PC hard-shutting off, CPU may be hitting thermal limits

[0xDEADBEEF]

Hey folks!

 

I'm trying to diagnose an issue remotely for a friend. She has a 9700k and an RTX 2070 from iBuyPower. The cooler is some kind of 120mm AIO with push/pull. She's had this build for a bit over a year, and very recently the PC has started to hard shut off while she's playing (FF 14). No blue screen, just cut to black. At first I assumed power supply instead of overheating, as I had a similar issue in the past. I wanted to be able to reliably induce the issue so we could tell if we fixed it, by stressing the parts and getting the PSU to shut off.

 

I had her download and run Furmark to start. The GPU handled it fine, but weirdly a single CPU core spiked to high utilization, and the entire CPU got real toasty (100+C on a core). You can see the temps and utilization in the attached pic. AFAIK Furmark does barely anything to CPUs; I ran it on my rig today and my CPU stayed at largely idle, so that was weird. But in any case, the CPU shouldn't be getting that hot, especially not when a single core is fired up. At that point I favored stopping instead of trying to induce the shutdown.

 

Potential things:
1. The cooler is bad, maybe the pump
2. The cooler and/or thermal paste is installed poorly/insufficiently
3. Some kind of hardware fault causing the CPU to get abnormally hot

4. Some kind of software issue causing excess current?

 

She can feel the the AIO tubes vibrating. I had her turn off the PC, tilt the case a couple of times to get air bubbles moving around, and listen for water flow as it turned on, but we can't yet confirm that the pump is definitely moving water.

 

#2-4 I think get progressively less likely, as I'd be surprised as to why it's suddenly becoming so power hungry, but I couldn't give you any examples of 3 and 4. And even if the cooler install was insufficient, why would any of 2-4 start becoming a major problem _now_?

 

I'm having her order thermal paste to try reseating the CPU cooler with fresh paste, as the simplest and cheapest option.

 

Any thoughts/ideas on what else we should consider or test? Thanks in advance.

 

[Deli]

An inexpensive air tower cooler will have this problem solved. That's why I hate prebuild comes with AIO.

[0xDEADBEEF]

2 minutes ago, Deli said:

A inexpensive air tower cooler will have this problem solved. That's why I hate prebuild comes with AIO.

Hey I've been off the AIO train, but I'm also yet unconvinced that it's the cooler itself until we've actually confirmed the issue sticks around.

[Deli]

Just now, 0xDEADBEEF said:

Hey I've been off the AIO train, but I'm also yet unconvinced that it's the cooler itself until we've actually confirmed the issue sticks around.

CPU utilisation 26%, power consumption 46w, CPU temperature 102C....

[Deli]

AIO is fine. However prebuild usually tries to cut corners and comes with the worst quality AIO. That's the problem.

[0xDEADBEEF]

5 minutes ago, Deli said:

CPU utilisation 26%, power consumption 46w, CPU temperature 102C....

I know I know. It's just one of those "make no assumptions" situations. I can't tell yet if the pump is working, so for all I know it's just seated poorly or the paste is bad.

 

Someone suggested grabbing the two tubes of the AIO and see if there's a notable temperature difference.

[0xDEADBEEF]

I'm still scratching my head that:

 

1. Furmark is pegging a CPU. On my 6700k right now, Furmark does _nothing_ to my CPU utilization or temps (as expected).

2. One core being pegged is causing everything to heat up _that much_

 

I'm paranoid that there's some kind of software or hardware issue that I can't identify that's causing it to bake, like too much current. A new cooler wouldn't solve that.

[Deli]

Just now, 0xDEADBEEF said:

I know I know. It's just one of those "make no assumptions" situations. I can't tell yet if the pump is working, so for all I know it's just seated poorly or the paste is bad.

 

Someone suggested grabbing the two tubes of the AIO and see if there's a notable temperature difference.

Even if the pump is running. The coolant could be low because of permeation. That isn't enough flow across the cool plate.

[0xDEADBEEF]

3 minutes ago, Deli said:

Even if the pump is running. The coolant could be low because of permeation. That isn't enough flow across the cool plate.

After a bit over a year? It _has_ gotten very hot where she is, but I'd be surprised if it could evaporate that much. I had a Corsair H50 that ran reliably for several years before I retired it (I know that's just one anecdote).

 

Anyway, I'll still hope that some more ideas pop up in this thread, so I have more to consider and/or test than just a bad cooler.

[Deli]

9 minutes ago, 0xDEADBEEF said:

I'm still scratching my head that:

 

1. Furmark is pegging a CPU. On my 6700k right now, Furmark does _nothing_ to my CPU utilization or temps (as expected).

2. One core being pegged is causing everything to heat up _that much_

 

I'm paranoid that there's some kind of software or hardware issue that I can't identify that's causing it to bake, like too much current. A new cooler wouldn't solve that.

Could be something running in the back ground. But if HWMonitor is correct. The CPU is only drawing 46w. No way the CPU package will hit 102C. Even if two cores are at max load. The only conclusion is the cooler isn't functioning as it should be.

[Deli]

2 minutes ago, 0xDEADBEEF said:

After a bit over a year? It _has_ gotten very hot where she is, but I'd be surprised if it could evaporate that much. I had a Corsair H50 that ran reliably for several years before I retired it (I know that's just one anecdote).

iBuypower doesn't use the best quality AIO. I don't think that's any surprise.

[0xDEADBEEF]

Bump, just in case anyone ends up with an inspired response.

[LordVetinari]

Occam's razor - Hear hooves think horses. 

 

CPU at 26% and 100°+...... Pick up a cheap air cooler at the same time as the thermal paste. Will soon let you know if the AIO  has gone bad or if it's anything deeper.

 

But yeah I doubt any zebras are around. 

[0xDEADBEEF]

Can confirm, NH-D15 is better than a broken CLC.