=2429

    Based on your 17C (63F) room temperature and rge's findings, I think RealTemp is probably closest to the truth. I tested my E6400 B2 stepping with an IR thermometer and found that TJMax = 85C which all of the other programs assume, is usually too low for these processors.

    Intel TAT is a laptop testing tool and was never meant to be used for desktop processors.

    During the original Intel presentation last year they told us that their TJ Target for your processor was 70C. When some users complained, they bumped that up to 80C and said that TJMax might be a little higher than that number but never said how much higher. My testing and rge's testing leads me to believe that the truth is closer to 90C.

    The original 65nm sensors aren't too bad at all once you start using the correct TJMax. The Core i7 sensors are excellent and the 45nm Core 2 sensors seem to have a significant amount of variation in TJMax from one processor or core to another and also have some sticking issues at low temperatures.
     
  10. rek981

    rek981 New Member

    Messages:
    7
    Likes Received:
    0
    GPU:
    EVGA 9800GTX+
    I cannot get this plugin to display my temps in the hardware monitor. What am I going wrong?

    1. I downloaded plugin, and I unzipped the contents to the plugins/monitoring directory
    2. open rivatuner, hardware monitoring, setup, plugins, and I selected rtcore.dll

    Nothing happens. No temps, no options to display core temps.

    Please help

    Ross

    Q6600, 9800gtx+ running vista x64

    PS the actual realtemp program works flawlessly
     

  11. unclewebb

    unclewebb Guest

    Messages:
    71
    Likes Received:
    0
    GPU:
    VAPOR-X HD5770 1GB
    http://www.fileden.com/files/2008/3/3/1794507/RTCore.zip

    Here's the latest version of the RealTemp plug-in. If you check the file it should show version 3.06.

    I have the same setup as yours, Q6600, 9800GTX+ but I'm using Vista x86 and this plug-in works fine. Copy this new RTCore.dll to your

    C:\Program Files\RivaTuner\PlugIns\Monitoring directory and see if this makes any difference.

    [​IMG]

    Once you do the above, you should have some options available:

    [​IMG]

    If this doesn't work, let me know. Maybe my friend burebista is using Vista 64 bit so I'll ask him.
     
    Last edited: Feb 20, 2009
  12. burebista

    burebista Ancient Guru

    Messages:
    1,740
    Likes Received:
    36
    GPU:
    MSI GTX1060GAMING X
    Kevin, as you know for me everything is fine with your plugin and now with perfect CPU load I don't have any complains. :D
     
  13. chinobino

    chinobino Maha Guru

    Messages:
    1,139
    Likes Received:
    75
    GPU:
    MSI 3060Ti Gaming X
    Thankyou unclewebb & Unwinder!

    Nice work unclewebb, I now have all 4 CPU core temps and both GPU core temps on my Logitech G15 LCD display where I want them!

    I calibrated my CPU temps with Realtemp over a year ago to set the offset 2°C higher than what is being reported, as my CPU cores throttled at 93°C and the Tjmax for the QX9650 is 95°C.

    The only problem with setting the +2°C is that is appears in the LCD screen as 'XX.X °C (+2.0)', which is taking up too much LCD screen width.

    I have not bothered setting the offset because it is a minor issue but I am wondering if there is an option to not show the offset (+2.0) in the LCD/OSD?

    Regards, chino
     
  14. unclewebb

    unclewebb Guest

    Messages:
    71
    Likes Received:
    0
    GPU:
    VAPOR-X HD5770 1GB
    All Core processors typically start to throttle at a Distance to TJMax of 2 or 3. They are designed to do this to keep the core temperature from ever reaching TJMax even during some extreme conditions.

    I ran Prime95 Small FFTs on my E8400 for 3 hours with the fan disconnected. The processor went up to 98C and throttled off and on for 3 hours until I got bored and stopped it. Both cores hit 99C a few times but after 3 hours, only one core hit 100C and that only happened once.

    TJMax is not a fixed value but for a QX9650, I think it is probably closer to 100C. Intel calls these values TJ Targets and actual TJMax is typically higher. There is no single number that covers all QX9650 processors so you can use whatever value you think makes the most sense for your CPU.

    If you want to use TJMax = 97C then set that in the RealTemp 3.00 Settings window, click on OK and RealTemp will automatically use these values in the plug-in as well as long as you clicked on the RivaTuner button and told RealTemp where you have RivaTuner located. That way you won't have to look at (+2.0) in your OSD anymore.

    [​IMG]
     
    Last edited: Feb 21, 2009
  15. rek981

    rek981 New Member

    Messages:
    7
    Likes Received:
    0
    GPU:
    EVGA 9800GTX+
    Unclewebb,

    Thanks for you reply! I figured out my problem. I had to go to power user --> RivaTuner\Monitoring --> Set "EnablePlugins" to 1

    Ross

     
    Last edited: Feb 21, 2009

  16. Shinzok

    Shinzok Guest

    Messages:
    21
    Likes Received:
    0
    GPU:
    EVGA GTX 770 2GB
    Ehm.... So what you are saying .. is that I now also need to download realtemp ?????
     
  17. unclewebb

    unclewebb Guest

    Messages:
    71
    Likes Received:
    0
    GPU:
    VAPOR-X HD5770 1GB
    Using RealTemp to install the plug-in and set up your calibration settings is definitely the easiest thing to do. It's a free program and only one click away. Here's where the latest beta is always hiding:

    http://www.fileden.com/files/2008/3/3/1794507/RealTempBeta.zip

    If you have any problems with beta version 3.06 then use the last stable version 3.00. That version is in the download section here at Guru3D.
     
  18. burebista

    burebista Ancient Guru

    Messages:
    1,740
    Likes Received:
    36
    GPU:
    MSI GTX1060GAMING X
    New version, 3.0.7.0.
    Separate load for each core. I don't know more details (maybe uncle come with a thorough explanation :D) but his method seems more accurate than "classic" ones.
    Give it a try, compare it with load meters from CPU.dll, or Task Manager or Process Explorer or any load meter. :)
     
  19. Grendel_66

    Grendel_66 Guest

    Messages:
    791
    Likes Received:
    0
    GPU:
    MSI GTX 970 Gaming 4G
    Thanks for the heads up, much appreciated !
     
  20. stangowner

    stangowner Guest

    Messages:
    607
    Likes Received:
    11
    GPU:
    2xMSI N550GTX-Ti Cy II OC
    I'm running a Q9550 at home, and all 4 graphs in the hardware monitor are near identical to the task manager. But I have a E6300 at work, and the values reported by RealTemp in RivaTuner constantly seem low.

    This isn't the best pic, but it does show the difference. I cleared the history of both bars, and started them at the same time. If you look in the middle, you can see that Windows is reporting near 100% for both cores, but RealTemp barely breaks 50%.

    [​IMG]

    Any ideas? I've tried 3.0.7.0 and 3.2.0.0 beta. Same results. I could not find anything in the readme or cfg file. And searching returns millions of hits on temp, but not load.

    Thanks
     

Share This Page