Overheating i7-4790k

[Met]

Hello all. Recently my Windows 10 system has been having some stability issues, which I think I've pinned on the CPU overheating. When running particularly CPU heavy games, I get sudden shutdowns, and was seeing temperatures of over 90C through MSI afterburner. I was using an Asetek 570LX liquid cooler at the time, and the thermal paste hadn't been replaced in the 5 years since it was built, so I reapplied to no difference. I then switched to a Coolermaster LC240E, completely replaced the thermal paste again, and I'm having the exact same readings.

Just doing a full scan with Windows Defender is enough to overheat the chip to the point of shutdown. Idle temps are around 33C, the beginning of the scan when only one core is being used ramps it up to around 55-57C, and within a literal second of all cores being engaged, the temperature skyrockets up to 90+C and the computer shuts down. If I wasn't staring at the temperatures... or blinked... I would have had no idea what happened.

As far as I can tell, both coolers are functioning properly, and that there's no issue with the thermal paste (it covers the entire chip's surface, and the sink is pressed as closely to the chip's surface as the mounting brackets would allow.) The only other oddity I've been noticing is that the BIOS seems to give me the Welcome/Setup screen every time I go into it where it hadn't before. It still remembers settings I've changed, like my ram speeds, though.

Unfortunately, with my limited knowledge, I'm left out of troubleshooting ideas to test, and I wondered if anyone else could have any ideas as to what's wrong, or just further testing I could do.

Much appreciated.

 

[pgdeaner]

My first thought would have been the same in that there must be an issue with the cooler. Beyond that, the fact that your bios is acting strangely would concern me. Does the bios lose settings or just send you to the welcome/setup screen? Is it related to the system resetting from heat?

 

[Zerileous]

I'm wondering also if it's a BIOS or board problem and it's misreading the sensor causing the shutdown. First thing I would do is reset the CMOS (not just factory defaults option in the BIOS itself). Figure out of you can make it stable at all without any sort of overclocking or XMP first, then work from there. Any way to test the CPU in another system? If it's not a motherboard issue / and the reported temperatures are real, then it sounds almost like a problem with the TIM under the IHS.

 

[pgdeaner]

I'm wondering also if it's a BIOS or board problem and it's misreading the sensor causing the shutdown. First thing I would do is reset the CMOS (not just factory defaults option in the BIOS itself). Figure out of you can make it stable at all without any sort of overclocking or XMP first, then work from there. Any way to test the CPU in another system? If it's not a motherboard issue / and the reported temperatures are real, then it sounds almost like a problem with the TIM under the IHS.

I'm in agreement with you. This sounds like it could be a motherboard issue. Never had an issue personally that lead me to need a delid but I guess it does happen.

 

[Met]

Alright, this is a bit of a doozy. I reset the cmos and chose "load optimized defaults and boot". It then gave me this error:

https://imgur.com/a/SmKFTR3

I pressed enter, and it began booting into windows. I immediately entered my login info, and it shut down trying to load into desktop.

It then rebooted, and I waited on the screen before login for a minute. I was then able to log in, it loaded my desktop, and I opened afterburner to see my temperature at 60C. Then the chrome tabs I had open at the time of shutdown reopened, which immediately caused my temperature to shoot to 100C before another shutdown. On its subsequent automatic reboot, it then gave me that same error as shown in the previous link.

Next I logged in, I was barely able to kill chrome before shutdown, but not before reaching 96C in half a second. Woof. Now I'm idling at 35C on the desktop, seemingly stable so long as I do nothing.

Additionally, I experienced a shutdown when I first tried to boot it, after having it unplugged for around 20 minutes resetting the cmos. It also occurs to me that this same situation has happened at least twice before, once being right after I swapped coolers.

Ah, I also forgot to mention that I unfortunately don't have another computer available to test my cpu in.

I'm able to open webpages one at a time, spiking readings to 60C each time I load a page.

 

[trents]

Met, we'd prefer you attach images directly to you posts using the tools in Go Advanced. That way they will not disappear over time and it also makes one less step to do for those helping you.

 

[Met]

Apologies, I inserted it.

 

[freeagent]

Sounds like your cpu is getting a buttload of vcore. Maybe pull the cmos battery after you unplug the rig and go for a snack. When you come back plug the battery back in and plug in the psu hit go and see what happens.

 

[Met]

Already reset the cmos that way. I left it out for around 20-30 minutes.

 

[Johan45]

If you can enter the BIOS and stay there without it overheating, I would try flashing an updated version. But you don't want it shutting down during the update process.

 

[Mandrake4565]

What about manually setting the Cpu voltage to something like 1.15-1.2 v and see how it reacts like that?

 

[Met]

Leaving Vcore on Auto, using CPU-Z, I saw my core voltage bounce between around .9 and 1.38, mostly staying around 1V. Idle temperatures were about 30-35C. Setting the Vcore to 1.2 now has me idling about 20 degrees higher, at 50-55C. Opening a blank Chrome tab shoots the temp up to 62C, as it had before.

 

[Mandrake4565]

What happens under load? When left on Auto 1.38 V seems a bit excessive. I have a 4770k with a full custom loop and at 1.23 V it will push my temps under load above 90c. You can also try and lower the set voltage and see if it remains stable.

 

[Met]

Well, chrome running as a background process was periodically pushing the temperature up to 100C, so I don't think it'll take much more load than that. Previously, though, a shutdown would happen within two minutes of starting a Defender scan. At between 40-50% utilization is when I hit 100+C.

When you say lower the set voltage, is that different than setting the Vcore voltage?

 

[Met]

Delidded the CPU and found the thermal paste underneath to be intact, but flakey like dried paint. It came off very easily, and in chunks. I replaced it with a thin layer of Corsair TM30 non-conductive thermal compound, spreading it over the main chip before pressing the IHS into it firmly. Didn't apply adhesive. Booting back, I didn't experience a shutoff mid boot, as I consistently do, but the thermals are identical to the results I was getting previously.

 

[freeagent]

Did you play with the multiplier or anything? You said its giving 1.38v on auto. Mine does that on auto if I change the CPU multi to 47. Seems like a lotta juice. If its giving that much, might as well start playing with that multi if you haven't already

Hate to ask the obvious but you are sure your pump is plugged in and is working?

 

[Mandrake4565]

If possible download and open Hwmonitor. Id like to see what the voltages look like when it's spiking in temps. If delidding it and re doing the paste under the IHS didn't help, it does sound like the chip is getting to much voltage. We need to figure out why.

 

[Met]

Sorry for the long delay, I'd decided to update the bios some time ago and see if that worked, however during the process of rebooting into the bios, I hadn't realized that Windows had queued an update. With my system as unstable as it was, this update froze and the computer needed to be hard reset. Afterwards I did update the bios, and the Vcore went from 1.404 to 1.200 on Auto. After that, the problem seemed to be fixed.

What wasn't fixed was the fact that my OS was now corrupted from the queued update being interrupted, which triggered a very lengthy series of attempted fixes, calls to Microsoft tech support, and then finally being forced to reinstall Windows.

That being said, while I was trying to fix the corrupt and unstable OS, I had enough room to test if the CPU was still overheating, which I could not get it to do under full load and 4.4GHz clock. It looks like the BIOS update was what finally fixed things.

 

[Zerileous]

Nice, glad this worked out!