Water cooling loop issue.

Discussion in 'Water Cooling' started by vegeta535, Jun 18, 2018.

  1. vegeta535

    vegeta535 2[H]4U

    Messages:
    2,229
    Joined:
    Jul 19, 2013
    So I just put together my first water loop with a 7820x @4.8ghz running 1.25v with a 1080ti. I have 2 280 30mm thick in the loop. When I run prime95 temperature hold around 70c then suddenly shoot up to 100c and starts throttling after 5 or so minutes. What would cause this? The 1080ti stays around 32c which is about 2c more then idle. CPU sits around 30c at idle.
     
    Last edited: Jun 18, 2018
  2. VanGoghComplex

    VanGoghComplex [H]ard|Gawd

    Messages:
    1,269
    Joined:
    Apr 5, 2016
    70c seems pretty hot to begin with for a chip running at 1.25v. I'm going to guess a bad CPU block mount, but that's taking a shot in the dark.

    What version of Prime95 are you running? I personally don't use Prime95, but I've read that after a certain version, it uses AVX instructions, which causes most CPUs to furnace.

    Can you provide pictures or a better description of your loop? What pump are you using and how are you powering it? What CPU block are you using? What kind of case are you using?
     
  3. Zarathustra[H]

    Zarathustra[H] Official Forum Curmudgeon

    Messages:
    25,279
    Joined:
    Oct 29, 2000
    That would be my guess too, but its just based on the fact that block mounts are a common issue. Also agreed, we could use more details.

    One thing I would check is to see if your radiators are actually seeing the heat from the CPU. You could to this with a fancy thermometer, but even just keeping your hand on one during test checking to see if it feels any warmer when the CPU gets toasty could give you some more hints as to what could be going on.

    Poor CPU block contact is the obvious first thing to check, but you could also have some sort of water flow issue. If the radiators are getting hot, maybe there isn't enough airflow across them? Check how your fans are installed. Are you doing push-pull? If so, did you accidentally install the fans so they are fighting eachother, rather than helping eachother? When doing a large complex build its the stupid oversights you miss when you are rushing to get things together that usually come back and bite you.
     
  4. NoOther

    NoOther [H]ardness Supreme

    Messages:
    6,668
    Joined:
    May 14, 2008
    The 1080ti is on the same loop and it maxes at 32C? Could be a combination of things, but like others, I would try remounting the CPU block. The other culprit would maybe be a flow issue, but if your 1080ti is on the same loop and is staying around 32C, then it seems to point to the CPU mount. Could not be enough contact on the block, which means too much thermal paste, not screwed down tight enough, or the surfaces are not flat enough to conduct heat properly between them.
     
  5. vegeta535

    vegeta535 2[H]4U

    Messages:
    2,229
    Joined:
    Jul 19, 2013
    The 1080 ti sit at 32c idel when stressing the CPU. It goes to 60ish when stressing it. Yes the radiators get extremely hot when the cpu shoots to 100c. I am running a EK-XRES 140 DDC 3.2 PWM Elite. Flow is pump/res>1080ti>cpu>rad1>rad2>pump/res. I must of derp the way I ran the tubes or the block has a bad mount. Flow rate looks good through the loop. I'll message with it after work today. Was getting late when I ran into this problem while stress testing last night.
     
  6. VanGoghComplex

    VanGoghComplex [H]ard|Gawd

    Messages:
    1,269
    Joined:
    Apr 5, 2016
    What kind of 1080ti and block are you using?

    I ask because my 1080ti never breaks 35c under load, and I've got an otherwise comparable cooling system in terms of rad space... I'm beginning to get the feeling you may be having a whole-loop problem instead of just a CPU problem.
     
  7. vegeta535

    vegeta535 2[H]4U

    Messages:
    2,229
    Joined:
    Jul 19, 2013
    Ek-1080gtx ti
     
  8. VanGoghComplex

    VanGoghComplex [H]ard|Gawd

    Messages:
    1,269
    Joined:
    Apr 5, 2016
    Okay. Your EK PWM pump - where do you have the PWM wire plugged in? My second guess is a low flow issue.

    I ask because if you're controlling your pump off a motherboard header, and that header is a fan header set on a "silent" fan curve or something similar, you may be just barely spinning your pump.

    If you're using a full cover waterblock, you should be getting much better cooling than 60c... Heck, my 1080ti barely broke 60c back when it was air cooled.
     
    Zarathustra[H] likes this.
  9. NoOther

    NoOther [H]ardness Supreme

    Messages:
    6,668
    Joined:
    May 14, 2008
    If it was me, I would probably swap the order and have the loop go to the CPU first. But I don't think that is the issue.


    If this was a whole loop issue, the GPU would be running much hotter as well. It still sounds like a mount issue for the block on the CPU. If it was a whole loop issue, then the culprit would most likely be the radiators. Not sure why you have 2, and if you have 2 it might work better to put one them between the GPU and the CPU.
     
  10. Zarathustra[H]

    Zarathustra[H] Official Forum Curmudgeon

    Messages:
    25,279
    Joined:
    Oct 29, 2000
    I doubt it. Order is mostly irrelevant if you have sufficient flow. The only time order would become a problem is if you have a low enough flow that you see significant differences in water temp in different spots of the loop.
     
  11. VanGoghComplex

    VanGoghComplex [H]ard|Gawd

    Messages:
    1,269
    Joined:
    Apr 5, 2016
    I only have my 1080ti to compare to, but 60° is 'much hotter' than I would expect from a properly functioning loop with a full-cover GPU block.

    Mine tops out at 35°, and I'm only working with 420mm of rad with a CPU in the loop as well.
     
  12. Zarathustra[H]

    Zarathustra[H] Official Forum Curmudgeon

    Messages:
    25,279
    Joined:
    Oct 29, 2000
    Possibly. For whatever reason though, GPU blocks tend to have much lower temps than CPU blocks do, so I could imagine a situation where very low flow is causing an issue, where the GPU stays in a OK operational temp, but the CPU does not.

    His GPU temps ARE very high. My GPU with an EK fullcover block idles only very slightly above room temp with fans tuned down. With fans on full blast even full load the GPU is in the low 30's. If it is going up to 60C there is definitely something going on with the GPU as well, which suggests whole loop issue.
     
  13. NoOther

    NoOther [H]ardness Supreme

    Messages:
    6,668
    Joined:
    May 14, 2008
    I guess I am confused then, because his very first statement says it stayed around 32C. I thought when he was talking about stressing it the CPU first goes to 60C, then tops out at 100C while the GPU stays at 32C, which would coincide with his first comment.

    vegeta535 Can you gives us a better indication of which components are at which temperatures and when they get there?

    IE:

    IDLE:

    GPU: 30C
    CPU: 32C

    Start Prime95:

    GPU: 32C
    CPU: 60C

    Prime95 Full Stress:

    GPU: 32C
    CPU: 100C

    or whatever the actual numbers are.
     
  14. VanGoghComplex

    VanGoghComplex [H]ard|Gawd

    Messages:
    1,269
    Joined:
    Apr 5, 2016
    Ah, I see the confusion. I read a GPU idle temp of 32° and a GPU loaded temp of 60°.
     
  15. NoOther

    NoOther [H]ardness Supreme

    Messages:
    6,668
    Joined:
    May 14, 2008
    Exactly, if both his GPU and CPU are spiking, then I would say it's a radiator and/or flow issue. It would be better to use 1 larger radiator than 2 radiators. The 2 radiators could be killing the flow. Could also be air pocket issues, but with a reservoir I find that hard to believe as long as the reservoir is still filled at least halfway after priming the loop.
     
  16. VanGoghComplex

    VanGoghComplex [H]ard|Gawd

    Messages:
    1,269
    Joined:
    Apr 5, 2016
    Hm... I could see the dual rads killing flow if he had a crappy (or underspun) pump, but with a DDC he should be fine with two. My old build ran a 140 and a 280 rad with a detuned "silence optimized" Alphacool DDC, meaning it was considerably slower than normal DDC pumps, and flow was no issue there.
     
  17. NoOther

    NoOther [H]ardness Supreme

    Messages:
    6,668
    Joined:
    May 14, 2008
    Yes, but I have no information on the radiator. If it is a good radiator, then there shouldn't be an issue. If it is 2 shitty radiators, there could be a big issue. Honestly, a single good radiator should be able to handle both. It also depends on what his answer to the temps are. If the GPU is not spiking, then I still think its a CPU block mount issue. Either way the most important information is exactly what the temps are for the components at the respective times/situations with Prime95.
     
  18. vegeta535

    vegeta535 2[H]4U

    Messages:
    2,229
    Joined:
    Jul 19, 2013
    I don't believe it is a flow issue. Running the pump at full speed barely changes anything. I am using 2 280 black ice nemesis radiators with a EK res with a DCC pump attached. Using a heatkiller 3 and a full cover ek-gtx 1080ti with compression fittings on everything. Everything brought new. 2 stock fractal r6 fans set up as intake pulling through the front radiator. Top radiator has 2 noctua fan set up as exhausting out through the radiator. Pump is connect to the pump header on the MB and all fans connected to the R6 fan hub which is connected to cpu header.

    7820x running at 4.8ghz 1.25
    32gb Gskill trident @3600 cas 15
    MSI tomahawk x299 MB

    Loop order is.
    Res/pump>rad1>GPU>cpu>rad2>res

    Idel temps
    CPU high 30s on all cores
    GPU 38

    Stress CPU
    Almost all cores hit 100 instantly after remounting
    GPU stays around 38

    Mounting looked good with a even spread of tim after removing the block.
     

    Attached Files:

  19. vegeta535

    vegeta535 2[H]4U

    Messages:
    2,229
    Joined:
    Jul 19, 2013
    They got worse after remounting. Temps go too 100c instantly now on most cores. Using mx-4 tim. I just put a dab in the middle. It spread pretty evenly to the edges.
     
  20. pendragon1

    pendragon1 [H]ardForum Junkie

    Messages:
    11,568
    Joined:
    Oct 7, 2000
    are you sure you've got water flow?
     
  21. vegeta535

    vegeta535 2[H]4U

    Messages:
    2,229
    Joined:
    Jul 19, 2013
    Yes 100% sure I got flow. I can see it moving from the silver additive I added.
     
    pendragon1 likes this.
  22. pendragon1

    pendragon1 [H]ardForum Junkie

    Messages:
    11,568
    Joined:
    Oct 7, 2000
    how did you bleed the loop? is there a bleed valve on the top rad? what happens when you stress the gpu? ^gpu wouldnt cause that kind of spike without it experiencing one.
     
  23. vegeta535

    vegeta535 2[H]4U

    Messages:
    2,229
    Joined:
    Jul 19, 2013
    The res is slightly higher then my radiator and using the fill port to bleed. I have been getting a little air out after tilting the case around. GPU maxs out at 41c under stress and CPU is around 61c running furmark. So I guess it is air/mount issue with CPU then. Ambient temperature is around 24c.
     
  24. pendragon1

    pendragon1 [H]ardForum Junkie

    Messages:
    11,568
    Joined:
    Oct 7, 2000
    ok cant tell from the pics. you could try putting the case on its side and get the res out and up as high as you can to help bleed it. it looks like you have the tubing length to do it
     
    Billy the Dimwit likes this.
  25. vegeta535

    vegeta535 2[H]4U

    Messages:
    2,229
    Joined:
    Jul 19, 2013
    So I have my computer on the side right now and hear a lot of gurgling sounds. I should assume that means there is air in the system. It doesn't make that sound standing up right. I'll try pendragon1 suggested. Block mount should be solid now. It is not shooting too 100c right off the back now but it will still reach it after 5 mins.
     
  26. pendragon1

    pendragon1 [H]ardForum Junkie

    Messages:
    11,568
    Joined:
    Oct 7, 2000
    airlock in the top rad I would assume...
     
  27. pendragon1

    pendragon1 [H]ardForum Junkie

    Messages:
    11,568
    Joined:
    Oct 7, 2000
    looks like he has enough slack to do it
     
  28. pendragon1

    pendragon1 [H]ardForum Junkie

    Messages:
    11,568
    Joined:
    Oct 7, 2000
    the res still needs to be upright and highest point to bleed properly. or he could bring it lower and open a bleed valve on the top rad if it has one.
     
  29. Zarathustra[H]

    Zarathustra[H] Official Forum Curmudgeon

    Messages:
    25,279
    Joined:
    Oct 29, 2000

    You need to do full case gymnastics. Don't just lay it on its side and leave it. Have it running and essentially tip it side to side back and forth,, jiggle, roll, lean, repeatedly for as long as it takes until all air is out of the loop. This is typically not a quick process, especially if you aren't using a surfactant in your fluid.

    I wouldn't do this. You want to make sure the pump has contact with fluid at all times or you may do damage with it. Rock it side to side and back and forth, and even put it on its side, but never upside down.

    What kind of dye is that? It almost looks like handsoap? Unless you are planning on just running straight distilled water with a few small additives, it is better to not DIY your fluid. Either use plain distilled water with some algicide and a mild surfactant, or use a dedicated marketed water cooling fluid. Don't try to DIY fluid color. It usually ends in tears.
     
    pendragon1 and Billy the Dimwit like this.
  30. pendragon1

    pendragon1 [H]ardForum Junkie

    Messages:
    11,568
    Joined:
    Oct 7, 2000
    good point

    what silver additive?

    look closer, it looks like its full of shimmery hand soap.
     
  31. vegeta535

    vegeta535 2[H]4U

    Messages:
    2,229
    Joined:
    Jul 19, 2013
    It is mayhem silver color additive. Added to distilled water.
     
    Last edited: Jun 18, 2018
  32. pendragon1

    pendragon1 [H]ardForum Junkie

    Messages:
    11,568
    Joined:
    Oct 7, 2000
    that stuff will gunk up, shouldnt be this quick though. did you test it with plain distilled before adding it? you might want to drain it and flush it and try just distilled.
     
    vegeta535 likes this.
  33. vegeta535

    vegeta535 2[H]4U

    Messages:
    2,229
    Joined:
    Jul 19, 2013
    Temps are getting better but still shoots after a few minutes.
     
  34. pendragon1

    pendragon1 [H]ardForum Junkie

    Messages:
    11,568
    Joined:
    Oct 7, 2000
    move/shake it around and flick the block. can you see that stuff swirling in the gpu block?
     
    vegeta535 and Billy the Dimwit like this.
  35. vegeta535

    vegeta535 2[H]4U

    Messages:
    2,229
    Joined:
    Jul 19, 2013
    I just replaced the with plain distilled water and temps are already better. Almost all cores are sub 70c now
     
  36. pendragon1

    pendragon1 [H]ardForum Junkie

    Messages:
    11,568
    Joined:
    Oct 7, 2000
    ok that's better. that stuff may have gunked up your cpu block. its a pita but i'd drain it and check that block.
     
  37. vegeta535

    vegeta535 2[H]4U

    Messages:
    2,229
    Joined:
    Jul 19, 2013
    Meh still shoots up to 100c after 5 mins. I'm done for the night. Got to go to bed. Ill disassemble the loop and block tomorrow. Thanks for the help today.
     
    pendragon1 likes this.
  38. pendragon1

    pendragon1 [H]ardForum Junkie

    Messages:
    11,568
    Joined:
    Oct 7, 2000
  39. WATERCOOL-Jakob

    WATERCOOL-Jakob Watercool Community Manager

    Messages:
    17
    Joined:
    Nov 2, 2017
    For your original problem, there seem to be several reasons that could potentially be at work here:
    - You chose a very old block. The Heatkiller3.0 was released in 2008. While it's still a potent piece of hardware, combining it with Intel's latest gen Extreme edition seems a bit harsh :D
    - You mounted the block the wrong way round. We recommend using the central port as inlet, and the UPPER port as outlet. The HK3.0 had a dead end on the other side of the out-port. I assume that you have two serious air bubbles on that end. Try flipping the block around so that the outlet is ABOVE your inlet, this should seriously improve your thermals.
    - If I understand you correctly, you are using Mayhems Aurora. This fluid comes with multiple warnings that it's a show fluid and not intended for daily use, or for inexperienced customers. This fluid uses small floating particles that are well known to clog every microstructure. The HK3.0 used not only 0.3mm wide microchannels, but also a complex jetplate design (CHeck out the pics in this review from back 2009). I would assume that both structures are already pretty heavily clogged.

    In conclusion, here is what I would suggest: open and clean the block thoroughly, mount it the other way round, and use distilled water with a corrosion inhibitor as fluid. These three simple steps should drastically improve your thermals.


    (And, although an EK block is none of my business: I would recommend plumbing that the other way round, too. Should also0 improve both thermals and flow.)
     
  40. pendragon1

    pendragon1 [H]ardForum Junkie

    Messages:
    11,568
    Joined:
    Oct 7, 2000
    good catch and advice!
     
    vegeta535 likes this.