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

CPU Electron Migration and You!

Discussion in 'Overclocking & Cooling' started by TheMailMan78, Nov 30, 2009.

  1. TheMailMan78

    TheMailMan78 Big Member

    Joined:
    Jun 3, 2007
    Messages:
    21,185 (7.75/day)
    Thanks Received:
    7,708
    This thread is for the discussion of CPU voltage tolerances, temperatures and anything else that effects the performance and life span of a CPU. I want this thread to be a learning tool as to the facts and experiences of many for the members of TPU.

    I'll start.

    Its been stated in other threads that the Phenom II can sustain voltages far beyond the rated max of 1.5v as long as the temperature stays acceptable. I do not subscribe to this theory as chip degradation starts the second you add power. The more power the more degradation no matter what the temperature is.
     
  2. erocker

    erocker Super Moderator Staff Member

    Joined:
    Jul 19, 2006
    Messages:
    39,896 (13.07/day)
    Thanks Received:
    14,304
    Sure, they can go up to 1.55v. If temps are good, temps are good. AMD still uses SOI, the same as of my AMD 3200+ s754. That chip uses 1.65v and has done so (while being at 54c all the time) for 5 years now. I subscribe to this fact and have experienced it firsthand.
     
  3. TheMailMan78

    TheMailMan78 Big Member

    Joined:
    Jun 3, 2007
    Messages:
    21,185 (7.75/day)
    Thanks Received:
    7,708
    Well thats far more than what AMD is stated as being safe. Our number one enemy is temperature. Keeping that under control is relatively easy. However degradation occurs the second you add power. With that being said you have been lucky for 5 years. ;)
     
  4. erocker

    erocker Super Moderator Staff Member

    Joined:
    Jul 19, 2006
    Messages:
    39,896 (13.07/day)
    Thanks Received:
    14,304
    Nope, 1.65v is stock voltage for that chip. ;) AMD, says 1.55v is the max voltage for their 45nm chips, hence why they allow 1.55v in Overdrive. Also, according to AMD, any overclocking voids the warranty anyways.
     
  5. TheMailMan78

    TheMailMan78 Big Member

    Joined:
    Jun 3, 2007
    Messages:
    21,185 (7.75/day)
    Thanks Received:
    7,708
    Sorry I was referring to the Phenom II.
     
  6. Binge

    Binge Overclocking Surrealism

    Joined:
    Sep 15, 2008
    Messages:
    6,982 (3.08/day)
    Thanks Received:
    1,752
    Location:
    PA, USA
    It depends entirely on the process used to make the chip. Sadly each chip will present different tolerances to electricity.
     
  7. TheMailMan78

    TheMailMan78 Big Member

    Joined:
    Jun 3, 2007
    Messages:
    21,185 (7.75/day)
    Thanks Received:
    7,708
    So going above AMD said tolerances is a roll of the dice in other words.
     
  8. erocker

    erocker Super Moderator Staff Member

    Joined:
    Jul 19, 2006
    Messages:
    39,896 (13.07/day)
    Thanks Received:
    14,304
    It's always a roll of the dice. I've recieved chips that are entirely unstable at stock. A 720BE and a e7200.
     
  9. Binge

    Binge Overclocking Surrealism

    Joined:
    Sep 15, 2008
    Messages:
    6,982 (3.08/day)
    Thanks Received:
    1,752
    Location:
    PA, USA
    :toast: I'd say that's a safe assumption. Most companies make a certain degree of margin for extra stress. The margin is used to maintain minimal deviation in performance while using the base-line product for it's intended use. The moment you're overclocking you are testing the hardware margin of your specific chip.
     
  10. Error 404

    Error 404

    Joined:
    Apr 14, 2008
    Messages:
    1,777 (0.73/day)
    Thanks Received:
    169
    Location:
    South Australia
    At the moment I've got my CPU running at 3.5 GHz, 1.45 volts vcore, 1.4 volts FSB. It really isn't a good clocking chip, the FSB wall is ~325 MHz... :ohwell:
    Are these safe voltages? I've had it running for about 8 months at 3.8 GHz, 1.59 volts vcore :D, then took it back to stock for a while. The only difference I've noticed is that the temps have increased overall, even at stock. Could this be due to electromigration?
     
  11. Wile E

    Wile E Power User

    Joined:
    Oct 1, 2006
    Messages:
    24,324 (8.17/day)
    Thanks Received:
    3,778
    I, on the other hand, have killed 3 Brisbanes at 1.55V under water cooling in less than a year.

    No, temps have little to do with electron migration.
     
  12. Jstn7477

    Jstn7477

    Joined:
    Aug 30, 2009
    Messages:
    3,916 (2.05/day)
    Thanks Received:
    1,598
    Location:
    Sarasota, Florida, USA
    It sounds like your thermal grease has dried up. You should probably remove the old paste and apply new paste.
     
    Crunching for Team TPU More than 25k PPD
  13. cadaveca

    cadaveca My name is Dave

    Joined:
    Apr 10, 2006
    Messages:
    14,098 (4.47/day)
    Thanks Received:
    7,282
    Location:
    Edmonton, Alberta
    It really depends on the doping of the silicon, how much EM should be of a concern. AMD already talked about what they did to deal with EM, many moons ago, on the 940BE press events.

    And temps have ALOT to do with EM. The hotter silicon gets, the more current it will leak. The higher the leakage, the more EM becomes an issue.

    Here's a great site for specific infos:

    http://www.iop.org/EJ/journal/SST

    Some more reading:

    http://necsi.edu/projects/yaneer/physrevlett52-1129.pdf

    http://www.ioffe.rssi.ru/SVA/NSM/Semicond/Si/electric.html

    http://www.chironholdings.com/chirontechnology/references/App Notes/Via EM.pdf

    http://staff.wwcc.edu/jeffrey.watson/CT105/Chapter08.ppt
     
    Paulieg says thanks.
  14. Error 404

    Error 404

    Joined:
    Apr 14, 2008
    Messages:
    1,777 (0.73/day)
    Thanks Received:
    169
    Location:
    South Australia
    I've already tried that, it makes little difference. My CPU will reach about 55C under load, but before I did a "suicide run" at 4.1 GHz (max speed for this chip), it used to reach a max temp of about 48C or less. I plan on replacing it soon anyway, darn thing wont clock for nuts. :banghead:

    So how long would it last at around 1.51-55 volts vcore? I might as well clock it back up to around 3.8-4 GHz, since it does give a nice speed increase.
     
  15. cadaveca

    cadaveca My name is Dave

    Joined:
    Apr 10, 2006
    Messages:
    14,098 (4.47/day)
    Thanks Received:
    7,282
    Location:
    Edmonton, Alberta
    Error, your "suicide run" may have melted the IHS solder a bit, causing the higher temps.

    anyway, are you talking about Intel or AMD chips? They are VERY different.
     
    Error 404 says thanks.
  16. Jstn7477

    Jstn7477

    Joined:
    Aug 30, 2009
    Messages:
    3,916 (2.05/day)
    Thanks Received:
    1,598
    Location:
    Sarasota, Florida, USA
    The thread started off about AMD and their SOI process (presumably from 754 through AM3 processors).
     
    Crunching for Team TPU More than 25k PPD
  17. cadaveca

    cadaveca My name is Dave

    Joined:
    Apr 10, 2006
    Messages:
    14,098 (4.47/day)
    Thanks Received:
    7,282
    Location:
    Edmonton, Alberta
    AMD has changed the doping in thier process many times over the years. We know the C2 Phenom2 are ok at those voltages, but no info is really out yet for C3.

    AFAIK, that's what the revision is all about, different doping. Based on the scaling we've seen so far on C3, it seems to leak less(scales better with voltage, runs cooler), so it might even handle more...but at the same time, it might not.

    I've got a C3 here to put in my DX11 rig, but I haven't even pulled it out of the box yet. hopefully Santa wil lgive me the last parts I need.:laugh:


    Typically, with all silicon, when you stop to see linear scaling, you are entering territory that may cause EM.

    But like mentioned, even with les EM can be an issue, but this is typically due to poor PWM design, or "dirty" power.

    Error only has INtel rigs listed, so I had to ask.
     
  18. TheMailMan78

    TheMailMan78 Big Member

    Joined:
    Jun 3, 2007
    Messages:
    21,185 (7.75/day)
    Thanks Received:
    7,708
    In other words each chip is completely different. So saying 1.6v on any given chip as long as the temps are fine could leave you with a unstable or worse dead system sooner than later.
     
  19. cadaveca

    cadaveca My name is Dave

    Joined:
    Apr 10, 2006
    Messages:
    14,098 (4.47/day)
    Thanks Received:
    7,282
    Location:
    Edmonton, Alberta
    100% correct...unless you know how the doping affects the silicon. AMD has told us C2 is ok @ 1.55v, but yeah, 1.6v may be close. Of course, the binning process helps us know which chips will handle it, and which will not...those that cannot are typically downbinned into tri- and dual- cores.

    That said, it will be quite some time before that damage is seen...years most likely. We do feed the internal PLL 2.5+ volts...this is much higher than Intel's 1.9v, which should tell you something about the strength of AMD's silicon in comparison to Intel. AMDs chips also require less current(albeit a mediocre amount) to "switch" states.

    The overall temps a cpu runs at is also an indicator...higher leakage=more EM. Intel chips run hot, AMDs much cooler, although cache also plays a role in temps too. It's a complex thing.

    Finding that point is why they run the chip under LN2 and such...once temps are eliminated as a factor, you can investigate what point voltage will have an effect regardless of temps. When AMD says 1.55v is ok, I assume they've done this...and when they said this, they were showing chips sub-zero.
     
    Last edited: Nov 30, 2009
  20. Binge

    Binge Overclocking Surrealism

    Joined:
    Sep 15, 2008
    Messages:
    6,982 (3.08/day)
    Thanks Received:
    1,752
    Location:
    PA, USA
    You should have stopped after '100% correct' because most of what you wrote which was true was already stated, and what hasn't been written is completely misleading.

    Lower leakage only means that there's a greater oxide barrier used to create a gate, but how much stronger are you implying the gate needs to be before it sees "less" EM... even if the gate is stronger it is ignorant at best to assume that EM is "better" with gates that leak "less". There is a point in all variables such as: temperature, gate leakage, process, and materials that will contribute to EM, but hinting vaguely at how much is not something you should state without the facts. Assume EM will occur under any situation beyond the rated specs of your chip, or you're fooling yourself into thinking overclocking/overvolting is safe on every chip.
     
  21. cadaveca

    cadaveca My name is Dave

    Joined:
    Apr 10, 2006
    Messages:
    14,098 (4.47/day)
    Thanks Received:
    7,282
    Location:
    Edmonton, Alberta
    You bet. But as mailman has said, just putting voltage on the chip can cause EM, even @ .00001v. If you want ot be completely correct, EM can happen with the chip just sitting on your desk, with radiowaves passing through it/over it.

    I'm kinda using warranty period length as an assumption...

    Lower leakage can mean that the covalent bond between the atoms can more easily accept the passing elctrons, and more leakage can mean that there are more atoms with all electron levels filled.

    To me, "electron migration" is the movement of electrons into these "holes" in the atom, which thereby restricts electron flow, and the electrons will move to the easiest path...which may not be the path of the circuit. This can happen with any amount of electrons..whether a constant supply, or just a single electron passed over the circuit.




    Edit: I guess, to me, this just seems like fearmongering. If you overclock, you've already voided your warranty, and should be prepared to replace your chip, out of pocket. EM is going to happen...whether it's 20 years from now, 3 years, or tomorrow. Yes, the chances of it happenning increase as you raise the voltage, but like stated...each chip is different. It just might happen at stock...it might happen underclocked, or overclocked. It's what kills chips, period. Overclocking, in any form, is going to increase this risk...doesn't have to be with more voltage...doesn't even need overlcocking.

    So even saying 1.425, or whatever, is safe, isn't really true either, if you want to look at like you guys are. Theres' a chance I will walk out my front door and get struck by lightening, too, though, but I don't worry about that either...my warranty was void as soon as I mounted the chip without factory cooling, so it's unimportant to me. I've yet to hear of single C2 phenom II dying, except at REALLY excessive votlages..1.75++, except for erocker's here...

    Might as well just never overclock, never use anything other than standard cooling, with what I see you guys saying. How many of you are gonna do that?
     
    Last edited: Nov 30, 2009
  22. Binge

    Binge Overclocking Surrealism

    Joined:
    Sep 15, 2008
    Messages:
    6,982 (3.08/day)
    Thanks Received:
    1,752
    Location:
    PA, USA
    Understanding the dangers allow for adventurous persons to assess risks. Simply put most of us OC with the knowledge that it is outside of the norm. You would be wrong in assuming electron migration = leaking electrons. Maybe this is the wrong term because I've always known this to be known as "Electromigration". It's a term for describing a phenomena associated with moving electrons through very small channels of metal. The metal channels will accept the current but the electrons will actually cause the metal to move molecule by molecule. This usually occurs in a metal trace.

    Here's a picture in time lapse of a metal trace being fed high currents:

    [​IMG]

    Video and images: http://people.ccmr.cornell.edu/~ralph/projects/emig_movies/

    The above statement is absolutely complete bull. I do not appreciate people making bold claims especially when they are even confusing EM (electromigration) with EMR (electromagnetic radiation).
     
    Last edited: Nov 30, 2009
    extrasalty and Hunt3r say thanks.
  23. cadaveca

    cadaveca My name is Dave

    Joined:
    Apr 10, 2006
    Messages:
    14,098 (4.47/day)
    Thanks Received:
    7,282
    Location:
    Edmonton, Alberta
    Exactly, but the molecule movement is caused by those stray electrons filling the outer levels of the covalent bond, causing the molecular bond to break...

    I've confused nothing...the end effect is the same. You assume too much here...

    ELECTROMIGRATION...Migration caused by electrons. Electromagnetic radiation is merely a source of electromigration. NO confusion at all.;)


    :toast:
     
  24. Binge

    Binge Overclocking Surrealism

    Joined:
    Sep 15, 2008
    Messages:
    6,982 (3.08/day)
    Thanks Received:
    1,752
    Location:
    PA, USA
    There is a confusion. You are a confusing person. On one hand you want people to feel safe while overclocking, and on the other you try to confuse the average user with a complex wording of a simple phenomena. You are wrong about EM being caused by stray electrons. The electrons are being pushed through the metal in the form of current, and there needs to be a current before there can be Electromigration. In fact all Electromagnetic Radiation is AC in nature and would have to be coming from a high power transformer for a current to even exist. EM will never occur from natural or household sources with current day consumer electronics/computer parts.

    You would be correct about electrons causing the break in a trace of metal, but you are again wrong about how it happens. It only occurs in areas where the trace is too small or becomes small after years of passing electrons. Like I've said before this has nothing to do with gate oxide density which would affect leakage. The idea is these traces are meant to carry electrical currents. There is supposed to be a saturation of electrons as that's what happens when you pass current through metal, but while you grasp the concept of molecular structures you're failing to prove your original statement. Instead you are rewording a simple phrase into something much more complicated and I doubt you are even qualified to make that statement.

    Thank you for your PM earlier but there's a bit of an issue,

    In order for the right info to come out someone does not need to be wrong. There only ever needs to be the truth. If you are purposefully pushing out the wrong information to entice conversation than you are doing the people of TPU a great disservice. I don't take your comments as disrespectful, but your blatant disregard for being at least semi-accurate or poignant is room for me to consider your responses to be more trollish than helpful. I don't like to throw around the term troll, and I hope if you are not then you assume that you are not. The issue I have is not with your source (even though you provided none I assume you may have some sort of background... at least a highschool degree since you use the term Covalent Bond) but why you present your information and what you hope to achieve with your bias which has proven to me and my sources to be harmful. Remaining innocent or helpless in a discussion like this is not what a reader should feel. This subject is not beyond anyone, and it's an issue that can be explained with little to no special vocabulary.
     
    Last edited: Nov 30, 2009
  25. DanishDevil

    DanishDevil

    Joined:
    Oct 6, 2005
    Messages:
    10,217 (3.06/day)
    Thanks Received:
    2,116
    Location:
    Newport Beach, CA
    I'll just chime in real quick here. I used to go through E8500's almost daily. I probably clocked over 50 of them. I noticed after about 6 months of running my "golden" chip at 4.75 GHz @ 1.52v under water, for occasional benches, and 4.5 GHz @ 1.45v under water, it needed a voltage bump to keep hitting those speeds.
     

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

Share This Page