New custom loop, high temps.

Discussion in 'Extreme and Water Cooling' started by Dave, Mar 18, 2019.

  1. Dave

    Dave Member

    Joined:
    Jun 26, 2001
    Messages:
    294
    Location:
    Morayfield (4506)
    Just moved from a NH-D14 over to a custom loop, EK Coolstream PE 360 rad, with Noctua NF-F12 industrial (3000rpm) fans, D5 pwm pump, EK velocity CPU block, in a Define R6.
    Ryzen 2700 @3.8ghz 1.325v. Ambient temp 24c
    Idle CPU temp 33c, rad fans 600rpm, pump 4750rpm (pwm 100%)
    Prime95 (Small FFTs), Max CPU temp, 66c, rad fans 2000rpm, pump 4750rpm (pwm 100%)

    With the NH-D14 max CPU temp was 76c, I was expecting better out of this loop, have double checked the CPU mounting, heatsink paste (hydronaut) looks fine. Run it for a few hours, pretty sure I've got all the air out.

    Any ideas what to try next, or am I just expecting to much?

    Thanks.
     
  2. havabeer

    havabeer Member

    Joined:
    Dec 12, 2010
    Messages:
    4,556
    Isn't the boost clock of that cpu 4.1mhz?
     
  3. OP
    OP
    Dave

    Dave Member

    Joined:
    Jun 26, 2001
    Messages:
    294
    Location:
    Morayfield (4506)
    Yep, but I have that disabled.
     
  4. Ratzz

    Ratzz Member

    Joined:
    Mar 13, 2013
    Messages:
    8,052
    Location:
    Cheltenham East 3192
    I think you are expecting too much personally. You have a 10°C drop over a D14, one of the best coolers in the business. That's a pretty decent improvement. Water is good, but it isn't magic.
     
    money_killer likes this.
  5. self_slaughter

    self_slaughter Member

    Joined:
    Dec 12, 2010
    Messages:
    1,441
    Location:
    4165
    Yeah, I think you're expecting too much out of it. That's a good drop.

    When you start pushing high voltages etc. water shows its strengths as it doesn't get overwhelmed so easily.
    At low voltages, its good, but not gunna halve your temps or something compared to a monster like a d14.

    It is like magic on GPUs though, with those big dies in direct contact with the water block. oooh baby :weirdo:
     
    money_killer, melon and im late like this.
  6. Plonkflopped

    Plonkflopped Member

    Joined:
    Feb 3, 2012
    Messages:
    285
    Location:
    Brisbane
    Nvm. Read your post again and I'm wrong.
     
  7. OP
    OP
    Dave

    Dave Member

    Joined:
    Jun 26, 2001
    Messages:
    294
    Location:
    Morayfield (4506)
    Thanks for the replies everyone.
    Made a few changes, now running.
    Idle CPU temp 32c, rad fans 550rpm, pump 1250rpm (pwm 25%)
    Prime95 (Small FFTs), Max CPU temp, 66c, rad fans 1500rpm, pump 2600rpm (pwm 50%)
     
  8. m3k

    m3k Member

    Joined:
    Dec 29, 2007
    Messages:
    651
    add more RADDDD - i had a 1080rad on the back of my case once- my CPU never could get above ambient- those were the days.
     
    SomeBoredGuy and mad_mic3 like this.
  9. Kaine[zof]

    Kaine[zof] Member

    Joined:
    Mar 22, 2004
    Messages:
    2,178
    Location:
    Melbourne
    This is actually impossible.

    More rad has diminishing returns. Because thermodynamics. Unless you're running some crazy hot cpu/overclock, any more than 360 is going to see very little difference in temps as you will start to be limited by the ability of the block to conduct heat away from the CPU.
     
    Deeb0 likes this.
  10. melon

    melon Member

    Joined:
    Jul 7, 2015
    Messages:
    14
    Location:
    Brisbane
    You can add two 360s in the R6. The ryzen 8 core uses a bit of heat at load. My 1800x under water is around the same, funny I also expected a better improvement moving from a d14 lol. The bonus is that you can run the fans/pump low rpm.

    As self_slaughter said GPU's under water are awesome. Should give you much more headroom for OC
     
  11. Pazzu

    Pazzu Member

    Joined:
    Jul 16, 2015
    Messages:
    65
    Location:
    2602 Canberra
    There really isn't much you can do unless you want to start lapping. Even then you would only be seeing 1-2 C drops. Direct die would ultimately give you the best, albeit with extremely high risks.
     
  12. radeon_freak

    radeon_freak Member

    Joined:
    Sep 19, 2007
    Messages:
    3,152
    Location:
    Sydney
    Lapping days have come and gone unless you will be doing both the IHS and block.

    Majority (if not all) today's blocks are intentionally made with a slight convex shape on the heat block. Been this way for some time now.

    Haven't purchased a new block in some time but EK used to (don't think it's changed) ship theirs with different style and thickness jet plates to alter the convex shape. Aquacomputer used to use rubber O rings to achieve the same result and shipped with an alternate if you wanted more/less convex shape.
     
  13. m3k

    m3k Member

    Joined:
    Dec 29, 2007
    Messages:
    651

    it was an exaggeration but idle was room temp, and load was like 30ish and would only hit 50s when both GPUS were under load.
     
  14. OP
    OP
    Dave

    Dave Member

    Joined:
    Jun 26, 2001
    Messages:
    294
    Location:
    Morayfield (4506)
    I'm finding that is pretty common when it comes to water cooling posts. :)

    So I overclocked the 2700 a little bit more to 4.0ghz, which required 1.35v to be 100% stable, which of course raised the temps even more.

    Ambient temp 24c.
    Idle CPU temp 33.5c, rad fans 650rpm, pump 1750rpm (pwm 25%)
    Prime95 (Small FFTs), Max CPU temp, 74.5c, rad fans 1500rpm, pump 4750rpm (pwm 100%)

    I then added an XSPC EX240 Crossflow radiator to the loop, which lowered the temps a bit more.

    Ambient temp 24c.
    Idle CPU temp 32.5c, rad fans 625rpm, pump 1735rpm (pwm 25%)
    Prime95 (Small FFTs), Max CPU temp, 70.5c, rad fans 1500rpm, pump 4750rpm (pwm 100%)

    So from my findings, it seems a single EK PE360 radiator is not sufficient to remove all the heat from a Ryzen 2700 @ 4ghz 1.35v, even with Noctua Industrial fans running at 3000rpm.

    It' shouldn't be case cooling as the temps hardly change when opening the front door, or removing the side panel.
    It doesn't seem to be the CPU block, as adding the second radiator dropped the temp by 4c.
     
  15. Sunder

    Sunder Member

    Joined:
    Apr 26, 2012
    Messages:
    3,285
    Sorry foe the thread necro, but something doesn't seem right to me here.

    I have a EK Phoenix 360 radiator, with an overclocked 1080ti and an overclocked 8086k in the loop.

    An overclocked 1080ti should be pumping out 280w, and an overclocked 8086k adds another 150w.

    That's 430W. According to this table, I should have a delta T (of the coolant) of a hair over 10*C
    20190425_180004.jpg
    That gels with reality, as on a 25 degree day, I was stablising the CPU around low 50s with Furmark Stress Test and Prime 95 running at the samr time. (Can't remeber exacy settings)

    According to here:

    https://www.legitreviews.com/amd-ryzen-7-2700-8-core-65w-processor-review_205588/9

    Even going silly volts (not needed to hit 4.1ghz), they only used 303w for the entire platform, not just the CPU.

    In other words, thats less heat going in, but you have a higher temp, and adding more radiator helps.

    Is your pump actually performing okay? Is it cavitating or churning air bubbles?

    Your outcomes don't seem right to me.
     
    Last edited: Apr 25, 2019
  16. OP
    OP
    Dave

    Dave Member

    Joined:
    Jun 26, 2001
    Messages:
    294
    Location:
    Morayfield (4506)
    Pump seems fine, no air bubbles, moved a good volume of water when filling, emptied the reservoir in a few seconds. I don't actually have any way of measuring flow though.

    You sure about them figures?

    https://www.techpowerup.com/reviews/EKWB/EK-MLC_Phoenix_360/6.html

    With the EK Phoenix 360 radiator and only an 8700k, they where seeing stock (3.7ghz) temps of 54c and overclocked (4.8ghz) temps of 71c with Prime95, and that's with an ambient of 22c, yet your claiming low 50s, on a 8086k running Prime95 with a 1080ti also running Furmark at the same time, with an ambient of 25c.
     
  17. Sunder

    Sunder Member

    Joined:
    Apr 26, 2012
    Messages:
    3,285
    Shit. No, you're right. Just looked at the old screenshots. It was 78*C, 5.1Ghz, 1.24v

    Not sure where I got low 50s from, but that number was stuck in my head.
     
  18. OP
    OP
    Dave

    Dave Member

    Joined:
    Jun 26, 2001
    Messages:
    294
    Location:
    Morayfield (4506)
    I'm still waiting for a coolant temp sensor to arrive, seems to be lost in the mail, but I got around to adding thermistors to monitor inlet and exhaust air temps.
    I've calibrated the sensors, and are only seeing a difference of 2c between inlet and exhaust, after running Prime95 Small FFT's for 20 Mins.
     

    Attached Files:

  19. Sunder

    Sunder Member

    Joined:
    Apr 26, 2012
    Messages:
    3,285
    I am curious to see what this turns out to be. It's been many years since I've done custom loops, I'm totally reliant on sealed systems these days. I do recall there were all sorts of issues in designing from scratch. I recall one case, where a person on the forum used a huge aquarium pump (which most of us did those days), and after weeks of scratching their head and checking all manner of things, it was that it was pushing water too fast through the radiator, not giving it time to actually exchange heat. Switching to a smaller pump actually helped. Not suggesting it's your issue, since things are much more standardised these days, but I'm still curious. Hoping one day I'll have more time to build and maintain custom loops again.
     
  20. havabeer

    havabeer Member

    Joined:
    Dec 12, 2010
    Messages:
    4,556
    Not being able to grab/lose heat out of a block isn't a thing, I dunno where it came from but it's not true.

    In an open loop (total loss) maybe, not in a closed loop
     

Share This Page