Fact or Fiction? CPU and Mobos slow down over time.

Discussion in 'Intel x86 CPUs and chipsets' started by KonMan, May 2, 2019.

  1. PersianImmortal

    PersianImmortal Member

    Joined:
    Jun 27, 2001
    Messages:
    9,054
    Location:
    Canberra
    Good point, that's a factor.

    It made me think of another scenario where performance can substantially degrade over time for no apparent reason, and that's if you fill an SSD to close to capacity. That will definitely noticeably reduce its performance, and this problem may be more common given the number of people who buy smaller SSDs as a boot drive and don't notice or care if it gets filled up. But that's really a software/maintenance issue rather than a degradation of hardware.
     
  2. evilasdeath

    evilasdeath Member

    Joined:
    Jul 24, 2004
    Messages:
    4,855
    Yes i agree with most of the thread, they don't slow down, but they may error more, depending on programming errors will either result in a crash, or potentially result in an error check picking it up and running again. Depends on the software. So it may take longer to get to the same result. If the software is like that. Most will crash.
     
  3. qwertylesh

    qwertylesh Member

    Joined:
    Aug 21, 2007
    Messages:
    8,576
    Electronmigration

    (Spelling corrected)
     
    Last edited: Jul 1, 2019
    Bold Eagle likes this.
  4. bob05

    bob05 Member

    Joined:
    Sep 20, 2009
    Messages:
    245
    Location:
    NSW
    Intel CPUs have slowed significantly over the past years with all the security flaws being found. If it's an AMD CPU, you are safe.



    SSDs are hard to slow down, so you wouldn't notice it.
     
    terroristone, l_ QuadX_l and nope like this.
  5. d_prac

    d_prac Member

    Joined:
    Apr 12, 2015
    Messages:
    79
    I'll argue the counter-consensus here.

    I had an old budget gaming PC that I built in highschool from when Foxconn first tried to sell motherboards directly under their own brand name.
    This thing was truly a buggy piece of shit and the only motherboard I've had fail in-service. Over several years it had three stages of failure, the last was a complete failure to POST but the two before it resulted in system performance reductions and loss of mobo connectivity/functionality. I wasn't savvy enough at the time to figure out exactly what had happened but I knew burning when I smelled it. I'm fairly sure something to do with power delivery to the CPU had partially fried each time.

    I've also had a GTX 670 that randomly fried some capacitors that dropped about 20% performance when it happened but still ran stable. Went back up to 100% when it came back from RMA.

    Although this isn't "slowing down" per se, it may as well be to the layman. Perception is reality when you don't understand what's happening out of sight.
     
  6. Bold Eagle

    Bold Eagle Member

    Joined:
    Jun 28, 2008
    Messages:
    8,992
    Location:
    Brisbane
    Exactly; 'Electromigration'
    https://en.wikipedia.org/wiki/Electromigration
    Depending on the usage and composition of that system will dictate how great an influence this variable will have on a system over time. For the vast majority of users it would not be measurable a decade later but in a system like an MRI or as others were noting earlier a LN2 system there would be changes.

    I think your initial question needs context to make it more meaningful.

    Would a domestic PC by a typical home user ever note this? No, not in anyway they could actually ever measure as it would be in nanoseconds at best after a decade 'after normal everyday use' with all other factors remaining equal/constant.

    Could someone who OC'd their PC, pumped extra/excessive voltages into the CPU and increased the frequency. The likelihood will scale relative to the voltages put in (energy) and the cooling applied - thermodynamics.

    Analogy: I have a Daihatsu Cuore (3cyl 1.0L) which I have had for the last 10 years. I have maintained oil, filters, spark plugs, etc. and it still goes as fast today as it did 10years ago. I have not done any major services and not changed rings, timing belts, bored it out etc. If I took that same car and then bored it out, ported it, over sized piston and rings, put a turbo on it and a nitro kit and then went and took it drifting every w/e really giving it stick (let's call this a highly OC'd PC) 10 years later unless I keep constantly rebuilding that engine it will be going a lot slower and have significant wear.

    If you have a PC that is utilised under such extreme conditions it will wear over time and show 'impairment'.
     
    Last edited: Jul 1, 2019
    qwertylesh likes this.
  7. qwertylesh

    qwertylesh Member

    Joined:
    Aug 21, 2007
    Messages:
    8,576
    Thank you.

    Only a processor that can defy the laws of physics/thermodynamics and not electrically degrade over time could withstand the performance degradation caused by electronmigration.

    The more electron loss = the more internal noise and the higher error rate in calculation on the silicon level

    Without use-case context, the answer to ops question is still a Yes CPUs slow overtime, granted scale of time and factors such as ocing/mining/burning in or decades/centuries of within spec use are all important considerations.

    Even with 100+ years of within spec use and the answer still isn't a definite No, only that it's highly likely other parts to drive the system would be bad long before the silicone (ie electrolytic smds) and as my bold eagled friend states the measure of slowdown could be tiny in this case. The measure could be benched but not affect real world usage

    A good burnin example (good example, bad practice) is when Linpack is ran improperly ( as in lots more then a once off standard 3 pass test ) will add the equivalent of upto centuries worth of dice ageing to a processor - even with adequate cooling

    I can't stress enough the 'use with caution' when testing CPU stability with any burn tool that leverages linpack (ie Intel burn test)

    People often run it (you know who you are and shame on you if you flipped it afterwards) in the same manner as prime95 for huge periods of time on end (ironically to test OC's) , where actually one 3 pass run is all that's needed to determine stability (the residual norm table should have the exact same character output per line/column/row) - any difference equals unstable

    Furmark presents the same misuse risks only for GPU

    Another example is CPUs that have had their TJmax thermal threshold exceeded - when these chips don't immediately die they're still ruined because the thermal stresses affect the structure of the dice. (it even happens faster than Mean Time Between Failure specifications, if your temp is within 20 percent of the TJmax value - ie TJmax=100c & dice temp=80c<)
     
    Last edited: Jul 1, 2019
  8. Hater

    Hater Member

    Joined:
    Nov 19, 2012
    Messages:
    3,632
    Location:
    Canberra
    does software bloat cause the perception change?
     
    Hotrod2go likes this.
  9. chook

    chook Member

    Joined:
    Apr 9, 2002
    Messages:
    1,196
    I think that one of the key points in the OP is 'non-overlocked' as extreme overclocking can cause degradation of the CPU and/or other components over time.
     
  10. cvidler

    cvidler Member

    Joined:
    Jun 29, 2001
    Messages:
    12,213
    Location:
    Canberra
    software bloat is a big problem*, as there's more hardware horsepower programmers write less efficient code. Efficient code is harder (thus more expensive) to write, and not needed when there's more horsepower to mask the performance loss.

    But OP says a bit perfect copy of the original system on day 1 is re-used for the long term test, so software bloat is not in play for this scenario.


    *as someone who can remember windows coming on a few floppies, and not multiple gigabytes of downloads.
     
    Hotrod2go likes this.
  11. 2_stroke

    2_stroke Member

    Joined:
    Aug 9, 2008
    Messages:
    1,277
    Location:
    cranbourne 3977
    Shrooms help alot XD or was it just some 420 debate. What if it fell into a black hole would it get faster with time? or is this a conspiracy to make more threads in the intel sub section to beat the AMD sub section Due to Ryzen 3000? Is the OP working for intel?
     
    Last edited: Jul 2, 2019
  12. bob05

    bob05 Member

    Joined:
    Sep 20, 2009
    Messages:
    245
    Location:
    NSW
  13. power

    power Member

    Joined:
    Apr 20, 2002
    Messages:
    58,971
    Location:
    brisbane
    cvidler likes this.
  14. qwertylesh

    qwertylesh Member

    Joined:
    Aug 21, 2007
    Messages:
    8,576
    Would the rate of cpu slow down be negated by my onset of old age brain slowdown, i mean like, perceptually the machine would get faster yeah? :lol:
     
  15. Bold Eagle

    Bold Eagle Member

    Joined:
    Jun 28, 2008
    Messages:
    8,992
    Location:
    Brisbane
    Now your talking about perception.......that is becoming subjective...........
     
    qwertylesh likes this.
  16. dr_deathy

    dr_deathy Member

    Joined:
    Apr 27, 2007
    Messages:
    2,606

    yeah nah, I have run intel burn in test for literately hundreds of runs on a POS with stock intel cooling in a row as well. How many till she dies? these are cpus with already 5-10 years of service as well.

    Yes i have tried to kill cpus. Your shitty motherboard and all the second hand ones will give out long before 99% of CPUs.
     
  17. Ratzz

    Ratzz Member

    Joined:
    Mar 13, 2013
    Messages:
    8,601
    Location:
    Cheltenham East 3192
    OT but should you ever decide to go drifting in said Daihatsu, please let me know so I can subscribe to your YT video of the beast in action !!!
     
  18. callan

    callan Member

    Joined:
    Aug 16, 2001
    Messages:
    4,787
    Location:
    melbourne
    FWIW I was in the computer industry for over 20 years - both retail and business (server).
    In all that time, and the many thousands of machines I worked over - ONE (1) Intel CPU failed in service. Motherboards??? many, dozens I'd reckon.

    Callan
     
  19. dr_deathy

    dr_deathy Member

    Joined:
    Apr 27, 2007
    Messages:
    2,606
    And a lot of these machines have been in harsh environments.
     
  20. Sunder

    Sunder Member

    Joined:
    Apr 26, 2012
    Messages:
    3,596
    Damn, you didn't get extreme enough. I did watercooling and back phase change cooling between the PII and Q6600 days. Replaced a LOT of CPUs on the same motherboard. I'm not sure exactly when, but I think it was when Intel started multiplier locking CPUs and putting self protection on them, that it became almost impossible to damage a CPU through overclocking.

    That said, I recognise the context was about stock and looked after PCs.
     

Share This Page

Advertisement: