One core abnormally hot (3770k on water)

Discussion in 'Die-hard Overclocking & Case Modifications' started by MM10X, May 24, 2017.

  1. MM10X

    MM10X Guest

    Messages:
    4,240
    Likes Received:
    1
    GPU:
    3080 FE
    Hello,

    I just installed an EK P240 DIY kit on my de-lidded 3770k.
    The supremacy evo CPU block was disassembled and I had installed the suggested water jet by the EK manual.
    The radiator is installed into the side panel with the two Vardar 120mm fans mounted externally pushing air into the case.
    Everything is leak inspected, bled, and free of kinks or other obvious issues.
    This isn't my first water cooling setup. :)

    At first it was performing great, initial testing at stock clock speed showed around 60c load vs 80c+ with my hyper 212 Evo.

    After some tuning and overclocking I arrived at 4.6Ghz, 1.308vcore with load temps around 72c max in prime 95.


    However a few days later, core #1 started loading closer to 90c.
    I pulled the water block, removed the CPU lid, cleaned both surfaced, lapped the IHS, and reinstalled everything with new thermal paste.

    Load temps were back down to around 70c(ambient temp varies a lot here) and the difference between the coldest core (#3) to the hottest core (#1) was about 5c.
    Played some videogames (CSGO) especially on hot days and max load was around 79*c after a few hours of playing and the radiator nicely heat soaked.


    Come home from work last night after a hot day so room temps are pretty high, ambient temps here vary from 20c in the morning to 35c mid day.
    Core #1 is peaking at 90c on the desktop running just browser and chat clients! Varying 5~15% load.
    Ran P95 and MAX temps reported by HWmonitor showed the following:
    vcore 1.308
    core #0 81*C
    core #1 94*C
    core #3 78*C
    core #3 71*C
    Package 93*C
    :eek:c:
    I saw as high as 99*C peaks after playing around with some stuff on the desktop and turned off the computer to inspect everything and lower the overclock for now. There are no obvious mechanical issues with the loop. Everything is as it was when I installed it.
    I tried unplugging the PWM on the water pump to run at full speed and it made no difference on load temps.


    I woke up this morning and re-ran the test with lower ambient temperature and clocks dropped to 4.5Ghz @ 1.284vcore.
    core #0 72*C
    core #1 85*C
    core #3 71*C
    core #3 61*C
    Package 85*C

    At idle or normal desktop usage, all the cores are at similar temps. The only significant difference is at full load.

    Anybody have an idea of why core #1 is so much hotter?
    The rest of the core temps seem to be loading consistently, but #1 seems to peak much higher a few days after re-seating the CPU & thermal paste.


    The setup.
    [​IMG]

    delidded 3770k, after lapping. My goal was to get it flat, not mirror finish.
    [​IMG]

    Airflow with the side panel on, should be adequate.
    [​IMG]


    TL;DR de-lidded & lapped 3770k on water, one core is abnormally hot after a few days of usage with the other cores loading normally. Temps dropped to normal for a few days after re-seating block and CPU, but now they're back up.


    Thanks for reading.

    Also, its good to be back on water. Thanks to HH. :cool:



    edit, I tried running coretemp and it agrees with hwmonitor.

    I'm reading through information on the sensor test, but it still doesn't make sense why one core is so much hotter than the others.

    Core movement test has strange results.

    core #0 29
    core #1 45
    core #2 36
    core #3 32



    edit,

    I re-seated the IHS & water block again and temps are back to normal with max 5*c difference from coolest to hottest core.

    I guess it is an issue with the mounting.

    There seems to be a bit of junk in the lines already maybe from the radiator, so I'm going to flush everything out and hopefully shouldn't have any more issues.

    Still strange to me I've never had this kind of inconsistent issue before.
     
    Last edited: May 24, 2017
  2. Catroqui

    Catroqui Guest

    Messages:
    743
    Likes Received:
    9
    GPU:
    RTX 3080/1080 Ti
    Glad you apparently solved the problem.

    Anyways, what version of Prime95 are you using?
     
  3. Agent-A01

    Agent-A01 Ancient Guru

    Messages:
    11,628
    Likes Received:
    1,119
    GPU:
    4090 FE H20
    What did you use for die?
     
  4. MM10X

    MM10X Guest

    Messages:
    4,240
    Likes Received:
    1
    GPU:
    3080 FE
    Thanks.


    It seems to be "solved" But I still don't understand why the issue reoccurred after the first time I re-seated the cooler.
    Maybe I used too much thermal paste after lapping the IHS?
    But why did the issue take several days to manifest?

    I'll have to continue testing for a few more days and see if the issue comes back.

    Stock clocks load temps were just under 60*c, at 4.6Ghz w/ 1.278vcore I didn't see higher than 69*c with maximum of 5c variation from coolest to hottest core, ambient temp was about 22*c.
    These are the same results I got when I first installed the cooling system.

    I'm using the latest version of P95 from gimps, 28.10 x64.
    The max temps I provided are observed after a few minutes of the blend test.

    I'm sorry I'm not sure what you mean?

    Thermal paste? I was using the 'thermal grizzly' stuff that came with the EK kit.
    I am currently using the Antec formula 7 stuff.
     
    Last edited: May 25, 2017

  5. Agent-A01

    Agent-A01 Ancient Guru

    Messages:
    11,628
    Likes Received:
    1,119
    GPU:
    4090 FE H20
    When you delidded you used thermal grizzly on the CPU die?

    Regardless, liquid metal TIM is a much better option for that.
     
  6. MM10X

    MM10X Guest

    Messages:
    4,240
    Likes Received:
    1
    GPU:
    3080 FE
    Yes, I just use whatever I have laying around. I realize this isn't ideal.

    Maybe I'll go direct die mount with the EK block. it isn't too hard to setup properly.

    Results show NT-H1 within 1*c of liquid ultra.
    I'm using Antec formula 7 so should be similar to IC diamond.
    The compound difference doesn't account for individual core inconsistency.

    https://forums.anandtech.com/thread...d-3770k-an-h100-and-9-different-tims.2285595/
     
  7. Agent-A01

    Agent-A01 Ancient Guru

    Messages:
    11,628
    Likes Received:
    1,119
    GPU:
    4090 FE H20
    He did a very poor application of liquid metal.

    Results of proper application can achieve 5-10c better than the best tim.
     
  8. MM10X

    MM10X Guest

    Messages:
    4,240
    Likes Received:
    1
    GPU:
    3080 FE
    Well, after a few days of testing it seems the last re-seat & new thermal paste took care of it.

    After a few hours of gaming I don't see temps over 75c on a hot day overclocked.

    I can't explain the previous results, but there is no longer an issue.
     
  9. MM10X

    MM10X Guest

    Messages:
    4,240
    Likes Received:
    1
    GPU:
    3080 FE
    A little update on this,

    From your suggestion, I just applied Thermal Grizzly Conductonaut on the CPU die to IHS, I am happy to report excellent results!
    Load temps dropped significantly immediately after application.

    This 3770k is clocked at 4.6Ghz with 1.296 volts.
    However I did notice for some reason my bios is dropping to 1.284 volts during the Small FFT test.
    AC unit was reporting ambient room temperatures between 25c and 26c during these tests.

    Prime95 SmallFFT Preset went from a maximum of 98*c down to a maximum of 65*c with liquid metal.
    Prime95 Blend preset went from a maximum of 80*c to a maximum of 59*c with liquid metal.



    This is the amount I used on the die with about the same amount on the IHS itself, thoroughly spread about with the supplied cotton swabs.
    [​IMG]


    These are the results before the liquid metal, with thermal paste Antec Nano Diamond Formula 7 on both the IHS & CPU water block.

    Antec Nano Diamond Prime95 SmallFFT preset, 25c ambient.
    [​IMG]3770k_TIM_P95smallFFT_25c


    Antec Nano Diamond Prim95 Blend preset, 25c ambient.
    [​IMG]3770k_TIM_p95blend_25c



    After applying Conductonaut to the IHS, with the same Antec Nano Diamond applied to the CPU water block.

    Conductonaut Prime95 SmallFFT preset, 26c ambient.
    [​IMG]3770k_LM_P95smallFFT_26c


    Conductonaut Prime95 Blend preset, 26c ambient.
    [​IMG]3770k_LM_p95blend_26c


    I'm very happy with these results, it seems for CPU die to IHS application liquid metal is the ideal compound.

    The reason I was using thermal paste on my IHS is because I had lots of it around, and thought it would have less impact than the de-lid itself. It seems I was mistaken!

    Based on my personal results and others I've seen around the net, I highly suggest anyone planning to de-lid orders some liquid metal thermal compound.


    Edit,

    Now running nicely @ 4.8Ghz with 1.344 volts.
    Maximum load temp 74*c with 26*c ambient while stress testing.
     
    Last edited: Jul 9, 2017
  10. INSTG8R

    INSTG8R Guest

    Messages:
    1,659
    Likes Received:
    95
    GPU:
    Nitro+ 5700XT
    FWIW there’s always gonna be on hotter core but you’ve pretty much fixed the real issue.
     

Share This Page