CPU temps. BIOS vs Softwere

Discussion in 'hardware' started by Oleg, Jul 17, 2016.

  1. Bill_Bright

    Bill_Bright Registered Member

    Joined:
    Jun 29, 2007
    Posts:
    3,536
    Location:
    Nebraska, USA
    And a fast talker!
     
  2. CrusherW9

    CrusherW9 Registered Member

    Joined:
    Dec 27, 2012
    Posts:
    517
    Location:
    United States
    They can do some cool stuff because they have the funding other channels don't. I also like Paul's Hardware and Awesomesauce Network. I still watch JayzTwoCents' videos but they seem to be less informational recently.
     
  3. Joxx

    Joxx Registered Member

    Joined:
    Sep 5, 2012
    Posts:
    1,510
    hehe
    Many out there snob those channels but personally I highly enjoy them, even when not informational they're recreational.
     
  4. Oleg

    Oleg Registered Member

    Joined:
    Mar 24, 2005
    Posts:
    442
    Location:
    USA
  5. CrusherW9

    CrusherW9 Registered Member

    Joined:
    Dec 27, 2012
    Posts:
    517
    Location:
    United States
    Nice link but I wish people explained how they determined what voltages and temperatures are 'safe'. It seems that all guides either pick an arbitrary number or just reference values they heard somewhere else. For example, the voltages referenced in that post, at least for Haswell (1.30vcore), are lower than the generally accepted "1.40vcore on air and low end liquid coolers" rule (again, for Haswell) that is widely prevalent on overclocking forums. And statements like this: "80C sustained Core temperature is too hot." bug me too. Some people say lower is too hot. Others (like me) say to stay under 95C during stress tests. Realistically, I guess there isn't a value that can be calculated for these parameters because any extra voltage, for example, technically reduces life span so I guess it's up to the individual to draw the line.

    Personally, I've run my 4770k at 1.4v on air trying to hit 4.7GHz. I've gotten it over 100C. I've run it for hours at a time (p95 with AVX 24 hour tests, some failed ~18 hours in) over 90C continuously (had to leave my window open in the winter.....). That was about 3 years ago when it first came out and it's been sitting at a constant 4.6GHz core / 4.4GHz uncore with 1.33 vcore max (has dynamic voltage so idles around .1v iirc) and a few other upped voltages. My ram is a 2x8Gb kit at 2400MHz C10, so my IMC is being stressed, too. I can't even remember the last time I blue screened. I guess we'll see how much life I get out of it but so far so good. If I get 5 years I'll be happy because it should be about time for an upgrade by then, and I'll finally be able to do a mini-itx build.
     
  6. Bill_Bright

    Bill_Bright Registered Member

    Joined:
    Jun 29, 2007
    Posts:
    3,536
    Location:
    Nebraska, USA
    The author does state he referenced CPU-Z and Intel ARK lists.

    But...
    I agree with concerns about users picking out arbitrary numbers, but frankly, stating 1.40V is "generally accepted" is just as arbitrary! And so is your own personal threshold of 95°C!

    That is NOT meant as criticism against you. It is meant to illustrate the REAL problem. And that is that there are no industry standards for measuring or reporting these values - in particular, temperatures. For example, Intel and AMD have their own ways of measuring and reporting acceptable temps. Some measure at the junction, some at the case. Some in all cores, some one core. Some average the cores.

    Sample rates for monitoring programs are not standardized. Most confusing (for me anyway) is there is no industry standard for sensor labeling either. AMD and Intel have their own ways and neither is consistent across their entire production lines. :(

    And there is no standard among motherboard makers for sensor type or placement either.

    This is why you can be monitoring temps in the BIOS, with Speccy, HWiNFO64, CoreTemp, SpeedFan, or any number of monitoring programs and they will all report something different!

    And of course, every CPU model is different - even among siblings in the same model line or even model, there can be significant differences.

    The only thing that is consistent is how those thermal diode sensors work. They detect a temperature value and represent that with some hexadecimal number. And that hex number is constant (for that temp and model sensor) and is published information for all manufactures to use when interpreting those values in their monitoring programs.

    But again, because there are no standards for sampling times or rates, and because temperatures in a processor can swing to either extreme and back again in just a couple clock cycles (and at 3+GHz, that's a lot of clock cycles in 1 second), it is impossible to get consistent readings across different programs.

    Lastly complicating the issue is these sensors used in computer systems are very low-tech and probably cost a penny each, if that. We are not talking about top precision, medical-grade quality here. Between identical sensors in the same production run, there could be several degrees differences in the reported temperatures.

    This total lack of industry standards forces us to use arbitrary numbers. I have mine too. For example, I don't like my CPUs to get over 60°C. Even though today's CPUs can operate safely at higher temps, if my CPU temps start sitting above 60°C for more than a couple seconds at a time, that's my personal threshold (an arbitrary number) that tells me I need to clean my filters, clean the interior, add another case fan, decrease the overclock, turn down the AC thermostat or look for some other cause.

    So the problem is, there are no standards for people to reference. So we have to go by arbitrary numbers.
     
  7. CrusherW9

    CrusherW9 Registered Member

    Joined:
    Dec 27, 2012
    Posts:
    517
    Location:
    United States
    That was the entire point of the post. The 1.4v limit get's spread around all over forums with nobody actually knowing why, and then a guy on a forum makes a nice post saying 1.3v. Someone should sacrifice two 6700K's and do a test. One at stock settings versus one with a massive overclock at 1.4v+ running hot (90C+) both running p95 24/7 until they die. See how much of a difference there is. Hopefully the excessively high voltage, clock, and temp would accelerate any degradation. Would obviously take years to know the outcome, but at least we'd have some info eventually.
     
  8. Bill_Bright

    Bill_Bright Registered Member

    Joined:
    Jun 29, 2007
    Posts:
    3,536
    Location:
    Nebraska, USA
    I think I would only trust the CPU makers or a professional lab. It could not be "someone" on a forum. There are enough folks on forums who consider themselves experts who've had absolutely no formal education in electronics or any formal training in computer repair and really have no business sticking their paws inside a computer.

    Proper testing and reviews takes some considerable expertise and resources - including proper testing facilities and $$$.

    Testing two 6700K in that manner would be great, but there are dozens of other processors that need testing too. And next year, dozens more. This is why it would be great if there were industry standards the processor makers followed, and published.
     
Loading...
  1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.