Announcement

Collapse
No announcement yet.

Maximizing the heat dump of radiators with slow pump speeds

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Maximizing the heat dump of radiators with slow pump speeds

    Wasn't sure where to park this thread, but I'm curious to hear what others think of what I've learned in the last couple days as I benchmark my new system.

    Hardware details up front: not-yet-overlocked 9990K processor and a 2080Ti, D5 pump feeding the loop that has 2x560mm rads with 8x140mm fans (push only).

    Inline sensors: 2 water temps, one before the rads and one after, later installed 4 more thermometer based ones to capture exhaust air temps at mid points on the rads' exhaust paths. These are all tied into an Aquaero 6 who's aquasuite software has been instrumental in testing the factors below.

    I was expecting to see the water temp drop a couple degrees between the two water sensors but no, its much much greater. Initially I had a 6c delta under load between the two but over time I tweaked the pump speed down and kept the fans running as close to 400-800rpm as I could while aiming to keep the post-rad sensor's readings as close to ambient air temp as possible. I've aimed to keep the pump at its slowest setting of 800rpm as I adjusted the slopes. The four air temp sensors let me see how much more room I've got by showing me four data points of temperature as the water passes through the radiators and if the last one is a very small difference that gave me confidence to dial things back even more.

    I got an 18c delta by the end of this process. Water coming in to the radiators at ~41-43c and leaving at ~23-25c. Wow. Just wow. The GPU gets to about 56c under these conditions and the CPU hovers around 65-70c and the whole thing is so quiet. Its just weird to feel the temp difference between the tube heading to the CPU vs the one coming out of the GPU.

    I've always seen people talking about medium/high pump speeds to keep component temps down but far as I'm concerned both of the components are well within their tolerance. What I am contemplating though is replacing the tube between the GPU and the rads, its PETG and my understanding is around 60c would get to where it'd slowly deform. Changing it to even just Acrylic would give a lot more head room in case something goes side ways.

    Anyone else play around with their settings with this as a goal over component temps? I was not expecting these results at all. Really cool stuff.

  • #2
    You shouldn't concentrate on how well the liquid is cooled before or after going through the radiators. CPU and GPU temperatures are the only metrics that really matter here. i doubt there would be a substantial, if any difference between the CPU or GPU temperature with maximum or minimum pump speeds. In fact, higher pump speeds should reduce CPU/GPU temperatures in an ideal setup.

    Remember that by lowering the pump speed, liquid also has more time to dwell inside the CPU/GPU block, reducing cooling efficiency.

    Comment


    • #3
      Yes I understand that Cyclops, if I raise the pump speed even to 1200rpm I can drop the GPU down to the mid 40s, but at 56c its not in any dangerous position. Meanwhile that's also slowly heating up the whole loop, I'm sure there's an upper threshold where it'd level off.

      This is really more of an experimental thing, this could run perpetually in theory because the water as a whole isn't slowly heating up and the GPU/CPU are still well within their comfort zones. When I get to OC'ing the GPU it will change the values but my estimations based on the air exhaust temps I have more head room to achieve further delta and maintain that ambient output temp goal.

      Comment


      • #4
        What are you trying to achieve? The heat dumped by your radiators will equal the heat added to your loop by the waterblocks. Newton's law of cooling gives you a linear relation between the air-to-radiator temperature delta and the radiator heat output. Therefore, halving your flow rate at the same fan speed should roughly double the temperature delta (approximating the radiator temp as equal to the water temp). Cyclops points out the reason for a custom loop is to reduce temps of the hardware, so most people want that temperature delta as low as possible. Increasing it does reduce pump noise, but you haven't mentioned that as a problem.

        Comment


        • #5
          Right, getting lower component temps is definitely the goal of any water loop. Now picture you've achieved that to a satisfactory level, the next trait that's important is noise factor of the system, that's what I'm gunning for now. With the active cooling components running at such low RPM its very quiet, so very quiet. Since its also expelling all the heat via the rads its not slowly heating up the reservoir and re-introducing that heat energy back to the components via warmer water going through them.

          As a tweak I raised the pump to 2600RPM and am running with just the CPU under load, it holds the same temps as running at 800RPM (65-70). It hasn't been a big contributor compared to the GPU in any of this any ways. I imagine though if the GPU were put under load as well I'd see them both rise beyond that as the water meant for cooling them shows up warmer then via the other setup.

          Caveat: I'm relying on the folding at home client for load testing and they've had trouble pushing work out to all the new clients they've gotten so I'm kind of at their mercy for testing unless I pick a different suite.

          Originally posted by Sir Meowric View Post
          The heat dumped by your radiators will equal the heat added to your loop by the waterblocks
          No, it will *try* to achieve equality but if there isn't enough surface vs air available it will come up short and that heat will stay in the water until the next attempt, but that's compounding your held energy in that water over time. Right?

          Comment


          • #6
            I'm sorry guys, I've been thinking about your responses and I think some further clarification of the layout and goals are would help: The case is distinctly divided into two portions, the front for the MB and components, reservoir etc and no active cooling. Then the back half where the radiators and fans live (its a TT 900 tower). The goal is to transfer as much heat towards the back as possible without re-introducing it to the front. In that regard the aim is to dump all the heat out the back so that the reservoir and front-half loop are as ambient temperature as possible to avoid heating up the the front half via the reservoir etc acting as passive heat sinks.

            Comment


            • #7
              I mean, the reservoir and the front of your case won't care if it's running 20 degrees hotter or not. It's not like they're gonna explode or go into a runaway thermal reaction.

              Comment


              • #8
                Originally posted by Cyclops View Post
                I mean, the reservoir and the front of your case won't care if it's running 20 degrees hotter or not. It's not like they're gonna explode or go into a runaway thermal reaction.
                No no of course not Tell you what though I'll look for another window where both the CPU and GPU are getting simultaneously capped out via the FoH and jump up the pump speed to like 75% to see how it all plays out.

                In retrospect though yes the sound of the pump is a significant factor that I'm aiming to minimize too.

                Comment


                • #9
                  I got an opportunity this morning to run the pump at the 2400RPM under full load and didn't like where it was going: The liquid entered the rads at ~33c and exited at 31c which in turn meant that warmer water was going through the CPU block and it was getting into the 80c range, the GPU was happier of course, sitting around 48c. This was with keeping the fans under the same curve of hitting ~800rpm but with that lower temperature target. (I have the curve pinned against that intake water sensor's values)

                  So I notched it down to a 1200RPM pump speed and this is looking great. The rad sensors are showing 34-35c intake with 26c output, the CPU raised a hair to floating around ~67-72c but the GPU really benefited and stayed around 45c. I think the trade off of a slight hit to the CPU to give the GPU even more room is worth keeping. Realistically in my day to day activities I doubt I'll ever see that CPU come close to maxing out, its the GPU that gets all the work in most games any ways.

                  I also looked into better understanding why I was seeing the radiator exhaust temperatures do a non-linear reduction as the water passed through and this makes sense: Of course the radiators have an easier time dumping heat that's much hotter then the ambient air, going from 40->32 is easier then going 32->24. Its the same in reverse where the lower your coolant is the easier it is for it to absorb heat off of a hot component.

                  Comment


                  • #10
                    I'm in this for the art
                    Last edited by bungwirez; 04-08-2020, 09:02 AM.
                    HAF932 Mods
                    C70 Mods

                    Comment


                    • #11
                      Its been almost two months now and I figure I should come back and 'close the loop' on this little adventures in slow flow rate experimentation.

                      It was fun but there was a lingering problem where because the liquid was spending so much time in the GPU block it was heating up to the point where it'd create air bubbles, a very minor boiling effect. You know like when you boil water on the stove how you see little bubbles form long before the whole thing reaches full boil? Picture that over the course of many hours but the water flow is too slow to push those bubbles further along. These bubbles were building up into sizeable air pocket and I'd have to dial up the pump speed to force the bubbles through the system.

                      In the end I raised my pump to 1500rpm as a baseline, the increased water flow reduced the water's temperature to top out around ~34-35c going into the two rads and exiting around 28-29c with the fans still running nice and slow at 700-800rpm under load. The delta starts off at about 6.2c but as the water over all gets warmer that delta settles to more around 5c. GPU & CPU temps really didn't budge much vs the slower pump speeds, but no more bubbles forming inside the GPU block.

                      This has made me think about switching up the next system's loop order so I have the reservoir/pump between the heat load(s) and the radiator(s) along with tubing that would encourage the bubbles to go towards it.

                      Oh and I replaced my EK Vardar 140mm fan array because they were grindy and wouldn't smoothly transition through a curve, the random sounds they'd make was downright irritating.

                      Fortunately Noctua fans tucked away in the back half of the case don't matter for their looks in this case, they are performing like real champs:
                      Click image for larger version

Name:	2020-04-07 NewFans.jpg
Views:	132
Size:	60.0 KB
ID:	90912

                      Comment

                      Working...
                      X