Help with voltages/temp delided i7 6700k

Discussion in 'Overclocking & Hardware' started by paulos, Dec 16, 2016.

  1. OP
    OP
    paulos

    paulos Member

    Joined:
    Dec 16, 2016
    Messages:
    22
    haha yeah i will probably go off and search more about others on 1.5v overclocks as messing something up now would not be worth it!

    http://imgur.com/URixUiN thats a link to 4.9Ghz

    when i collect more data my graph will show how the cpu responds to various oc's so far 4.6 is the sweet spot for voltage required from the jump from 4.5 to 4.6 is very minimal but before that and after that the voltage increase needed is large and increasing as the speed increases ( as expected ).

    yeah let me know and il give them a crack, that being said this isnt a claimed p95 24hour stable OC simply a OC for my needs ( just to clarify for anyone who might read this post and mis-interpret what my overclocks mean).
    I wonder if i could get the same OC's with less voltage by changing Baseclock and multi rather than just multi...
     
  2. wraith.50

    wraith.50 Member

    Joined:
    Sep 25, 2016
    Messages:
    115
    Basically it comes down to having enough voltage for a given frequency, under maximum load that determines stability.

    If you really want to see if your cpu is completely stable at a certain frequency and voltage, then test it under Prime95 V28.7 or higher.

    This will generate a lot more load and heat, than what Realbench does, so first try it out at stock clocks and see what temps you get and then increase your overclock by 0.1ghz at a time, until you get to 85-90 degrees max temps, when running Prime95.

    Another quick but good test of your overclock stability, under a lot of load, is Intel XTU Benchmark. This also places a lot more load and generates a lot more heat from your cpu, than Realbench does.

    You may find that your overclock at a certain voltage, will pass Realbench, but fail in Prime95 or XTU Benchmark.
     
  3. bennyg

    bennyg Member

    Joined:
    Dec 16, 2005
    Messages:
    3,153
    Location:
    Melbourne, Oztraya
    Yes I was about to say that 90W consumption @ 4.7GHz is a very light benchmark. I reckon you'll bsod instantly on a "power virus" benchmark like AIDA64 FPU but it's so far beyond real world there's debate over whether it's a fair measure of stability (like furmark for gfx cards where all the manufacturers scare users off from using them and implement driver restrictions on normal use of the gpu in those utils)

    Mine's ok so far does 45x at -50mV adaptive which settles to about 1.26V in most benchmarks but AIDA64 FPU stress test needs -25mV before it's stable. But not (yet) being delidded and being in a notebook it's thermally limited, 45x thermal throttles in AIDA64, so I run at 4.0 @ -150mV 24/7 with 4.4 @ -60mV via hotkeys using Throttlestop profiles when I want extra

    e.g. what the tests consume on my notebook @ 4.4/-60mV:
    wprime 1024M bench = 79-80W
    AIDA64 FPU only >110W
     
    Last edited: Dec 21, 2016
  4. OP
    OP
    paulos

    paulos Member

    Joined:
    Dec 16, 2016
    Messages:
    22
    I tried AIDA64 FPU only, on 4.8 and 4.9 ( just a quick 15mins each)
    passed both fine, however max wattage was 105 and 115 respectively.
    I did notice that my multiplier was changing from 46 to 49 ( for 4.9Ghz) temps didnt hit over 74, aida64 didnt report any cpu throttling. apart from reducing multi for power saving options all other stress tests and my games will keep the multi locked at what i set when under load, why the change now..? is it in correlation to a powerdraw safety feature? or a limitation of my motherboard mabye..?
    so far both overclocks are very stable for games but at a max of 68watt draw at 90% cpu load i guess that is to be expected!
     
  5. maldotcom2

    maldotcom2 Member

    Joined:
    Feb 18, 2006
    Messages:
    1,875
    Download Handbrake and cue up a few hours worth of x264 video encodes. x265 might be even better.
     
  6. OP
    OP
    paulos

    paulos Member

    Joined:
    Dec 16, 2016
    Messages:
    22
    FPU only AIDA64
    for some reason my CPU is downclocking its self on this test at 4.8 or 4.9Ghz im not going over 75'c monitoring suggests no cpu throttling
    i tried going from 1.380v to 1.400v for 4.8Ghz and it didnt change a thing

    The cores do the following under FPU only load
    4.8x1 4.5x3
    4.8x1 4.5x2
    4.8x3 4.5x1

    4.9x3 1x4.5
    4.9x2 4.5x2
    and at one stage for 4.9Ghz all cores held 4.4Ghz and then didnt change off 4.4

    what could this be?
    -motherboard powerphase limited?
    -psu not holding stable power under load mabye..? (casecom 800watt max)

    also FPU only is showing 105watts and 117watts for 4.8 and 4.9Ghz ( does that sound like enough power drawn...?)

    im trying to figure this out before i try and find 5Ghz
    that being said in Realbench bench and stress, in x264( tweaktowns skylake tester) , IBT and games the Freq holds just fine.


    and i also have the cryorig r1, so far only a few degree's diff not overly impressed possibly i will attempt other TIM's.
    Cheers Guys (one day il be out of questions)
     
  7. wraith.50

    wraith.50 Member

    Joined:
    Sep 25, 2016
    Messages:
    115
    Can you try run Prime95 and see if it does the same thing with the cpu cores.

    Also check that your CPU Core Current Limit is set to 255W, incase its current throttling the cores.
    Also if you have the setting RSR (Reliability Stress Restrictor), make sure its disabled.

    In regards to your aircooler, you are not going to see a huge difference in temps, even with a Noctua DH15 which is the best air cooler you can get.

    You need to go to water cooling, if you want to achieve a substantial drop in cpu temperature.
     
  8. OP
    OP
    paulos

    paulos Member

    Joined:
    Dec 16, 2016
    Messages:
    22
    So i found my problem, the short/long duration power and some other function were set to auto in bios limiting my wattage draw to 117watts. once this was changed, 4.8 and 4.9 held stable on all cores on all tests
    however as expected with an increase in wattage the temps also grew!
    this explains my deceptively low temps for some of my tests using the Hyper212x ( or atleast any test that needed over 115watts).

    4.9Ghz at 1.460v max temps (83,82,80,76) 2 hour AIDA64 FPU only at 127watts
    P95(26.6 used this version on purpose), RealBench, passmark, x264(tweaktown edit) all came in with lower temps and wattage

    that being said even with unlocked Wattage no Vcore increase was needed to keep the cores at a stable Freq!

    i tried 5Ghz at 1.5v booted fine, ran Aida64 FPU only and it ran for 2mins before the program reported an error but no bluescreen with temps of 87'c. but realbench stress test caused bluescreen. im not willing to go over 1.5v so i wont bother to find a stable 5Ghz...yet

    Just some comments:
    5-8mins CPU settling in period (where the temp rises the efficiency goes down requiring more power which then increases temp further) so from my results atleast testing a safe working temp should only require 10-15min stress test. if any test makes it past this 10-15mins then it also passes at 2 hours (atleast for me) this is with the absolute lowest voltage i can acheive for a given Freq (as in a 0.005v drop for a given freq will not pass at this settling in point).
    can i notice a difference from 4.8 to 4.9..? yes but at the cost of 12'c not worth it!
    once again the best performance to temp and Vcore ratio is 4.7Ghz on 1.310v at full load under 64'c
    all of my so called overclocking was still as basic as can be just changing Vcore and Freq everything else is set to Auto

    Thanks again everyone for giving me the info i needed to be safe and improve temps ( aka CLU under delid, never a normal TIM)
     
  9. wraith.50

    wraith.50 Member

    Joined:
    Sep 25, 2016
    Messages:
    115
    Glad you sorted out your power problem.

    Out of interest, can you try running Intel XTU Benchmark, with your 4.9ghz and 1.46volts and see what temps you get and if it completes the benchmark successfully.

    Is your 4.9ghz overclock, with your cache at 4.9ghz as well or still at 4ghz stock frequency ?
     
    Last edited: Dec 25, 2016
  10. OP
    OP
    paulos

    paulos Member

    Joined:
    Dec 16, 2016
    Messages:
    22
    you got me! it failed at 1.460v on XTU
    bumped it up to 1.480v to try and got a score of 1290marks at 73'c highest temp ( probably 26-28ish ambient )
    another day il dial in what 4.9Ghz needs voltage wise for XTU as a lowest!

    and cache Freq is stock 4Ghz, no point changing it if it doesnt affect games no..? ( apart from getting better test scores )
     
  11. OP
    OP
    paulos

    paulos Member

    Joined:
    Dec 16, 2016
    Messages:
    22
    actually i dont think its accurate, im having a signed driver issue which i thought disabling on start up would fix but it even now i press benchmark and it looks like its loading or benchmarking but its not...
    and on the benchmark last night i noticed the cpu was never really at 100% but fluctuation madly
    only with XTU am i having this issue!

    trying now on 4.8 1.380v after a few restarts i got the be XTu benchmark to actually start this time but the program stoped 10secs in, HWINFO64 showed on a short CPU useage and then back to idle, even though XTU was showing that it was currently benchmarking
     
    Last edited: Dec 26, 2016
  12. maldotcom2

    maldotcom2 Member

    Joined:
    Feb 18, 2006
    Messages:
    1,875
    Sounds like a niggling instability. Bump the voltage a notch and see if it disappears. Unstable overclocks can corrupt system files or programs, so if you're failing benchmarks on settings that previously passed, that could be it.
     
  13. OP
    OP
    paulos

    paulos Member

    Joined:
    Dec 16, 2016
    Messages:
    22
    Thought id update just in case this information helps others

    looks like XTU just doesnt run with this signed driver issue but i let that go
    nothing else i use has such issues, as to which running my comp at 4.6-4.8ghz has been good for the last month or so.
    no crashes in games like BF1, overwatch or Universe2
    Having now applied Thermal grizzly Conductonaut to the IHS/chip/R1 heatsink
    my temps are pretty good i will never see a temp over 68'c (i average 62ish'c) no matter the ambient (which lately has been 35'c+) at night with cooler temps i wont see anything over 60'c (52'c average) this is with playing hour after hour of BF1
    this is running 4.8ghz at 1.390v, the cryorig R1 fan curve is steep running on average 80-90%
    running at 4.6ghz is a dream with temps 50'c and under (gaming load)


    just as a comparison a friend of mine got his 6700k and can so far manage 4.6ghz on 1.370v but on custom water loop temps around 65ish (full gaming load) not delided, no liquid metal used.

    removing CLU from the ihs, chip and heatsink was fairly easy it does behave like a magnetic liquid it likes to ball up but not come off the surfaces, but ISO-proply does the trick. no staining on any of the surfaces, but one thing to note is the CLU between the IHS and chip did at one point run slightly off the chip onto the PCB into a little ball i had no idea until i took it apart to reapply but it made no difference to temps(indicating even less than the small amount of clu i used is still too much!). the nail polish i used over the two scratches from deliding look unchanged.

    i have also used CLU on my G1 1080, pretty good results and after a week check no eating into the heatsink (just had to apply nail polish around the chip its self)

    considering that i got into computers with this build (3months ago), liquid metal and deliding is pretty noob safe iff you do enough internet searching, make a plan, and take your time (in reference to spending 2 hours to de-lid with a stanley knife blade, bad choice of blade btw).
     
  14. bennyg

    bennyg Member

    Joined:
    Dec 16, 2005
    Messages:
    3,153
    Location:
    Melbourne, Oztraya
    I delidded mine with a 3d printed tool, the guy on these forums sells them on ebay. Less than 30 seconds to complete and only because I took it real slow with the vice.

    I dropped 20C at least by doing it with conductonaut which I also have on top of the IHS.

    For a desktop socket + SLI laptop this thing is now extremely quiet and cool at idle
     
  15. OP
    OP
    paulos

    paulos Member

    Joined:
    Dec 16, 2016
    Messages:
    22
    finally some cooler temps so i thought i would try 5Ghz on AIR!!!

    at 20c ambient iv managed 1.55v 5Ghz max temps of 67'c with heavy games and 75'c with P95
    as expected not worth it when 4.9ghz comes at 1.48v

    The voltage isnt dialled in, its just the first one i chose based on previous increments needed for 100mhz gain, i imagine there's not much left in reducing Vcore though.
    4.6Ghz has been dialled in at 1.23v temps never above 48'c

    Thought id share in case anyone else was feeling risky with the colder weather, colder temps= lower resistance= higher stability for a given voltage.
     

Share This Page