1. Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

CPU Degradation

Discussion in 'Overclocking & Cooling' started by Outback Bronze, Jun 2, 2012.

  1. Outback Bronze

    Outback Bronze

    Joined:
    Aug 3, 2011
    Messages:
    489 (0.46/day)
    Thanks Received:
    103
    Hi guys. Im trying to read up on some articles on cpu degradation. Me, personally i dont think ive owned a cpu that has done it (yet). I thought id pick some TPU brains about this topic so what i am asking is:

    A) What exactly happens to the cpu.

    B) How many of u have experienced degradation.

    C) Is the cpu f...ed once it starts happening.

    Thanks.
  2. sneekypeet

    sneekypeet Unpaid Babysitter Staff Member

    Joined:
    Apr 12, 2006
    Messages:
    21,453 (7.12/day)
    Thanks Received:
    5,967
    I have to assume you mean while overclocking.

    A. Usually the Overclock may fail, and the CPU will require more volts to continue to run that speed, or may not be able to OC that far after a time.

    B.It happens, took me a while (year or two of beating the piss out of em) before I had to increase volts or lower the OC to get stable.

    C. I think its covered in A and B.
  3. LGV

    LGV New Member

    Joined:
    Jan 2, 2011
    Messages:
    260 (0.20/day)
    Thanks Received:
    64
    Got several socket A before.
    They cache die. like 512 kb ( barton) then half the cache is dead.
    also see one 2800+ not run on manufacture set up anymore, its like 100 mhz less after 6 year of non overclock use.
    Overclocking 939, 775 did make some of them die in terms of mhz . like when you run say 1.8 ghz and you can overclock to 2.8 ghz in first try, but second and so on just 2.6. ( opteron)

    you aware of the bin process of the CPU-s? I think thet just give you some answer.
    No such a thing perfect manufactured CPU.
  4. lilhasselhoffer

    lilhasselhoffer

    Joined:
    Apr 2, 2011
    Messages:
    1,545 (1.29/day)
    Thanks Received:
    920
    Location:
    East Coast, USA
    CPU degredation is generally a manufacturing issue, which becomes more substantial over time.

    Please excuse the simplistic explanation, but materials science can bore people to death.


    Instead of electricity, imagine that the processor works with water flows. Each valve (transistor) is ideal, so will functionally never break down. The valves are connected by plastic tubing. This tubing will functionally last forever under rated flows (voltage).

    Overclocking increases the flow through the tubing, putting more pressure on it. Now, consider that the tubes are all made by the same manufacturer, but on different production lines. The end result of this means some tubes are thicker, and thus can stand greater flows; conversely, there are tubes with slight nicks, that will burst when too much flow is experienced.

    Depending upon where the tube bursts, different areas of the system will be influenced. If only a redundant part of the system is burst, the system can cut flow off to that area. If a critical area of the system is damaged then the system will not work.

    A CPU can experience non-critical failures which disable either components, or the whole chip. AMD used to do this during binning, so that x2, x3, and x4 chips were all segregated from the same base chip.


    In essence, degredation is slowly damaging the CPU. It can be to either critical components, or non-critical parts. If you are lucky, and unstable CPU can be stabilized by disabling a core. If you're unlucky the cache or IO is damaged. Welcome to the fun of manufacturing in the real world.
    Animalpak and Outback Bronze say thanks.
  5. St.Alia-Of-The-Knife New Member

    Joined:
    Mar 9, 2011
    Messages:
    195 (0.16/day)
    Thanks Received:
    32
    Location:
    Montreal, Canada
    google "electromigration"
    its basically what happens when your cpu gets too much voltage for a long time and parts of the circuit gets damaged
  6. cadaveca

    cadaveca My name is Dave

    Joined:
    Apr 10, 2006
    Messages:
    13,743 (4.56/day)
    Thanks Received:
    6,767
    Location:
    Edmonton, Alberta
    What happens in called "Electromigration". Basically, electrons tunnel through parts they shouldn't, and this creates greater leakage. CPUs will require more and more current to operate as electromigration happens, which also makes electromigration more likely, and it builds into a cascade until the tunneled hole prevents operation completely.

    Everyone has, although most won't notice it. My 3960X has degraded quite significantly.

    Yes, and no. Basically, the heat given off by the CPU, is leaked current, and by design. There are "provided pathways" for the electrons to exit, and as they do, this excites the atoms of the CPU, makes them move a bit, and generates heat.

    But, CPUs are made of silicon, a semi-conductor, whose electrical properties change based on the temperature it is at(it wil conduct better when colder, and leak less). We "overclockers" take advantage of this fact, and add more cooling, making the CPU a better conductor, and this allows it to operate at a higher frequency. The ability of a cooling device to remove heat directly affects stability because of this.

    So, if you cool a "broken" CPU enough, it will work just fine.

    When you cool a CPU, you lower it's ability to leak excess current. This can be offset by increasing frequency, and it can be a delicate balance to find teh best point for a CPU to run at. CPUs that are more leaky at room temps, overall, tend to overclock further when cooled, becuase there is still a greater ability for excess current to be shed when it is cooled better, and it's ability to leak excess is lowered. So overclockers that go sub-zero tend to look for high-leakage chips, even though technically they are using a less efficient peice of silicon under normal temperatures. So, a CPU that is poopy at room temp, might be great under cold, and vice-versa.
  7. phanbuey

    phanbuey

    Joined:
    Nov 13, 2007
    Messages:
    5,199 (2.14/day)
    Thanks Received:
    973
    Location:
    Miami
    Great bit of info :toast:
  8. Neoroy New Member

    Joined:
    May 30, 2012
    Messages:
    15 (0.02/day)
    Thanks Received:
    0
    Location:
    Bandung, Indonesia
    Nice explanations Cadaveca ;)
  9. robal

    robal

    Joined:
    Aug 5, 2008
    Messages:
    485 (0.22/day)
    Thanks Received:
    110
    Location:
    Hampshire, UK
    I've stumbled into something really strange (example of CPU degradation).
    It was the weirdest PC problem in all my PC-problem-fixing career.

    The system ran Athlon X2 5000+ (unlocked "BE")
    The CPU kept overheating.
    I reapplied TIM and reseated heatsink couple of times. Still the same.
    I thought that the heatsink is crappy, so I replaced it with high end one. Still the same.
    Even tried other PSU and MB. Same.

    It's not the case of damaged heatspreader. The CPU simply gives off that much heat.
    It must have at least tripled its TDP.

    I'd associate it with a history of overclocking and increased leakage through degradation.
    What astounds me is that it's still a stable CPU (prime95, memtest86), though you need hell of a cooler to keep it below 70 C.

    Long post... sorry
    cadaveca says thanks.
  10. cadaveca

    cadaveca My name is Dave

    Joined:
    Apr 10, 2006
    Messages:
    13,743 (4.56/day)
    Thanks Received:
    6,767
    Location:
    Edmonton, Alberta
    Perfect example, robal. ;)
  11. Outback Bronze

    Outback Bronze

    Joined:
    Aug 3, 2011
    Messages:
    489 (0.46/day)
    Thanks Received:
    103
    So these cpus that need more volts to be applied for a specific speed, how many more volts we talking?

    e.g. Cadavecas 3960x @ 4.5ghz with 1.45v

    Lets assume that the cpu is suffering from degredation.

    Will it need 1.5v to operate @ 4.5ghz or only slightly say 1.475v? Basically will it need a heap more juice or only a little?

    Also can the cpu ever be revived from degradation? Thanks.
  12. nleksan

    Joined:
    Mar 26, 2012
    Messages:
    239 (0.29/day)
    Thanks Received:
    36
    Location:
    Ohio
    Is this where the concept of "every 10*C you lower the operating temperature, you double the lifespan of the component" comes from?

    If so, I am definitely going to cool the bejeezus out of my 3930K!
  13. cadaveca

    cadaveca My name is Dave

    Joined:
    Apr 10, 2006
    Messages:
    13,743 (4.56/day)
    Thanks Received:
    6,767
    Location:
    Edmonton, Alberta
    Yes, the needed voltage, which inherently provides more power(Volts x Amps = Watts), can increase, but the value that is needed can depend on a lot of things. It's also possible that the CPU can get jsut one tunnel, and be dead.

    I could give a perfect example. AMD, with the HD4890, doubled up silicon vias(the path between the chip itself and the PCB, through the silicon), as they foudn that there was electron migration occuring at these points due to a faulty solder mask exposure. The could nto fix the exposure, so they doubled the vias up, as if the only one got too bad of a tunnel(and this was apparantly quite common), the whole chip itself was dead, because the current was lost before the chip was even able to run.

    And no, there is no "fixing" degredation that has occurred via electromigration. However, with these IvyBridge chips, I am sure many users are going to lose clocks, and gain heat...not from degredation, but by drying out the thermal paste.


    My 3960X did not degrade under natural usage models, so there is no need to worry about other SKT 2011 chips. I'm actually running a higher CPU voltage with my 3830 than I did with my 3960X. Let's just say though, that I know how to kill a SKT2011 chip. :laugh:

    Yes, exactly. Specifically, that it's a semi-conductor, is why that is tossed around, although, I do not beieve that that temperature scale is 100% correct for current silicon.
    nleksan says thanks.
  14. LGV

    LGV New Member

    Joined:
    Jan 2, 2011
    Messages:
    260 (0.20/day)
    Thanks Received:
    64
    NO

    increase voltage wont help always.
    defiantly not on cache damage.
    actually the increase of voltage may increase CPU leak. so get worst.
    every cpu degradation different.
  15. NC37

    NC37

    Joined:
    Oct 30, 2008
    Messages:
    1,165 (0.56/day)
    Thanks Received:
    257
    Such a nostalgic thread. Reminds me of when I'd clock Apple CPUs. G3s, G4s/etc...huge argument back then about electron etching. Most clockers then would swear by it, citing their results over time. But then others would deny electrons were doing anything to the CPUs.

    Thank you for this thread. Gave me a good read :).

Currently Active Users Viewing This Thread: 1 (0 members and 1 guest)

Share This Page