Hi - I have a new PC build with Gigabyte X870E Aorus Elite Wifi 7 motherboard and a Ryzen 9950X3D CPU.
I am cooling the CPU with a Arctic Liquid Freezer III Pro 360 AiO cooler.
The installation of the Liquid Freezer III was a very painful process, and now I am not sure that I got it installed optimally. When my Win11 system is sitting idle, I see CPU temperatures around 48 DegC. With the Arctic Liquid Freezer fan speeds set at 45% minimum, I would think that I should be seeing CPU temperatures closer to the System case temperature of around 30 DegC. I don’t have anything overclocked, and I have the X3D performance option disabled in the BIOS.
Are my 9950X3D temperatures too high with this cooler?
My led status display shows me the Tctl , i suppose that could be changed in the uefi, never looked.
Your temps are similar to mine , i have a Artic Freezer 3 360mm .
Probably not. My 9950X under LF3 Pro generally idles around +25 °C ΔT. Since the IHS dominates the thermal resistance chain the fan and pump speeds don’t have much of an effect. The LF3’s mainly pump limited on the lower RPM end.
I made some modest improvements with a reapplication of Arctic MX-7 thermal compound. I also got a bit of a surprise.
By reapplying MX-7, I saw the CPU temperature initially drop down to 39-40 DegC when I was sitting in the BIOS. Previously, I had seen the BIOS temperature hovering around 47 DegC, so my expectations were high. And when I booted into Win11, my idle temperatures were now hanging around 43-44 DegC.
Unfortunately, my expectations were inflated and short-lived. The surprise waiting for me was this: After I ran a GPU / CPU Benchmark in 3DMark, my CPU temperatures have now raised about 3 DegC hotter than before the benchmark. This means that for some reason, running the benchmark degraded the MX-7 thermal compound application. Now, I never see the CPU temperature drop below 43 DegC in the BIOS.
However, I am seeing my CPU temperatures are still lower than before the MX-7 reapplication - just not as much. Now, I see my Win11 idle temperature hovering around 46 DegC, which is better - just not as much as I saw before the 3DMark benchmark.
I don’t know if you can make that claim without more testing.
It’s more likely pump hysteresis, cold soak affecting testing methodology or even non-ideal paste application mixed with hotspots causing the small delta.
It typically takes over 100C to degrade most thermal pastes.
As to your original idle temps, they seem reasonable to me. AMD’s non-advanced packing chiplet processors don’t really idle well, they pump out 3-5 times more heat than a desktop Intel processor would at idle and will commensurately have a higher temperature; that mixed with the extra thick heatspreader on top of the silicon that @lemma mentioned and you have a recipe for high idle temperatures.
Hi, @twin_savage - I understand your point, but the only thing that changed between the before/after readings was running the CPU benchmark test (approximately 1 minute in length) that took the CPU up to 75 DegC for the 1 minute of CPU stress testing.
When applying the MX-7, I used the recommended “X” compound application pattern, which did provide better cooling results at first. Nothing physically or mechanically changed in my setup before/after the benchmark test.
Both before and after the 3DMark test, I had the system down cold to room temperature (about 30 minutes) before reading the CPU readings in BIOS and in Windows, and there is now a definite permanent rise in temperatures after the 3DMark test. The screws on the Arctic LF3 Pro mounting rails and the cold plate are as far down as they can go.
Is there a recommended curing period for the MX-7 compound for it to become more stable? Do I need to reapply the MX-7 and wait a few days for the compound to become “set” so that I don’t see the temperature rise?
Yeah, booting in the 30s C is normal for me. If the system’s been on for a while it still takes a few minutes for much heat to get past the first fan on the rad.
Going by Igor’s data, MX-7’s among the most stable pastes available. Always possible you happened to get a bad batch but Arctic’s quality control’s pretty good and what’s been mentioned so far suggests poorly controlled measurements within the first time constant of an AIO’s 10+ minute time to steady state.
I haven’t had occasion to do an MX-7 CPU mount yet but with pastes I’ve monitored from three generations back there’s been some minor jumpiness in logged temperatures under load in the first 15 minutes or so. Haven’t seen that with previous gen pastes like GX-14 or two gens back like pre-enshittification TF8, but if the application’s not great it’s probably always a possibility. See also Igor’s article on how application methods influence paste lifespan.
Since paste application is a physical and mechanical change, which can easily move temperatures a couple degrees if application and mounting methodology is inconsistent, a common minimum standard for TIM testing is testing of three applications. Cycle time depends on test protocol, but usually a few hours if not days.
Thanks, @lemma - I did one more (my last, hopefully) reapplication of MX-7 for now with less compound applied in the X pattern. My CPU temp dropped down to 39 DegC in the BIOS, and 46 DegC (approx 17 DegC ΔT from case temp) when idle in Windows. I’m not going to run anymore benchmarks, and I’ll track where my CPU temps go from here.
I did have one question based on your previous comment about fan and pump speeds not having much effect: Are you saying that I can keep my pump speeds and fan speeds constant (say, 60% for all three), and I won’t see much difference in overall CPU cooling even under load? I did some quick tests, and I didn’t see much of an effect (about 1-2 DegC additional cooling) when running the two fans and pump at maximum while sitting in the BIOS, but that was under no loading. If I get minimal cooling benefits from running the fans at full speed under load, I would like to keep my system quieter and run everything at something like 60% or so.
