An Analysis of the Z390 Socket Proves the Extra Pins Aren't Necessary

Discussion in 'HardForum Tech News' started by cageymaru, Jan 3, 2019.

  1. pcgeekesq

    pcgeekesq [H]ard|Gawd

    Messages:
    1,403
    Joined:
    Apr 23, 2012
    Maybe you should note that his OVERNIGHT test only had the 18 power pins taped off.
    And he didn't prove that the pins could "handle double the amperage" under the conditions that matter: in a hot case under a real workload for an extended period of time.
    Demonstrating you can run a pin above it's spec'd current capacity for a limited time in an open environment at room temp proves nothing.
    Did he even bother to measure the temperature of the PCB with a FLIR camera while running that way, to check for hot spots?
    'Cause you know, that kind of thing matters to real product engineers-- but not to people who merely play one on YouTube.
     
  2. pcgeekesq

    pcgeekesq [H]ard|Gawd

    Messages:
    1,403
    Joined:
    Apr 23, 2012
    The problem is with people making statements of supposed fact, like "the pins aren't needed," when they are ignorant of the engineering issues that must be considered when deciding how many pins a product should have.

    I'm not saying the pins are needed. I'm saying we haven't seen any credible evidence that they aren't.

    Understand: Intel's important clients (MB makers, Dell, and so on) don't like having to design and validate new MBs for new sockets. It's expensive and risky.
    That said, odds are pretty good it was going to be necessary when the 10nm CPUs come out, and they were supposed to be out now.
    So what may have happened is this:
    Intel provided a new socket layout and chipset for 10nm, probably a year or more before launch, and everybody designed it into their products.
    Then the fab guys dropped the ball on 10nm, and AMD produced good new products.
    Intel scrambled, and produced the 14nm 9000 series parts.
    But the MB manufacturers already had the 300-series chipset designs ready to go, so the 9000 series was targeted for that.
    Because after all, the 10nm parts where going to be in mass production Real Soon Now.
    And people would be pissed if they had to change the socket again when that unpredictable Real Soon Now became Now.​

    Was that what happened? Will 10nm CPUs drop into every 300-series MB with no problems?
    We'll see.
     
    VIC-20, Armenius, juanrga and 3 others like this.
  3. Nightfire

    Nightfire [H]ard|Gawd

    Messages:
    1,447
    Joined:
    Sep 7, 2017
    SEE BELOW

     
  4. Hagrid

    Hagrid [H]ardness Supreme

    Messages:
    7,996
    Joined:
    Nov 23, 2006
    So you know 100% what their decision was and why? Then can you provide a link, email or letter that supports your statement?
    You could have doctorates and still not know the answer. You are guessing, just like everybody else. ;)
     
  5. IndyColtsFan

    IndyColtsFan Limp Gawd

    Messages:
    282
    Joined:
    Jul 4, 2017
    *facepalm*
     
  6. pcgeekesq

    pcgeekesq [H]ard|Gawd

    Messages:
    1,403
    Joined:
    Apr 23, 2012
    Do yourself a favor in the future, and read the entire thread before responding.
    I addressed your concerns 12 minutes before you posted.
     
  7. Falkentyne

    Falkentyne [H]ard|Gawd

    Messages:
    1,551
    Joined:
    Jul 19, 2000
    No point. I'm just glad this forum is laid back and Kyle and Dan allow people to just get destroyed by facts and logic here.
    If this were LTT or overclock.net there would have been a bunch of permanent bans by now.
    Hardforum is probably the last old school forum left. Reminds me when everyone was playing Quake 3 and most people had thick skin.
    Anyway carry on.
     
    AlexisRO, Hagrid, pcgeekesq and 2 others like this.
  8. FrgMstr

    FrgMstr Just Plain Mean Staff Member

    Messages:
    48,094
    Joined:
    May 18, 1997
    Please be civil and stop attacking each other like children. This goes for everyone.
     
    {NG}Fidel likes this.
  9. juanrga

    juanrga Pro-Intel / Anti-AMD Just FYI

    Messages:
    2,489
    Joined:
    Feb 22, 2017
    That is the problem today on Internet. It is sad that sites with deep technical analysis (usually sites have been in the business for decades) are forced to close due to financial problems, whereas a plea of youtubers become celebrities thanks to videos with little technical content or plain wrong stuff. Welcome to the era of the mediocrity!
     
    Last edited: Jan 7, 2019
    auntjemima likes this.
  10. dook43

    dook43 2[H]4U

    Messages:
    2,904
    Joined:
    Sep 9, 2005
    What you call "shrewd business tactics" I call illegal. Such as bribing OEMs to exclude AMD processors entirely from lineups and paying game developers to optimize tessellation invisibly in a way that completely gimps AMD architecture. Not to mention Nvidia having AMD infiltrated with spies. These actions over a period of years significantly harmed AMD's ability to compete financially.

    It's arguable that hiring Dirk Meyer as CEO may have been more detrimental than either of those two things I mention, but these well-publicized incidents are why I am anti Intel and anti Nvidia.
     
  11. Dan_D

    Dan_D [H]ard as it Gets

    Messages:
    53,451
    Joined:
    Feb 9, 2002
    I'm not trying to debate the morality of company decisions. It wasn't the point of my statement. Also, anti-compete agreements and a host of other shit that's anti-competitive is actually quite legal. There are things done that were obviously illegal in the UK etc. because Intel was fined for that. However, companies infiltrating other companies with spies is not as uncommon as you think. If a company like ASUS can keep design elements of their motherboard lineup secret for 6 months they are happy. Each company buys the others products and attempts to reverse engineer them etc. I've heard plenty of horror stories told to me in confidence about how anti-competitive these companies really get with each other.

    None of them play nice. None of these companies are righteous underdogs that have clean hands are beyond reproach. You want to debate morality, here's this: Every company I know of in this industry has done things which many of you would find morally questionable. Obviously, this is a sliding scale with companies like Intel being the worst offenders, but AMD and others are far from being pillars of virtue.
     
    Crosshairs likes this.
  12. schmide

    schmide Limp Gawd

    Messages:
    202
    Joined:
    Jul 22, 2008
    I would say these same sites that did "deep technical analysis" were often beholden by NDAs and strong arm tactics (blacklisting), that skewed what they said and what could not say.

    One site that survived famously resisted such dirty tactics.

    The irony in the above statement is what Der8auer did was deep technical analysis. He took apart hardware, showed tests, showed results.

    If mediocrity looks like this, bring it on.

    We found out what being Principled on Technologies showed.

    Edit: youtubers celebrity punchline inserted
     
    Last edited: Jan 7, 2019
  13. pcgeekesq

    pcgeekesq [H]ard|Gawd

    Messages:
    1,403
    Joined:
    Apr 23, 2012
    And then reached an inflamatory conclusion for which he had absolutely no evidence, because he doesn't understand package design.
    Fake news, essentially.

    And that's a problem.
     
    Armenius and IndyColtsFan like this.
  14. juanrga

    juanrga Pro-Intel / Anti-AMD Just FYI

    Messages:
    2,489
    Joined:
    Feb 22, 2017
    He didn't did any deep technical analysis. That is just the root of problem I am illustrating...
     
  15. schmide

    schmide Limp Gawd

    Messages:
    202
    Joined:
    Jul 22, 2008
    Has any other site done analysis equal or greater that his? If so I would love to see it.
     
  16. schmide

    schmide Limp Gawd

    Messages:
    202
    Joined:
    Jul 22, 2008
    I do agree his conclusion is a bit over the top.

    I still like the methodology.
     
  17. Hagrid

    Hagrid [H]ardness Supreme

    Messages:
    7,996
    Joined:
    Nov 23, 2006
    if you are qualified, why not debunk his info? Why not show us your investigation? Post the data on your findings?
     
  18. mouacyk

    mouacyk n00b

    Messages:
    56
    Joined:
    Dec 5, 2015
    On hwbot, early testers of 8700k on z170 boards who did not tape off certain pins ended up charring them irreversibly.
     
  19. pcgeekesq

    pcgeekesq [H]ard|Gawd

    Messages:
    1,403
    Joined:
    Apr 23, 2012
    I debunked his conclusion, not his "info."
    Because I have a real job to do: getting people patents on their cutting-edge computer, circuit, and semiconductor inventions.
    Obviously, the rules governing attorney conduct bar me from going into the details of that.

    Besides, you don't have to conduct your own analysis to recognize the obvious flaws in someone else's.
    If you had any real experience in mathematics, science, or engineering, of any kind, you'd know that.
    People debunk "proofs" and discredit published results all the time, using nothing but logic and the pertinent prior art.
     
    Armenius and juanrga like this.
  20. {NG}Fidel

    {NG}Fidel [H]ardness Supreme

    Messages:
    5,919
    Joined:
    Jan 17, 2005
    He mentions in the video that he would still need to do more testing.
    You keep misstating his video and then using those narratives to attempt and trash him.
    You clearly did work at Intel.
     
    schmide and Hagrid like this.
  21. Hagrid

    Hagrid [H]ardness Supreme

    Messages:
    7,996
    Joined:
    Nov 23, 2006
    So in other words you have nothing, same as everybody else. Gotcha. :)
    How do you know I do not have any experience in any of those? Guessing again?

    He showed something and if somebody can prove him wrong, I am all for it.
     
    schmide likes this.
  22. pcgeekesq

    pcgeekesq [H]ard|Gawd

    Messages:
    1,403
    Joined:
    Apr 23, 2012
    A recent study showed that a persistent characteristic of fanatics is that it's impossible to change their mind about anything,
    no matter what logic and evidence you present them with. Clearly, we're seeing that in the Intel-bashers in this thread.

    As the title of this thread reflects, de8auer claimed to have proved the extra power pins weren't necessary. He didn't.
    All he proved is that he doesn't understand the issues involved in power delivery to fast-switching circuits.

    Going into the details on the mistakes he makes:
    1) his test of a pin involves SOLDERING it to HUGE THICK WIRES. (8:29) The cross sectional area of the solder joint is huge compared to the contact area of the pin to the CPU in real life, so already the "experiment" is irrelevant, because the resistance is going to be nothing like that of a socket in use.

    2) He talks about determining the temperature rise for that pin using this setup, but he's doing it for an isolated pin in free air, not a pin packed inside a lot of other hot pins inside an insulating socket. Plus he just ignores that those thick (relatively) cables are carrying a lot of heat away. It's similar to comparing the temp rise of an M2 SSD in free air to that of one stuffed underneath an RTX 2080. Clearly, he doesn't have a clue about thermal issues. He talks a lot, but it's all ignorant nonsense. "The pin ... can handle 5A no issue" he says: nonsense. Not in a socket, with a tiny little contact to a CPU pad.

    3) "Its only 10% higher current" (13:00) which means 20% more heat. For a constant resistance, power rises as the square of current. He seems not to realize that.

    4) Not once does he mention package inductance. That's a much harder issue than mere DC current, but his whole focus is DC current. That's because he doesn't have any clue about the issues here. But still, he concludes (14:56) "We wanted to check whether there is a valid reason why we would change the socket ... but ... there is not really a need to do that."

    His arrogance is made plain at 15:50: "for sure it would not have been needed." That's only because he doesn't have clue 1 about the challenges that CPU designers face in getting good clean power to a circuit that can generate big fast changes in the amount of current it draws.

    All he cares about is DC current (the easiest problem to solve) and pin heating (which he uses a completely bogus experiment to try to estimate.) He's an amateur (does he even know that package leads have non-trivial inductance?) and his conclusions reflect it.

    They are just plain not supported by his experiments.
    Like I indicated before, if he was a packaging engineer and wasted his time doing these kinds of ignorance-based experiments, he'd probably be fired, and rightfully so.
     
    Last edited: Jan 7, 2019
    Armenius, juanrga, auntjemima and 4 others like this.
  23. {NG}Fidel

    {NG}Fidel [H]ardness Supreme

    Messages:
    5,919
    Joined:
    Jan 17, 2005
    We get it you can ad-hom. Which Kyle asked us not to do in this thread....
    Now we are all fanatics if we ask you something....
    Point is de8auer says a lot about needing more testing.
    The title of this thread is not his opinion or statement but just some text a user here choose.
     
    Nightfire likes this.
  24. naib

    naib [H]ard|Gawd

    Messages:
    1,255
    Joined:
    Jul 26, 2013
    Actually what he showed is inconsequential.

    Half the entire video he concentrated on the current that will pass through the pins. The other half was some arty style "look at this CPU, Look at me sanding"

    He used 4mm leads that are rated at 15A continuous (so about 18AWG) soldered to a contact which has specification of < 19mR (EOL aging to 100mR...)... less than 80mW in that pin while the leads will be drawing the heat away.

    https://www.intel.com/content/dam/w.../guides/4th-gen-core-lga1150-socket-guide.pdf

    Do you have any idea how much current a piece of copper can handle?


    What he has failed to mention is the distribution of the power to the internals of the CPU. Did he take the lid off? did he note where the wirebonds were going? was aspects of signal integrity done? NO ...

    Am I saying that Intel has some special layout that definitely required the additional distribution of power across the die? no, certainly not. However, to stand behind what this sharletian was stating, and from someone not using an anti-static strap while handling the electronics is laughable
     
    Armenius and auntjemima like this.
  25. pcgeekesq

    pcgeekesq [H]ard|Gawd

    Messages:
    1,403
    Joined:
    Apr 23, 2012
    1) point is, de8auer seems to be so ignorant of proper packaging design that he doesn't know lead inductance "is a thing."
    2) you must not have watched the video, but let me give you the money quotes: (14:56) "We wanted to check whether there is a valid reason why we would change the socket ... but ... there is not really a need to do that." (15:50): "for sure it would not have been needed."

    "For sure." Proof positive of his ignorance and conceit: de8auer doesn't even know what he doesn't know.
     
    Armenius and IndyColtsFan like this.
  26. pcgeekesq

    pcgeekesq [H]ard|Gawd

    Messages:
    1,403
    Joined:
    Apr 23, 2012
    Your posts here makes it plain that you don't.
    But that's assuming you're not trolling. Are you?
     
    IndyColtsFan likes this.
  27. pcgeekesq

    pcgeekesq [H]ard|Gawd

    Messages:
    1,403
    Joined:
    Apr 23, 2012
    What part of "just waiting" includes posting personal attacks on me?
    And I've posted plenty of substance, none of which has been demonstrated to be wrong.
    de8auer's conclusions in the vid are fake news. "For sure."
     
    auntjemima and IndyColtsFan like this.
  28. {NG}Fidel

    {NG}Fidel [H]ardness Supreme

    Messages:
    5,919
    Joined:
    Jan 17, 2005
    You aren't sounding convincing just upset and angry and condescending to those asking questions.
     
  29. Hagrid

    Hagrid [H]ardness Supreme

    Messages:
    7,996
    Joined:
    Nov 23, 2006
    I was wanting facts not more guesses. You have proven that he is either right or wrong.
    Personal attacks? Where?
     
  30. pcgeekesq

    pcgeekesq [H]ard|Gawd

    Messages:
    1,403
    Joined:
    Apr 23, 2012
    I gave you plenty of facts, including quotes from the video.
    But go research the issues surrounding CPU power delivery yourself: you know, on the Internet?
    Get a basic understanding of the issues. Then come back and ask intelligent questions.
    You can do that, right? I'll wait.

    Or are you just trolling?

    "Since you are the smartest on the internet, it should not be that hard!" -- you posted that. Why?
    Do you actually believe I'm the smartest person on the Internet?
    Or are you insulting me by sarcastically implying I think I am?
    If so, that's a lie: there's at least a half dozen people I'm pretty sure are smarter than me on the Internet.
     
  31. Hagrid

    Hagrid [H]ardness Supreme

    Messages:
    7,996
    Joined:
    Nov 23, 2006
    Well I assumed you were since you think everybody is wrong and you are right. Even with 0 proof you have provided, you still think your are correct.
    Words are far different than actual testing and first hand knowledge. When you have one or both of those, then come back with facts.
    Until then he isn't wrong or right, yet.
     
  32. pcgeekesq

    pcgeekesq [H]ard|Gawd

    Messages:
    1,403
    Joined:
    Apr 23, 2012
    Wow, that's wrong on so many levels. Factually, logically -- how old are you?
    I get the feeling I worked as CPU designer for longer than you've been alive.
     
  33. Hagrid

    Hagrid [H]ardness Supreme

    Messages:
    7,996
    Joined:
    Nov 23, 2006
    3.50.

    Wrong as in your definition? I will take that as a compliment. The guy will hopefully post more info as he does more testing.
     
  34. pcgeekesq

    pcgeekesq [H]ard|Gawd

    Messages:
    1,403
    Joined:
    Apr 23, 2012
    Kyle liked my post of 2:11PM Pacific Time today.
    That trumps what you think of it, and what you think of me.
    One indication of respect from Kyle is worth a whole Twitter-pocalyps of negative comments from anonymous nobodies. :)
     
    Falkentyne and IndyColtsFan like this.
  35. Hagrid

    Hagrid [H]ardness Supreme

    Messages:
    7,996
    Joined:
    Nov 23, 2006
    Yeah well. Hmm. Since he knows this stuff more, then Kyle is right!
     
  36. FrgMstr

    FrgMstr Just Plain Mean Staff Member

    Messages:
    48,094
    Joined:
    May 18, 1997
    I love a lot of the work that dbauer does, but his analysis on this is flawed beyond belief.
     
    Armenius, juanrga, aaronspink and 5 others like this.
  37. naib

    naib [H]ard|Gawd

    Messages:
    1,255
    Joined:
    Jul 26, 2013
    He is completely wrong on the avenue he is pursuing.

    He went in questioning whether the reserve pins now as power is justified - reasonable supposition.
    He pursued it by showing some 1st year undergraduate test methodology with regards to DC capability
     
    Armenius and pcgeekesq like this.
  38. FrgMstr

    FrgMstr Just Plain Mean Staff Member

    Messages:
    48,094
    Joined:
    May 18, 1997
    I don't KNOW if he is wrong or right, but the way he came to his conclusions is incredibly flawed. My opinion is that he is way off base on this.
     
    Armenius and pcgeekesq like this.
  39. 1_rick

    1_rick Limp Gawd

    Messages:
    390
    Joined:
    Feb 7, 2017
    Not an EE, obviously, but you're talking about a 3% reduction in the number of ground pins. That's a huge difference relative to running a 33psi tire at 20psi, and your analogy would be a lot better if you talked about running a 33psi tire at 31psi.

    (ETA: None of this is to try to argue your overall point, just the details of the analogy)
     
    Last edited: Jan 7, 2019
    schmide likes this.
  40. pcgeekesq

    pcgeekesq [H]ard|Gawd

    Messages:
    1,403
    Joined:
    Apr 23, 2012
    The world is simply more perverse than amateurs imagine. Professionals worthy of the name rely on a vast body of past community experience that alerts them to that perversity.

    Old timers remember the tin/gold DRAM DIMM/socket debacle, and people older than that remember the copper/aluminum house wiring debacle (the latter of which actually killed people). Both were cuased by something that even the pros in the respective fields didn't realize: even a "non-corroding" metal like gold or aluminum will react with another metal when you put the two together and pass current through them in the presence of even a little moisture.

    I've seen DIMMs with gold from a socket chemically fused to the tinned leads, and I've seen circuit breaker panels that caught fire because an aluminum wire had been clamped (without any anti-oxidant goop) to a copper bus bar. It seems even the pros didn't realize these problems would occur, back in the day, but they do now. Would an amateur? Maybe. Maybe not.
     
    Armenius and Falkentyne like this.