Thermal throttling occurs suddenly at 9900K. Here’s a possible solution idea.
Thermal throttling occurs suddenly at 9900K. Here’s a possible solution idea.
Hello, let's discuss this issue. I'm running a 9900K processor with a 5GHz allcore chipset on manual VCore at 1.35V LLC, using a Cinebench R20 test. It's installed on an Asus Maximus XI Extreme with two separate water loops—one for the CPU and one for the GPU. Each loop has two radiators: a Nemesis Blackice GTX 560 and an EK D5 240mm RGB combo. The CPU uses an AC XPX Acryl waterblock, while the GPU uses an EKWB 1080Ti with BP. I've upgraded to a 6900XT with an AC GPX waterblock and switched to push-pull fans (Noctua AF 140) on all radiators.
Despite these changes, my thermal throttling remains severe, even at stock settings like 4.7GHz allcore with 1.25V. I've tried reseating the waterblocks, changing pressure, and adjusting the PCB bend near the die. Nothing seems to help. My suspicion is that the slight PCB flex might affect the indium solder contact, making it harder for the CPU to maintain good thermal performance.
I'm considering options like desoldering, using liquid metal to clean the die, or reapplying a new copper IHS. Another thought is whether heating the CPU slightly (to around 220°C) could help the indium solder regain better contact. What would you recommend? Thanks for your time and support!
Verify Vccio and System Agent levels. On auto mode this board generates around 1.45v, just above the minimum cooling limit. Keep each below 1.2v unless your memory is extremely high.
Vccio and vccsa work well at 2133mhz frequencies (around 0.9 and 1.05 MHz). I've already checked this. Even with the CPU running, I set VCCS and VCCIO manually to 1.2 and adjusted VCore to 1.25. The processor reaches about 90°C. I don't believe the CPU is failing immediately; it seems to heat up gradually. Replacing the GPU didn't improve things significantly in just ten minutes. Thanks for the guidance. Should I consider heating the CPU further so Indium melts? Could that help? Or is it similar to saying once good contact is lost, you can't restore it? I'm unsure, but my core temperatures differ by about 20°C during load. Some cores spike to 90°C while others drop to around 70°C. The CB20 score is near 4500, not 4800. At 5GHz I usually see core temps between 4950–5000.
Ok, I wasn’t sure about the amount of testing done. But with Maximus X and some OC work, I didn’t check IO and SA and saw really high settings. So the solder stayed solid. The LM handles heat almost like the solder does, so it makes no sense to worry unless you're really into it. How’s the pump speed? Is it running too slow?
To be clear, I’m not just looking for a quick fix and better temperatures. The issue is that I believe my CPU PCB might have bent slightly. Specifically, the die appears to be bending a bit outward in the socket direction. When you hold the CPU with the IHS at the bottom, it doesn’t look like a flat surface but rather a small curve around the die. It makes me wonder if the die has loosened from the solder joints. Could that happen? Is this even possible? I’ve noticed about 20°C higher temperatures now compared to before. That can’t be normal, right? So my thought is clear: there are only two possibilities. Either I heat up the CPU (around 160°C) and hope the indium solder becomes fluid enough to restore good contact and bring my temps back, or I could delude it and use liquid metal to remove the indium. I’ve heard that works, though it takes longer than using quicksilver. So essentially, what I’m considering isn’t pointless—it would only make sense if my temperatures are still within range. I just want to know what others think. I can’t tolerate this. 90°C on stock with this cooling isn’t acceptable at all.
The pump is operating at full speed. I notice the fluid moving quickly through the acrylic. No other changes as mentioned before. As you said earlier, Vcore and vccsa should be around 1.45 when using auto with XMP. With a RAM speed of 2133MHz, it would sit at 0.950 and 1.050. I read that indium solder melts near 156°C. That’s why I’m wondering—what if the CPU temperature reaches 160°C? Would the indium melt completely? Could it just become fluid but still stay between the IHS and die? Might it improve contact if the situation isn’t optimal? If the contact isn’t good enough, could heating it near the melting point help? Or would other components fail first, especially at close to 160°C? Since indium is a liquid metal, it’s possible it would behave similarly. That’s something I should consider.
You'd be better off to RMA the chip then to delid it. Intel cpu therm trip is 110-115c depending on the board. 100c I throttle temp. If you're not hitting 100c, it's not throttling. 90c is not a surprise temp on a chip capable of over 300w with all core OC. Most chips do look slightly bent. This is not alarming, I've seen this on most chips. As long as the pins touch the pads and it posts up, it's a non issue. Everything you see Der8auer do, I've already have done. In fact the thin razor he uses was an idea a young man had years before a die mate was invented...... which is doubtful the guy will ever get credit for, but can be found in a delid tutorial from early 2009 at EOCF.
You're right, the oven is no longer safe for cooking. It poses a serious health risk for the next few decades. Wishing you the best and hoping you don't have to hear this again...
Thanks for the alert from @Mark Kaine, though I don't actually use my oven much. Even so, I baked a GPU inside one a few years back. @ShrimpBrime if RMA was available, I would have tried that. Thanks for the details on the chips—most of them look slightly bent, but it's not a big deal as long as the pins stay in contact and they heat up. I didn't realize that. But I think my indium layer might be damaged; it probably got broken somewhere. My CPU starts throttling around 90°C, though it's only slight. After trying a 160° setting, it made things worse, so I gave up. I cleaned everything up carefully—even reapplying heat-resistant silicone. Now I have a 5GHz allcore, 1.36V LLC6. With 1.2VCCS and VCCI around 80°C in Cinebench R20 (208W under load). My "CORE 2" runs a bit hotter, just 2-4°C, but it's manageable. I might consider getting a new 12900K with a better motherboard if I want to improve longevity. But honestly, I already noticed the indium wasn't making good contact anymore. It was clear where it failed. Sorry, I didn’t take photos, but on one edge of the die it also seemed disconnected. Personally, it was worth it. I’ll probably inspect the IHS more closely or replace the copper part. The room temperature is 32°C at idle and 33°C under load—so it’s hard to keep the heat away from the die quickly. If you have any more questions, I’ll help. For now, I know my CPU is running a bit hotter than usual, but cleaning and reapplying the IHS with LM seems to work so far.