Odd Heat Issue on Ivy-Bridge
Odd Heat Issue on Ivy-Bridge
My i5-3570k isn't performing well in the silicon lottery, but my temperatures have been inconsistent lately and I'm struggling to understand why.
Short summary:
I run a 4.2 overclock on this model and it's been stable for years. Recently, though, the temps have spiked significantly. I've changed thermal paste multiple times without any improvement. I've used both generic and premium options, tried different cooling solutions, and even swapped out my cooler. Despite these efforts, temperatures in OCCT Linpack (with AVX disabled) now consistently reach the mid-80s, while they were in the low to mid 70s before. I also noticed that when I accidentally misaligned the cooler, the issue became apparent. The thermal paste application and die/lid adjustments didn't resolve the problem. Ambient temperatures are usually between 16-20 this season.
Let me know if this makes sense or if you have more details!
Is the overclocking performed manually or by choosing a percentage overclock in the BIOS or through a GUI tool?
I manually adjusted it in the BIOS with a +.005v offset and +.012v turbo, LLC 2 (according to what I remember), plus some minor voltage adjustments. The setup has remained stable, maintaining good temperatures since I built the machine in early 2013. The base clock stays at 100MHz, the multiplier is now at 42, and the speed step remains enabled. Windows power settings are set to high performance, but the minimum processor state is at 5%.
I haven’t made any changes to the overclock before this problem occurred, so it seems the issue developed on its own.
OK, good. (on the manual overclocking part)
1 - The H80 (and most 120mm radiators) usually can't handle enough heat for a moderate to high CPU overclock. Given that a standard (good copper or bronze) 120mm radiator only manages about 150w, I'd doubt an aluminum one could manage more than 125-130w. Stock TDP on your CPU is 77w at full load; when you increase the speed, you need to estimate the new TDP using stock and OC speeds along with vCore. I know some might disagree, but I’d question what temperature ranges are considered acceptable depending on how much heat each cooler actually removes. I generally aim for a 5-10C difference between coolant and air temperature, not just the reported CPU temps.
2 - It seems either the pump isn’t working or there’s an airlock in the cooler. Can you take the radiator off the chassis and spin it while the pump is running?
3 - How is the cooler positioned? On top? Behind?
OK, good. (regarding the manual overclocking section)
1 - The H80 (and most 120mm radiators) usually don’t handle enough heat for a moderate to high CPU overclock. Given that a standard (good copper or bronze) 120mm radiator only manages about 150w of heat, I’d doubt an aluminum unit can manage more than 125-130w. The stock TDP on your CPU is 77w at full load; after overclocking, you need to recalculate the new TDP using stock values, OC speeds, and vCore. I know some might disagree, but I’d question what ‘good enough’ temperatures are based purely on the cooling improvement each cooler offers for each individual. I generally aim for a 5-10°C difference, meaning the difference between coolant temperature and ambient air.
2 - It seems either the pump isn’t running or there’s an airlock in the cooler. Can you detach the radiator from the chassis and spin it while the pump is active?
3 - What’s the mounting position? On top? Behind?
Try positioning the tubes at the bottom if possible. This could be the main problem—airlock.
These coolers often aren't fully filled with coolant, leaving an air space. Even though the pump is running, it mainly just circulates the same coolant without progressing it.
Adjust the position so the tubes entering and exiting the radiator are at the bottom, ideally fixing the airlock issue. Sometimes these coolers aren't fully filled with coolant, leaving an air space. If the pump can't push the air—especially since it's low power—it might prevent coolant from flowing freely. The pump keeps running but just circulates the same coolant without progressing. I'm concerned that if the air pocket stays too long, it could damage the pump bearings.
Adjust the position so the tubes entering and exiting the radiator are at the bottom. This could be the main problem—airlock. Sometimes these coolers aren’t fully filled with coolant, leaving an air pocket. If the pump can’t push the air (which is compressible) because of low power, it might stop coolant from flowing freely. The pump is still running but just keeps recirculating the same coolant without moving it. This was one of the first signs I noticed when taking it out of the box—it wasn’t fully filled, and shaking the unit made it sound correct. I’ll try rotating it to see if that helps.
Mrfrizzy:
My i5-3570k isn't winning in the silicon lottery, but my temperatures are off lately and I'm stuck trying to understand why.
Short summary:
I'm running a 4.2 overclock on my i5-3570k. It's been stable for years, but now temps are sky-high. I've changed thermal paste six times without any improvement. I've used both generic Coolermaster and Coollaboratory Liquid Copper. I also tried a Hyper 212 Evo, but it stayed stable with temps around 70-75 in Prime95 (older version) and 60s in games. Now, with OCCT Linpack (without AVX), temps jump to the mid-80s and rise further. My CPU is overheating and I've reapplied paste every time I disassembled it. Ambient temps are usually 16-20 this season.
Longer explanation:
I've had a 4.2 overclock on a Hyper 212 Evo for some time. After re-lapping the IHS and removing it, temperatures were low to mid-70s during stress tests and mid-60s in games (back then it was cooler). Now the first core runs about 10-20 degrees cooler than the others, which are all in the low to mid-80s when gaming. This shift happened after I bumped my cooler and misaligned it by 10-20 degrees. Whether that caused the issue or just made me notice it isn't clear is uncertain, but it might help with diagnosis.
Originally, I used Gelid GC-Extreme, which worked well until it aged and temps rose, so I switched to generic CoolerMaster paste. That lowered temps but not enough. After trying generic paste again, I changed my SLI 570 for the 980ti and got Coollaboratory Liquid Copper to see if it helped. I then disassembled the system, reapplied paste to IHS and die, and even after that, temperatures stayed unchanged three times. My 212 Evo fell off the desk, heat pipes broke beyond repair, so I switched to an H80i GT. It still didn't fix things—temps stayed the same or got worse. My idle temps dropped with the H80i GT compared to the 212 Evo's mid-30s range. The motherboard temp sensors are usually around 30s, but it's unclear what's really happening.
I've also experimented with overclocking at 4.4V without changing voltage, and during tests, three cores hit 90°C+ while the first stayed around 70°C. This suggests the die isn't dissipating heat properly, possibly due to poor IHS contact or inadequate thermal paste. The cooler itself doesn't seem to affect temps under load, so that's unlikely. Ambient temps are typically 16-20 this time of year.
Does this look like an issue on my end or has my 3570k degraded? Or is there another factor at play?
Thanks a bunch!
System Specs:
(2) 140mm front intake fans (NZXT FZ)
(1) 140mm side intake fan (Thermalright TY-147)
(1) 140mm top exhaust fan (Thermalright TY-147)
H80i GT Push/Pull for rear exhaust fans (SP120 Performance Edition, not stock SP120L)
i5-3570k @4.2 1.24v (+.005 offset, +.012 turbo, LLC2, minor misc volt tweaks)
ASRock z77 Extreme 6
16GB (4GBx4) Crucial Ballistix Sport @1600 8-9-8-18 1.65v
Gigabyte G1 Gaming 980ti @1506/7900
Enermax Galaxy DXX 1000w PSU
Kingston HyperX 3K SSD
Windows 10 Pro x64
UPDATE:
It seems my problem is worsening. Running Prime95 26.6 smallFFT and saturating the watercooling leads to max temps of 91-96-95-91 per RealTemp 3.7. OCCT Linpack results in 87-94-94-91. At 35-40% load, I'm seeing 60-70°C. Ambient temps are around 16-18. What's going on!?
I also reverted to stock clocks and voltage settings (1.144v in CPU-Z) in BIOS, but still get 66-70-70-66 in Prime95 26.6 per RealTemp 3.7. That's the stock cooler range.
Also, the radiator airflow isn't warm—just slightly cool to the touch after saturation.
UPDATE 2:
It turns out my MB reports VCORE lower than it should be. When I set 1.240v, I'm actually measuring 1.28-1.29v on the capacitors at the back. I've adjusted my overclock settings from +.012v turbo, +.005v offset, LLC2 to +.004v turbo, -.060v offset, LLC1, and PLLv 1.732v, now getting around 1.23v at the back. It's stable in OCCT Linpack and standard modes so far. I'll keep tweaking voltages as long as temps stay stable, but they've dropped to mid-80s to low 90s.
I shared your long version because it contains key details.
You mentioned TIM on the IHS and die during remount, which is excessive since minimal paste is usually enough. Too much thermal compound acts insulating rather than conducting. I didn't even know about Cool Laboratory Liquid Copper until now, which makes me wonder why more research hasn't focused on Silver—despite its high conductivity.
Have you thought about a direct die mount? CompuTronix does this with the 4770K and has tested nearly all TIMs, finally settling on Cool Laboratory Liquid Pro, which worked great for him.
Caution: Liquid Pro or Ultra can damage aluminum. Did you remove adhesive from under the IHS as others did to avoid thermal pumping?
Why would you overclock with offset voltage if you could use manual voltage control instead?
I shared your extended version since it contains crucial details. You mention placing TIM on the IHS when remounting, which seems excessive because minimal TIM is usually sufficient; excess thermal compound can actually act as an insulator rather than a conductor. I wasn’t familiar with Cool Laboratory Liquid Copper until now, and this raises questions about why research hasn’t focused more on Silver, despite its superior thermal conductivity. Have you thought about using a direct die mount? CompuTronix has done this with the 4770K and tested nearly every TIM, ultimately choosing Cool Laboratory Liquid Pro, which performed exceptionally well for him. Caution: Liquid Pro or Ultra can damage aluminum. Did you remove the adhesive beneath the IHS, as many have faced thermal pumping after removing their CPUs? If you’ve invested so much effort to remove your CPU’s lid, why are you still overclocking with offset voltage instead of applying full manual control? Your approach might be misinterpreted. I applied paste directly onto the die, placed the IHS, and then added more paste on top as usual. Over time, I noticed less paste was squeezed out to the side during each attempt. Technically, diamond offers better thermal conductivity, but it’s not an element I’d consider. I’m hesitant about a direct die mount since I haven’t done it before, though I think it could lower my temperatures somewhat. My main concern is why my temps have spiked suddenly after years of stable performance with identical settings. I avoided Pro or Ultra because of the aluminum issue and the difficulty in cleaning/reapplying them later. Yes, the adhesive was removed during de-lid and lapping. The process wasn’t too difficult, and I enjoyed the work. At that time, I had a custom watercooling loop targeting 4.5GHz, but safe voltages/currents limited my progress. I rely on offset voltage to maximize speed stepping range instead of locking at a constant voltage with 1.2x at idle. No matter the method, I’ve maintained identical overclock settings for years, yet my temps have become unacceptable. If something unusual has occurred with my i5 and I need to replace it, I will. Otherwise, if another factor is at play, these are the points I’m seeking advice on.