New custom loop handles high GPU and water temperatures effectively
New custom loop handles high GPU and water temperatures effectively
I'm really pleased you enjoy the build. Take your time with it, and appreciate how it turned out.
The temperature sensor is situated at the spare intake port, on the back of the GPU. The vertical fans are for intake (there are vent holes on the back cover), while the top fans are for exhaust. The pump is already operating at full capacity, though it didn’t make much of a difference compared to 45%.
I wasn’t aware that the CPU block orientation affected performance either. The main differences between cores are quite small, so this shouldn’t be my biggest worry.
However, I tried a few things:
- No air bubbles were visible in the GPU block, but some appeared around the edges of the CPU block. I attempted to move them, but they remained stuck. There were no unusual noises from the pump.
- Removing a cap from the reservoir didn’t change the flow.
On the positive side:
- After giving it another good shake, flipping it over, and gently tapping the radiators, I noticed some air was present in the top radiator. As I cleared the air, I observed improved airflow around the loop (shown by the bubbles).
- After removing excess air from the system, I ran a GPU stress test again (no time for CPU today).
First test without any panels:
GPU: 47 | Liquid: 40 | Ambient: 27.5 | Fans: 1000 RPM (around 75%)
Second test with all panels on:
GPU: 50 | Liquid: 43.6 | Ambient: 27.5 | Fans: 1200 RPM
The data shows that adding panels improved the water temperature by about 4°C.
I’ll give it another good shake and turnaround tomorrow to see if more air can be extracted, but I’m unlikely to find any left. Still worth trying.
Thanks for all the helpful advice so far!
The 4C difference may stem from case airflow or insufficient ambient air entering through the chassis when panels are closed, leading to slightly higher core temperatures. This is unexpected given the nearly equal temperature changes I observed. Positive news is that your flow rates seem normal, which helps address this issue. Air bubbles might still be present but could dissipate more easily if you adjust the case angle while running—make sure to top off the reservoir first to avoid air re-entering. Watch for a drop in reservoir level; this indicates air is gathering in the loop rather than elsewhere. Just uncap and refill to push out the trapped air and stop it from returning.
I recently attempted to rotate the system in all directions, upside down and face down, while gently shaking it. I managed to capture only a very small amount of air from the radiators. There was no noticeable movement of air either. It seems there is almost no air left inside the radiators. What I do see are some residual air in the CPU, which appears reluctant to move, and a few tiny bubbles near the reservoir edges. The issue persists. After about 10 minutes on the bench, the water temperature reached 43.2°C and continued to increase gradually once the test stopped. Both radiators were warming up evenly to the touch.
I'm still following this discussion. The mystery remains. The radiator capacity is sufficient so neither should become overly warm as noted. Could the pump be unusually hot for some reason? It's unclear where this extra heat might be originating. Could you share your idle water temperatures and what GPU temperatures are seeing during current tests?
Also, how are CPU and GPU core temperatures being displayed? I'm also interested in what those numbers represent. Based on the room temperature, you're experiencing a significantly larger temperature difference compared to someone with more cooling capacity.
with the rads you possess, the temperature above 40C is quite high for just a room temperature of 27C.
the GPU temperature is acceptable.
i compare it to what i experience with a 33% reduction in rad surface (2x240), a higher ambient around 33C, and slower fans up to 1350RPM (assuming the same notctua PPC), though using a less power-hungry card—heavily overclocked and power-modded—so it consumes slightly less power.
still, i rarely notice the coolant reaching even 40C.
this represents roughly a 9-10C difference (ambient plus coolant).
a few degrees might be due to sensor variations, but not up to 10.
i think there could be a problem with the flow in the loop; it may not be related to air trapped inside.
the exact model of the temperature sensor would help—some are too long to fit on the GPU terminal, and the sensor depth including threads should be under 5mm.
by the way, i suggest improving the vertical rad exhaust as well—it would make performance more consistent with "removed panels," since the top rad wouldn’t reuse hot air from the vertical rad.
Also, how are CPU and GPU core temperatures being displayed? I'm interested in understanding what those numbers represent.
I'm not sure what you're referring to here. The values I'm seeing come from the HW monitor. My components aren't getting too hot. However, I can easily notice they warm up a bit when under load. The fans I'm using are the one linked at overclockers.co.uk, with a maximum speed of 1500 RPM. When my water reaches 44°C, they spin around 1200 RPM.
I'm not overly concerned about GPU temperatures at 55°C. In fact, I expected them to be higher, around the high 40s. If they could maintain lower fan speeds, that would be better. But I might switch them back to the side fans and run only the exhaust fans. If that had happened initially, I'd try it again.
The temperature plug I'm using is connected to my Corsair Commander Pro. I might consider getting another temperature sensor and placing it in the reservoir to check for different readings. I also posted this on Reddit, and some people were confused because they thought my system was above ambient when I turned it on. To clarify, when I power up the PC, the sensor shows water temperature matching the ambient temperature. After about 10 minutes of sitting on the desktop, the water temperature reaches around 34°C.
Is there any way to test the airflow through the loop? Several people have suggested this so far. I could just let some air circulate while recording a video for you. Would you be able to judge whether it looks good or not?
I don’t understand why the pump feels unusually hot. It’s unclear where this extra warmth is originating. Could you share your idle water temperatures? Also, what GPU temperatures are we seeing during current testing?
The pump itself isn’t generating much heat. I only notice slight vibrations, which is typical. From the moment I start the system, the water temperature matches the ambient reading (about 29°C). After ten minutes, it rises to 35°C, with fans running at a slow speed of 40%. The idle GPU temperature today is 31°C, which is cooler than the water—this seems unusual.
Once I launch Heaven, the GPU temperature jumps to 47°C. From that point, the GPU temperature gradually increases alongside the water temperature. After ten minutes in Heaven (as she mentioned), the GPU is at 55°C, water at 41°C, and fans spinning at 1200 RPM. This is happening while the panels are off today.
I believed you had observed varying load temperatures after a few tweaks. From your liquid temperatures, it seems you're experiencing a greater temperature difference than I anticipated for a system with the amount of radiator space you're seeing.