Processing at high load levels, consistently between 90 and 100 percent.
Processing at high load levels, consistently between 90 and 100 percent.
I’ve got an Arctic Freezer II 240 running on an i9 9900k. It originally came with an Icy Diamond pad, but lately most cores are hitting mid-90s and 100+ on the fourth core. I tried reapplying the paste today, spreading it thinly over the IHS as best I could. It worked—thinner layers kept the credit card intact, or else it’d strip the coating. I also reset the BIOS to match the settings from that video. Now I have a stable 5.0GHz clock at 1.28V, but the temps stay in the same spots. Even then, they don’t drop much. Only when I cut the speed to 3.6GHz do I notice a slight improvement. The screws are tight but not overly so, avoiding any damage to the threads or IHS. I’m unsure what’s really going on. The i9 9900k can handle high loads and stays cool at those temps, but I prefer the quieter 80s setting I had before. Evaporation is possible, but I doubt it’ll happen this low unless it gets even colder. I don’t hear any noise, though the fans are definitely running. That’s confusing—since it never made a sound before but had good temps, how can I be sure it’s actually cooling? I do see air movement, but it doesn’t seem to pull hot air away from the radiator like it should. After letting it run for over 20 seconds, I noticed warm air being expelled from the radiator. It seems the liquid is helping, but I’m still puzzled. Is there a way to confirm the pump is working beyond just the sound? The header on my CPU_Fan connects to the CPU, and the case has a fan at the top with others pulling air through the radiator fins. The idle speed is around 35-45, mostly 42-43.
When the pump is operating and there are no blockages, the tubes or pipes should have similar temperatures. If you notice a cold area in the tubes or radiator, it indicates a blockage.
Just a few more details now. It's operating smoothly (60s or below) except for non-AVX tasks. Performance peaks on OCCT or Cinebench, which pushes it up to this level. Games function normally, so I'm fine right now. However, I'm wondering what's causing this spike. If the cooling system is effective and fans are managing heat, it likely means the paste is working well. I've tried a KPX earlier, but it won't arrive for a week and won't help much. Still, I'm stuck and need another idea. Thanks for the suggestions so far! I'm curious about what could be making AVX loads demand more power and raise temperatures beyond 110°C.
You've already explored several adjustments, but the results haven't met expectations. It seems there might be an issue with the setup or conditions—perhaps something changed in the system or environment. I'm here to help figure out what might be affecting performance at high temperatures.
Well, things took an unexpected turn. I entered XTU and managed to hit a voltage of 1.200 in one attempt, but lowering it further triggered a BSOD. Once Windows stabilized, it didn’t hold steady near that point. Eventually I settled back up to 1.280, where stability returned. Switching my Cashe down to 42 didn’t improve the situation, so I adjusted it back to 45. I then experimented with different offsets, curious if pushing it further down to 3.6Ghz with minimal temperature changes would yield similar results—or something new would emerge. Interestingly, around an AVX offset of 9 I noticed temperatures dropping below 100°C again. Raising it to 10 kept me around 96°C or less. It seems an offset of 9 might be the sweet spot, balancing under 100°C while still supporting at least 4.0GHz for AVX instructions, a difference of about 100MHz. Currently it appears this setting is the optimal range.
For now, it seems the Arctic Freezer II (Model 1) might be struggling to maintain performance over time. I observed temperatures cooling rapidly after testing, suggesting the cooling system is handling the load well. Running fans and an AIO at full capacity still managed to keep things under control.
It’s puzzling, though—previously, even at full load, temperatures stayed high for days. This time, after a short burst, they dropped significantly under AVX load. The dust issue was likely minimal this cycle, possibly just a temporary accumulation. Still, it’s strange that the cooling didn’t improve much despite the fan being off. More to think about.
Here’s a revised version of your text:
To wrap up this discussion: I realized it all... Wow, I’m such a... Linus. ... Actually, I didn’t drop anything except my temperatures (*burn!*). However, the problem appears to be with my IHS and the thin layer of thermal paste I applied last time. This time, I opted for the Kingpin KPX—finally managed to get it after waiting a week from an Amazon seller with high ratings for speed. I also spread it more evenly this time and added a small extra on the center after ensuring the nuts were properly seated. Once everything was secured, I checked the BIOS and confirmed I was running version 5.0 with no AVX offset. I saved and restarted, then ran HWMonitor and Cinebench r23. The results were outstanding: top scores and lowest temperatures during load tests, both single and multi-threaded. I even managed to lower my maximum temperature to 82°C on the second try! Although I’m unsure what caused the thermal pad to fail over time, it seemed to work well initially. After uninstalling and reinstalling, I noticed the cold plate had almost no contact—so poor performance. When I tried pasting this system after a long wait, it barely connected, making me question how I could run without it overheating. Thankfully, after repasting with KPX instead of AS5, I achieved the coldest temps ever and discovered my IHS is concave. My past layer of paste was fine at first but degraded over time. Now I have a fresh coat, and my cooler is functioning perfectly. Great job everyone who offered advice—it was all about a simple repaste. Also, I forgot to adjust the voltage setting; it was at 1.280 instead of my usual 1.320. That wasn’t ideal, but it still worked. I’m grateful for the guidance, even if it was as straightforward as this fix. Thanks again!