Help with voltages/temp delided i7 6700k

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

  1. paulos

    paulos Member

    Joined:
    Dec 16, 2016
    Messages:
    22
    Hey guys hopefully im posting in the right place

    i have delided (razor method) my 6700k, two tiny scratch's of pcb ( put nail polish over) ( mx-4 line method for chip, x method for heatsink(hyper 212x)
    This is my first
    -Delid
    -heatsink install
    -thermal paste application
    -soon to be overclock

    Comp for past 4 days with heavy game use reaching temps of 58-63'c (stock 4ghz)
    however at idle the temp can spike from its normal of 35ish to 55, which is consistent with a larger observed workload at the time of the spike. (ambient temp of 24-34'c)

    my stock voltage on core temp and cpu-z reads 1.1998
    even when in a game at 90-100% cpu loads the voltage doesnt change it fluctuates more readily but stays the same average.

    today i have turned on my gaming boost in bios( msi pro carbon) which set cpu to 4.4ghz at 1.35v.
    idle temp is same as stock, however a quick 50% intel burn test the cpu reached max 66-75'c or 3cores under 69 and one core 75.
    the voltage however reads on core temp 1.2005 but on cpu-z reads 1.312 and through the burn test the reading's didnt change by more than 0.0005volts for both programs.



    my Questions
    -temp fluctuations at idle have i used to much/less thermal paste or bad heatsink install?
    -variation in temp of cores in terms of max temp reached can be 9c diff is it normal?
    -which program is best to trust for cpu voltage readings?
    -how does nail polish fair with heat from cpu?
    -if my voltages really havent changed with the boost why the jump in temp to 75'c?
    -has anything i have posted so far seem consistent with something potentially wrong..?

    Thanks guys- full pc modding overclocking noob

    Update
    today also on 4.4Ghz few hours of gaming 90-100%cpu loads temps came in at 70-76
    also tried 4.4 on stock 1.2 volts and it passed with max temp 65ish, currently on 4.6Ghz at 1.275volts highest spiked temps 71-78 ( most around 72) i did try 4.5 on stock voltage and failed 15mins into a test.

    Update2
    turned cpu fan onto auto 80-100% it was on stock setting of 40% with above tests and gaming
    at 4.7Ghz voltage 1.320 average temp=62 max temp=68-76( always core No1 hotter than rest!) cpu load 90% 15mins
    at 4.8Ghz voltage 1.340 average temp= 65 max temp= 70-79 cpu load 90% 15mins
    for my working overclock i was hoping to get temp spikes less than 85 with voltage less than 1.4 with unknown Ghz.

    i dont really know what to make of my result so far...?
     
    Last edited: Dec 17, 2016
  2. maldotcom2

    maldotcom2 Member

    Joined:
    Feb 18, 2006
    Messages:
    1,838
    You really dived in the deep end here. Your temps are to be expected with those voltages and your cooling solution. They're actually quite good. As far as voltage readings goes, note the difference between VID and Vcore. VID is the voltage the CPU wants, Vcore is the voltage you're giving it. With Skylake, general consensus is your limiting factors are about 1.4v and 80 degrees, unless you don't care about hardware longevity. Disable any of those auto overclocking tools and do it manually.

    I recommend you be more methodical with your overclocking procedure and test thoroughly before proceeding to the next multi. Keep a log of settings (multi and voltage), tests you performed, temps both load and ambient, and the result. Highlight "safe havens" so you can return to those stable settings when you need a break from overclocking and want a reliable PC. You can also quickly pickup where you left off.

    15 minutes of IBT is completely insufficient for stability testing. I've had overclocked PC's that passed many, many hours of Prime95, Aida64, SuperPI and the most intensive setting of IBT, only for them to fail 15 minutes into a video encode. For this reason, I recommend using Handbrake video encoder as your preliminary stability test. Alternatively, Asus Realbench utilises Handbrake as one of the tests.

    Here's how I do it; Raise multi, testing thoroughly after each increase until it fails. Then raise voltage 2 or 3 notches and retest, if it passes raise multi again, and so on. Do this until you reach your max safe load temperature. You can now tweak the voltage down increment by increment testing each time until it fails, then bump the voltage up a notch.
     
    Last edited: Dec 18, 2016
  3. isaakk

    isaakk Member

    Joined:
    Oct 10, 2014
    Messages:
    124
    Location:
    7315 (Turners Beach)
    Pretty much this. Temps sound fairly inline with the voltages you're giving it under what is consider a low-mid range cooler, my 6700K spikes like yours does as well under brief load however this is to be expected. 9°C differences in core temps may be accurate if you're only stressing certain threads/cores, if you were seeing differences like that under 100% load on all cores that i'd start being concerned. As to HW/temp monitoring programs, HWMonitor + CPU-Z are great for a general overview, but i've found that they generally seem to report VID rather than VCORE reliably. The only program that I have found to be reliable for VCORE is AIDA64's monitor section, however this is a paid program.

    It is worth noting that delidding isn't worth the risk and breach of warranty if you're only going to stick MX4 under there. If you really want to make it worth it, Coolabratory Liquid Ultra makes a huge difference with 15-20°C improvements if done correctly and is worth the effort. It's very much a PITA to work with though and dangerous if it gets anywhere other than where you want it.

    For a bit of a comparison, my delidded CLU chip under a H110i will happily sit at 55-60°C under full load @ 4.8GHz, with 1.375v. Have tried pushing it higher, but vcore increases to a not-so-fun level of 1.42v to even boot at 4.9GHz.
     
    Last edited: Dec 19, 2016
  4. TaroT

    TaroT Member

    Joined:
    Jan 18, 2002
    Messages:
    8,568
    Location:
    Hazelbrook nsw 2779
    mine runs 4.5 1.34 and gets around 72 on a hot day now i have the fans sorted.
    that is not delided....i haven't tried putting nearly 4 volts in it though....seems a little extreme wouldn't you say "p maybe you missed a 1 and a decimal point.
     
  5. paulos

    paulos Member

    Joined:
    Dec 16, 2016
    Messages:
    22
    Thanks for reply's guy's
    although it did take 40mins to delid with a shitty Stanley knife blade i figured it would be the logical choice as i intend on staying Aircooled and it looked very do-able.
    thanks for clearing up VID and Vcore

    since the day i wrote this post i have been writing down all my cpu temps, ambient temps, voltages, GHz. have found some very stable safe zones too such as 4.6ghz on 1.3v max spike at 70'c or average 55'c, 4.7 on 1.325v max at 77'c average at 65'c.
    but im having trouble finding a good spot for 4.8! i image plotting a graph with all variables considered would be my best approach to find the 'sweet spot' that point before voltage needs an exponential increase for a 100Mhz increase.

    haha and yeah i can see now IBT is no way to say a system is stable, however running IBT and furmark at the same time as my stability test has afterwards proven stable in games which for my uses is all i need (for now).
    That is running IBT at 95% and furmark on its standard 1080p test
    which throws gpu usage to 90-100% whilst cpu is under a constant 90-100% load which is about 10-20% more load than the games i play.

    CLU and cryorig R1 on the way to hopefully level out my spikes and a little bit for the "what Ghz can we reach at 1.45v". with the addition of this i will try stability testing with more programs "OCCT, Realbench..."

    on idle there can be a 10'c difference in max temp spikes across cores, under load the gap becomes smaller to 6-7'c just for one core the rest are within 3'c of each other at idle or underload.
    an example would be ( 68,75,69,68 ) i would be guessing i just need to redo my IHS and PCB thermal paste and mabye lap the IHS to lower this temp gap.
    If im still finding i have this one core spiking at a higher temp even with a change of TIM and cooler, i will just overclock with that one core being my temp limit. unless this single core temp is a sign of something...?

    i will update when i have more results with my current setup and with the CLU and R1.

    with all of this being said i am still just changing my multiplier and Voltage to overclock and keeping it in dynamic mode. is this still fine or is this what people mean when they say to get out of auto mode..?

    Cheers

    Update for today
    tried 1.35V on 4.8Ghz failed 10mins into combined furmark/IBT avg temps of 75ish with max at 79-88. considering the temps i wont bother with 4.8, instead untill stuff arrives i will hone in on lowest Voltage for 4.5-4.7Ghz and the fps in game correlating to those speeds.
    1.325v and 4.7Ghz passed Realbench max temps of 74-80
     
    Last edited: Dec 19, 2016
  6. maldotcom2

    maldotcom2 Member

    Joined:
    Feb 18, 2006
    Messages:
    1,838
    I'd remount the cooler, since you have that 1 hot core under load. Disregard idle and average temps, they mean squat. You're only interested in the hottest core at 100% load. You still have some voltage headroom but you've reached the temperature limit. You definitely need a better cooling solution, but that's a very good chip to get 4.7 at 1.325v if it is stable.
     
  7. isaakk

    isaakk Member

    Joined:
    Oct 10, 2014
    Messages:
    124
    Location:
    7315 (Turners Beach)
    Yeap, definitely a typo/brainfart, not sure what I was smoking this morning. :rolleyes: Have updated with actual correct voltages.
     
  8. paulos

    paulos Member

    Joined:
    Dec 16, 2016
    Messages:
    22
    so far i am at 4.7Ghz on 1.310v
    passed real bench and 100%IBT and stable in games

    Yeah my cpu could be deceiving i have herd of skylake having a mixture of low Vcore cpu's that run very hot and run out of puff fast (possibly what i am experiencing with 4.8Ghz) and cpu's with a Higher stock Vcore but then more linear Vcore to Mhz increases.

    Isaakk what was your lowest Vcore for 4.6 or 4.7..? ( just to judge the Mhz and Vcore needed between clocks)

    Apparently the reason for hotter cores with any cpu but in this case for skylake it has to do with core's positions amongst other components and cores which from this article ( http://www.tomshardware.com/forum/id-1800828/intel-temperature-guide.html ) says a 10'c difference between cores is expected. it also predicted my hottest core. that being said i am interested in seeing how CLU performs compared to MX-4.

    albeit i understand now that Vcore is the amount we feed the cpu and VID is the amount the cpu wants to be fed. when i set in Bios Vcore to 1.325v the value in HWinfo64 says my max Vcore is 1.312
    now on 1.310v my Vcore max is still 1.312 ( the value does change for different Ghz) this is all with a reported VID of 1.210 for either voltage. Does this 1.312Voltage that HWinfo64 shows me, "suggest" this is my ideal Vcore setting or that this is the minimum Vcore amount i need for this OC?

    Questions
    -does having my OC in dynamic mode and auto mode increase Vcore on its own making me think i am passing low voltage high overclocks when in fact my Bios is simply upping Vcore to meet cpu needs ( this would suggest HWinfo64 is wrong or the value is relative to another baseline)
    -should i ever pay attention to VID for different overclocks?
    -to lower temp on a hot core should i clock cores individually to get a smooth temp among all cores instead of a constant clock?
    -for gaming is it worth individually clocking cores to have something similar to (5 ,4.7 ,4.5 ,4.5)Ghz
    -can you have good temps and good voltage and still not get a certain clock or is it simply you need more voltage but any speed can be run if the cooling is good enough..?
    -can you be stable on say 5Ghz at a high voltage but not 4.8 on any voltage..?


    Cheers Guys!
     
  9. paulos

    paulos Member

    Joined:
    Dec 16, 2016
    Messages:
    22
    Well CLU arrived today along with that i lapped the IHS with 1500/3000 wet n dry sandpaper on a cork block( my attempt to minimise uneven sanding )


    before
    4.7Ghz Vcore 1.310v with MX-4 for either side of IHS ( hyper 212x two fans)
    Realbench temps ( 70,77,70,70)
    IBT 100% temps (70,78,71,72)

    After
    4.7Ghz Vcore 1.310v CLU for the chip and mx-4 for heatsink ( hyper 212x two fans )
    Realbench temps (59,60,60,57)
    IBT 100% temps (58,58,59,55) with 55'c max for games

    hopefully in the next day or two i will receive the Cryorig R1! but in that time i guess i can bump up the Vcore and find a stable 4.8! very happy with the temps considering the OC and cooler
     
  10. maldotcom2

    maldotcom2 Member

    Joined:
    Feb 18, 2006
    Messages:
    1,838
    There is nothing to derive from VID which will help ascertain the ideal Vcore. From what I've seen VID is far higher than actual required voltage for stability. It's probably Intel's way of setting a voltage scale that works for all CPU's in the silicone lottery, which you seem to have won btw. Post some screenies of those OC results!

    Also have a look into the cache multiplier. IIRC default is 41 (4.1). Keeping that low can provide more head room for the core OC. Cache overclocking doesn't affect game performance in a notable way.

    Edit: Also don't take too much notice of software reported Vcore. It's never 100% accurate.
     
    Last edited: Dec 20, 2016
  11. paulos

    paulos Member

    Joined:
    Dec 16, 2016
    Messages:
    22
    Easy no more trying to figure out VID's role in OCing then, il be taking a look into other OCing methods for sure.
    cheers for clearing up those questions

    Realbench benchmark any others i should post up..?
    [​IMG]
    http://imgur.com/OKkL5kj

    Edit:
    its not looking good! tried 4.8 on 1.375v failed Realbench benchmark on the last test -.- otherwise was holding stable at 65'c max
     
    Last edited: Dec 20, 2016
  12. wraith.50

    wraith.50 Member

    Joined:
    Sep 25, 2016
    Messages:
    115
    Your temps are good considering you have an air cooler, but if you want to achieve the lowest temperatures and also lower your voltages, then water cooling is the only way to go.

    I would highly recommend that instead of getting another air cooler, that you purchase a watercooling kit from EK, that comes with everything you need to cool your cpu and then allows you to expand it later on, to add in your gpu as well, when you are ready.

    From what you have written, you are obviously out to achieve the highest clocks you can, at the lowest temperatures and voltages. An aircooler is just not going to get anywhere near a custom watercooling kit.

    The following kits will keep your delidded 6700K temps as low as possible. Obviously the bigger the radiator, the better cooling capability you will have, when adding in a gpu as well. Just depends on what size radiator will fit your case.

    https://www.ekwb.com/shop/ek-kit-p240

    https://www.ekwb.com/shop/ek-kit-p280

    https://www.ekwb.com/shop/ek-kit-p360


    Also you should look at getting Thermal Grizzly Kryonaut, as the TIM to use between the top of the IHS and your cpu block , as that is the best non conductive paste you can get and will help lower your temps even further.
     
    Last edited: Dec 20, 2016
  13. maldotcom2

    maldotcom2 Member

    Joined:
    Feb 18, 2006
    Messages:
    1,838
    Looking good. Just go ahead and try 4.8 at 1.4v, at least you'll know then if it's even on the table.

    Once you've settled on a setting it's also wise to run a test that will load the CPU for a long period of time, a few hours at least just to ensure you are stable. I agree with Wraith, that chip will be dynamite under water.
     
  14. wraith.50

    wraith.50 Member

    Joined:
    Sep 25, 2016
    Messages:
    115

    Just to give you an idea of how much difference watercooling makes to your temperatures, I ran Realbench benchmark at the same settings as you have above, so 4.7ghz overclock.

    My cpu is also delidded like yours and I have used CLU under the IHS and Kryonaut on top. My ambient temperature was 22 degrees, and can be verified by looking at the SYSTIN Temp in Hwmonitor, which is the temperature of my motherboard and that is at 27 degrees, which is 5 degrees above ambient, as the board is just aircooled.

    As you can see watercooling lowers the temps by another 16-18 degrees, compared to your air cooling results. This in turn allows further overclocking headroom.

    http://imgur.com/a/4BQAH
     
  15. maldotcom2

    maldotcom2 Member

    Joined:
    Feb 18, 2006
    Messages:
    1,838
    How did you manage a min of 7 degrees below ambient?

    Edit: and why is realbench reporting a frequency of 800 MHz?
     
    Last edited: Dec 21, 2016
  16. paulos

    paulos Member

    Joined:
    Dec 16, 2016
    Messages:
    22
    Yeah for the moment atleast air cooling is as far as i will take this, a mixture of lack of funds and the fact my Gtx1060 is now the bottle neck!
    But when i get a chance to upgrade GPU il be looking at water cooling for all!

    Thanks for the links/advise on what comes next Wraith.50

    Update 4.8Ghz stable at 1.4v
    passed Realbench at (63,64,63,60)'c still blown away by CLU's difference or i really did mount it horribly (applied to much TIM)
     
  17. maldotcom2

    maldotcom2 Member

    Joined:
    Feb 18, 2006
    Messages:
    1,838
    That's quite an amazing chip you've got there, to be at 4.8, 1.4v and still be sub 65C on a low end air cooler puts it in the top 1% of chips easily. I wouldn't even bother with watercooling, in fact i'm jealous. Mine (6700K, delid, CLU) does 4.6 at 1.39v and 80 degrees after 1 hour of video encode, and that's under water. Granted 30 degree ambient, but still...

    Makes me want to roll the dice and buy another one, surely I can't get a worse one. :paranoid:

    CLU is pure liquid metal, a Gallium alloy that is liquid at room temp. Thermal transfer characteristics are far superior to any other type of TIM.
     
  18. paulos

    paulos Member

    Joined:
    Dec 16, 2016
    Messages:
    22
    4.4 on 1.200v 51'c max
    4.5 on 1.230v 52'c max
    4.6 on 1.250v 57'c max
    4.7 on 1.310v 62'c max
    4.8 on 1.380v 66'c max
    4.9 on 1.450v 73'c max

    Testing method for todays speeds (6 hours of just watching temps and changing voltage)
    -Realbench handbrake test solo first
    -Realbench benchmark
    -IBT 100% 15mins
    -Realbench stress test 15mins ( this seems to get the cpu the hottest by 2-3'c)

    is it just voltage to Ghz that controls stability or is there a natural frequency of each and every chip to which the chip prefers certain speeds to certain voltages aka not stable at 4.8 on 1.350 but stable at 4.835 on 1.355.

    when im not busy on my computer im going to run the 4.8 and 4.9 speeds for much longer on Realbench stress test just to make sure it can hold up!
    Im thinking mabye there is a 5Ghz in there but at what voltage and temp i wonder! i may give it a crack if it can be had at 1.5v and under 82'c. not something i would actually use but for the "mum get the kodak" and can it be done on a hyper 212x would be cool to see.

    intel does list max voltage 1.52v am i ok in thinking i am safe to try 1.5v ??
    i wonder if my motherboard MSI pro carbon has enough power delivery to support such demand (8 power phase)...?

    haha yeah i have read a fair amount of people being stuck at 4.4-4.5 even at 1.45v. i would be happy with 4.6 i was hoping to get that, but at the same time i invoked "the power of now" and told myself i would get lucky and here we are, at the same time i also said my delidding would be perfect but two scratch's says otherwise so who knows what is at work!. i hear kaby lake is the cpu to get 4.8-5ghz as a standard OC potentially, although somehow locked to windows 10 and above.
     
  19. wraith.50

    wraith.50 Member

    Joined:
    Sep 25, 2016
    Messages:
    115
    My min temp is always below ambient, when I first start my pc and then after the loop's water temperature has equalized, which usually takes about 30-40 mins then my min/idle temps are close to my ambient temp.

    The frequency realbench shows at the top, is a snapshot of the system specs, that it gathers when you first start it up. Because I had my C-states on in the bios and my Windows power plan on balanced, the system is able to lower its frequency to 800mhz and it did so while realbench was collecting the specs and snapshot the cpu when the cores where at 800mhz.
     
  20. maldotcom2

    maldotcom2 Member

    Joined:
    Feb 18, 2006
    Messages:
    1,838
    It would be reckless of me to recommend going much above 1.4v, but you seem aware of the risks so I'm willing to watch. A lot of people will shoot for a max OC and then wind it back to a more sensible 24/7 OC. At a certain point the voltage and heat required for another 100Mhz isn't worth it. Every CPU has a "sweet spot" before requiring significantly more voltage to go faster.

    This has inspired me, I'm going to grab Realbench this arvo and see if it will pass any settings that I know are unstable on my 6700K.
     

Share This Page