CPU temps. BIOS vs Softwere

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

  1. Oleg

    Oleg Registered Member

    Joined:
    Mar 24, 2005
    Posts:
    442
    Location:
    USA
    Have just learned BIOS temps. are not true or accurate. PC monitoring hardware apps giving you true readings. Back in the days it was otherwise. WTF?.
     
  2. Amanda

    Amanda Registered Member

    Joined:
    Aug 8, 2013
    Posts:
    2,115
    Location:
    Brasil
    Well, welcome to the new days, where motherboard sensors say your processor is at 60ºC when in fact the room temperature is 10ºC and your processor is actually a 9ºC ;) (consuming ~20-30W, 0% usage).
     
  3. CrusherW9

    CrusherW9 Registered Member

    Joined:
    Dec 27, 2012
    Posts:
    517
    Location:
    United States
    I'm assuming this is a joke since that's impossible but I didn't catch it.
     
  4. Joxx

    Joxx Registered Member

    Joined:
    Sep 5, 2012
    Posts:
    1,718
    Where did you learn that?
     
  5. Amanda

    Amanda Registered Member

    Joined:
    Aug 8, 2013
    Posts:
    2,115
    Location:
    Brasil
    Aparently my CPU is really cool when idle (20 to 30W consumption, 0% usage). Sometimes it stays at 5ºC. When I start browsing and doing other stuff, it goes to 10-15, sometimes 20. When compiling a Kernel, it goes to around 60 (never went pass that. On Windows with prime95, I couldn't get it to go beyond 45).
     
  6. Bill_Bright

    Bill_Bright Registered Member

    Joined:
    Jun 29, 2007
    Posts:
    4,041
    Location:
    Nebraska, USA
    This is true but they are as accurate as you are going to get at that point in time! The problem is, those temps displayed in the BIOS Setup Menu are not that helpful because running the BIOS Setup Menu is probably the least demanding task we can ask of our computers. We need to know the temps when our systems are working hard, not when barely coasting.

    All the software based monitoring programs must monitor the same hardware based sensors. The main problem is, these sensors cost less than a penny each and are not very high-tech. That said, they still are accurate to within ±5% (typically well within ±3%) which is pretty darn good. But they do need to be calibrated individually for the most accuracy and that's not happening except in critical medical/scientific devices.

    But they don't have to be that precise either. If you really NEED the 5°C the sensors may be off to keep your computer stable, then you have other issues, like inadequate case cooling.

    It should also be noted that there is nothing secret, magical or proprietary about these sensors. They simply convert the temperature they sense at that specific clock cycle into a hexadecimal number and provide that number to the chipset. Most importantly is that the way this is done is widely published. Whether the monitoring programs bother to look up and use those specs is another issue.

    What makes it bad for us is there are no industry standards for which sensor the hardware makers will use, where the sensors will be located, how it will be labeled, or how it will be monitored/sampled. Intel and AMD have their own sensors they use and places to put them. And they label them however they want. Same with NVIDIA and AMD for their GPUs. Same with motherboard/chipset makers for their "system" temperature sensor.

    So, for example, CoreTemp may sample the temperature of the "core" every 5 seconds while HWiNFO64 may sample the T-junction every 3 seconds while Speccy samples the CPU Tcase every 7 seconds and they all label them CPU Temp.

    And note in just 1 second, the CPU may have gone through 3 billion clock cycles!

    So because temperatures in processors can swing widely in just a couple clock cycles and because monitoring program timings are not standardized, different monitoring programs will report different temps at the same time.
     
  7. Bill_Bright

    Bill_Bright Registered Member

    Joined:
    Jun 29, 2007
    Posts:
    4,041
    Location:
    Nebraska, USA
    That is not likely. 5°C is just 41°F. It goes against the Laws of Physics to achieve temps cooler than the ambient (room) temps with conventional, passive (fans) cooling. Even with water cooling, unless the water is refrigerated, you cannot get cooler than ambient temps.

    So if live in the high mountains, or it is winter where you live and you have your windows open, it is not likely your room temps are that cold. It is more likely your sensor is bad or the monitoring program is not properly reading it.
     
  8. Amanda

    Amanda Registered Member

    Joined:
    Aug 8, 2013
    Posts:
    2,115
    Location:
    Brasil
    Room temps are really cool in the winter here. See, in Brazil people don't build proper insulated houses, not for the summer and neither for the winter (because it's expensive to do that here). So in the winter it could easily get -3ºC where I live (southest state in Brazil) and the room temperature doesn't go much beyond that. So if the temps are 7ºC outside, you can't expect them to be above 10ºC on the inside since we don't use radiators or anything to heat rooms :argh:
     
  9. Bill_Bright

    Bill_Bright Registered Member

    Joined:
    Jun 29, 2007
    Posts:
    4,041
    Location:
    Nebraska, USA
    If there is one place I want to be comfortable, it is in my own home. I would not be comfortable if my home was that cold inside. Good luck to you.
     
  10. Amanda

    Amanda Registered Member

    Joined:
    Aug 8, 2013
    Posts:
    2,115
    Location:
    Brasil
    That makes perfect sense, but trust me, if you were making 1200 bucks a month and electricity alone (without radiators) would rip off 250 of that, you'd think twice before making the house warm ;) Add radiators and it's easy to lose 300-350 bucks. Then there's rent, gas, food, clothing, internet, TV, etc, all expensive over here :D

    On the bright side, my temps are cooler than poeple who live near the equator in Brazil where it gets really hot in the summer :thumb:
     
  11. Bill_Bright

    Bill_Bright Registered Member

    Joined:
    Jun 29, 2007
    Posts:
    4,041
    Location:
    Nebraska, USA
    Hey, I fully understand. I was a single parent, working full time as a sergeant (meaning low pay) in the USAF, pulling a full load in school so I used to live on a very tight budget too. You do what you have to do.
     
  12. CrusherW9

    CrusherW9 Registered Member

    Joined:
    Dec 27, 2012
    Posts:
    517
    Location:
    United States
    This is what I was talking about earlier.
     
  13. Bill_Bright

    Bill_Bright Registered Member

    Joined:
    Jun 29, 2007
    Posts:
    4,041
    Location:
    Nebraska, USA
    Yeah, even with winter weather and no heat in the room, if the room is 10°C the CPU cannot be 9°C.
     
  14. Krusty

    Krusty Registered Member

    Joined:
    Feb 3, 2012
    Posts:
    10,209
    Location:
    Among the gum trees
    With the ambient temperature inside my house at 17°C and three tabs open in Cyberfox, Speccy says my laptop's CPU is at 48°C.
     
  15. Krysis

    Krysis Registered Member

    Joined:
    Dec 28, 2012
    Posts:
    371
    Location:
    DownUnder
    I've been using both for years – and their accuracy is ample as far as I'm concerned. I use a laptop and as most will know – air venting can be very poor. I start worrying when these programs report the CPU core temps exceeding 70C.

    As for Bios temperature reporting – never knew there was such a thing! o_O
     
  16. Bill_Bright

    Bill_Bright Registered Member

    Joined:
    Jun 29, 2007
    Posts:
    4,041
    Location:
    Nebraska, USA
    Understand the number of open tabs may impact your RAM usage but really does not tax the CPU. That said, 48°C is just fine. I don't start to worry until my temps start sitting above 60°C for more than a couple seconds. And when that happens, it typically is a reminder I need to clean my dust filters.

    Not all motherboard/chipsets provide temperature (or voltage) information in the BIOS but many, if not most do. It is often located under PC Health or something similar. Some also report fan speeds and allow you to set alarm thresholds too.

    FTR, Speccy is one of my favorite hardware information programs but it often has problems properly displaying sensor values (especially voltages). For example, right now, Speccy is reporting my voltages as:

    +3.3V = 2.028V
    +5V = 3.367V
    +12V = 0.048V​

    If any of my voltages were that far off (especially the +12V) this computer would not even boot. Note the ATX Form Factor allowed tolerances are ±5%.

    HWiNFO64 is reporting those same voltages as:

    +3.3V = 3.305V
    +5V = 5.010V
    +12V = 12.168V​
     
  17. Oleg

    Oleg Registered Member

    Joined:
    Mar 24, 2005
    Posts:
    442
    Location:
    USA
    Thanks everyone for the answers. As for monitoring software it's Speccy as well here.

    Now. I have Intel 4460 (Not overclocked). Room ambient temperature is 28.4C at the moment. CPU running at 32-34C Running stress test In-place with Prime 95 the temp. always stays at 60C, even if ambient temperature is lower. The question is are those OK temps.?
     
    Last edited: Jul 19, 2016
  18. Bill_Bright

    Bill_Bright Registered Member

    Joined:
    Jun 29, 2007
    Posts:
    4,041
    Location:
    Nebraska, USA
    Sure. Today's CPUs can actually run and remain stable at temps higher than 60°C, often higher than 70°C.

    Frankly, I am not a fan of using benchmark programs because the purposely over-stress your components. Can you take your family car to drag strip and race it down the track without blowing up the engine? Probably. But will that strain the engine, transmission and just about everything else? Yes. Will it increase friction and wear and tear? Yes. Will it increase aging? Yes.

    Instead of worrying what your temps are when pushed by synthetic scenarios, keep an eye on your temps when you do what you normally do with your computer.
     
  19. CrusherW9

    CrusherW9 Registered Member

    Joined:
    Dec 27, 2012
    Posts:
    517
    Location:
    United States
    Ehhhhh this is sort of proved false. Degradation due to heat and voltage (assuming they're within spec) don't gradually happen and will instead simply outright fail.
    https://youtu.be/44JqNJq-PC0
     
  20. Bill_Bright

    Bill_Bright Registered Member

    Joined:
    Jun 29, 2007
    Posts:
    4,041
    Location:
    Nebraska, USA
    Interesting video (as most of Linus' videos are). Thanks for that.

    But note I never said performance will degrade gradually over time. I said running benchmarks will "increase aging" and "wear" which Linus acknowledged starting about 6:25 into his video where he notes heat and overclocking make it worse.

    My point and the one Linus reinforces is not that performance degrades over time due to the abuse, or that benchmarking will speed up any degradation in performance. The point is such abuse can cause the card fail (totally stop working) sooner than it would if not subject to such abuse. That is (just pulling numbers out of the air), a frequently benchmarked card may fail after 5 years while a card not subject to such abuse may last for 7 or 8 years or longer.

    So Linus's video did not test for that, but implies it. To test that, you would have to have two cards (preferably 10 or more cards) and regularly run benchmark scenarios on 1/2 and regular gaming and computer task scenarios on the other 1/2 then note the life span of each. I am saying, and Linus supports this, that the benchmarked cards will have a shorter lifespan.

    So I'm sticking with my position and I am still not a fan. That is not to say I don't and will not use them. I do and will continue to run benchmarks on our new builds where I need to test and ensure the case cooling is sufficient. But I don't and will not run them to test the limits of CPUs, GPUs, hard drives, etc.
     
  21. Krusty

    Krusty Registered Member

    Joined:
    Feb 3, 2012
    Posts:
    10,209
    Location:
    Among the gum trees
    I took both of my laptops outside Monday my time and blew the dust out of them with my compressor. :)
     
  22. Oleg

    Oleg Registered Member

    Joined:
    Mar 24, 2005
    Posts:
    442
    Location:
    USA
    I agree with you on that %100. That's why i have run stress test only for half an hour. Some people do it for over 2 hours that's defiantly degrades CPU or age where it would quit working at all.
     
    Last edited: Jul 20, 2016
  23. CrusherW9

    CrusherW9 Registered Member

    Joined:
    Dec 27, 2012
    Posts:
    517
    Location:
    United States
    I've run Prime 95 (Which wasn't made for stress testing, btw) for over 24 hours straight multiple times a few years ago. Temps would hit 95C. Stable overclock hasn't changed.
     
  24. Bill_Bright

    Bill_Bright Registered Member

    Joined:
    Jun 29, 2007
    Posts:
    4,041
    Location:
    Nebraska, USA
    No but it is used as a benchmark program to "stress test" your hardware. In fact, as noted here,
     
  25. Joxx

    Joxx Registered Member

    Joined:
    Sep 5, 2012
    Posts:
    1,718
    Good video.
    Yeah, I'm a Linus fan myself, he manages to be both mainstream and original.
     
  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.