3 Replies Latest reply on Nov 15, 2010 4:35 PM by Doc_SilverCreek

    Design recommendations av lifetime vs core temp


      Dear Everyone,


      I hope someone can help.


      I have some considerations when optimizing a system for minimum noise.

      I've tried to find official intel numbers of lifetime vs temp but failed.

      In particiular my question is how the saftey margin away from Tjunction MAX influences the expected lifetime of the cpu?


      Is there somewhere an expected lifetime (say @ 24/7/100%load) vs CORE temp?


      I only find notes tht below Tjunction MAX the cpu should function "properly", but it seems not so sensible to design with zero margin, and in particular how can we expect the product lifetime to vary as a function of the effective maximum core temp of the design?


      I expect that intel has some measurements of runtime-life time (probably theoretical) or extrapolated from tests, of lifetime VS operating temp?


      I seek to find an optimization of thermal desgin and cooling that minimizes noise but ALSO somehoe maximizes lifetime an MTBF.


      I'm looking for  the answer specifcally for two cpus.


      P8400 (C2D 2.26 Ghz)


      i7 870


      (two different systems)


      Ie. how close to MAX temp can one let these go (in order to minimize noise) AND still not see detrimental effects of lifetime (systems running 24/7 for years; worst case at 100%load)



        • 1. Re: Design recommendations av lifetime vs core temp

          If you are looking to mantain the best CPU temp with least noise , GO  H2O , Corsair has two nice units the , H-50 works good on 95 watt CPUs and the H-70 works good on 135Watt CPUs . I have one of each and they are great at maintaining temps no mater the load.  I have A Overclocked Xeon that runs24/7 and stays 40 c under worst load never gets above 50 c.

          • 2. Re: Design recommendations av lifetime vs core temp

            Thanks for your reply! I do understand that there are good and bad cooling, but my question was that I want numbers on life time vs temp, so I can make a rational tradeoff. The general relation that lower temps = better is clear, but how the functional relation Lifetime=f(T) looks like is not so clear.


            The reason is that in some embedded units, you simply don't have the choice to change the cooler. It is what it is, the option may be the reduce the FAN speed or so.


            So the question becomes that of a compromise. Noise vs temp.


            If I know how the statistical lifetime vs load temp looks like, I would be able to make that decision.


            For example for one P8400, we experimented with dropping the FAN speed to improve noice, but temp went from 65 to 75.


            I want to quantify, how much lower lifetime one can expect due to that, or due to getting the temp up even to 80-85.


            The common reasoning that people replace their PC every 3 years or so is not the answer I need, since not all poeople od thta. Some embedded sytems or special purpose PCs can easily run for 24/7 20 years, or simply until they break down.


            This is why I want to quantify the tradeoff. The only people that I can imagine that migh be able to answer this is intels own development/testing. So I was hoping to get some official responce from intel on this forum.



            • 3. Re: Design recommendations av lifetime vs core temp

              As simple as the question sounds it is very hard to give you a straight, definitive answer you want in an open forum.


              The Thermal/Mechanical Specifications and Design Guidelines (here is one) http://download.intel.com/design/processor/designex/322167.pdf covers a lot around what you are asking.


              Customer support can also provide MTBF data on the processor will also help you to make a judgement call. http://www.intel.com/support/processors/sb/CS-030116.htm


              Processor class also has a major impact here. A Xeon server processor is designed for extended operation at high work loads which generally translates to higher temps. I have seen a number of the newer Green Servers delay fan ramping to maintain CPU temps in the 65-75 deg range.


              So, I am not a lot of help either, but hopefully I have pointed you in the right direction.