NP9870 - Overheating CPU

Not the whole picture, bro Fox. ICD with lower/almost half the Thermal conductivity vs. MX-4 will outperform MX-4. And Conductonaut don't outperform Liquid Ultra/Pro. More as a guide line.

 

Awesome! Keep us all posted.

I checked your other threads with the GPU issue, and it seems that your CPU having problems is entirely unrelated, which is good.
Just to be sure everything makes sense, I went back to the first post and looked for the signs for pump-out related overheating, and it looked like most things seem to fit:

1.) Problem is isolated to CPU. GPU and RAM survived stress test.
2.) CPU caused BSOD when stressed with OCCT (most likely with WHEA_UNCORRECTABLE_ERROR) and black screen when playing games (probably emergency thermal shut down). Specific settings not stated, but temperatures were said to be 85-90C before failure. Most likely temps were higher AT moment of failure.
3.) Underclocking + undervolting AND NO HT managed to control the thermal and made the system stable. 3.0 GhZ with a -0.130V undervolt and still managed to bring stressed temperature to 65*C, probably with FN+1 max fan speed. Christ, I think the gimped 6700K barely draw 40W at max load in this case.
4.) MX-4 was used as thermal paste (which is low viscosity I think), and there's also a moderately long period between last application and issue.
5.) Repasting after posting the thread with the suggested method by Papusan seemed to help. Honestly the amount required by this method is definitely more than pea-sized method.
6.) 4th core is 10C hotter than the coolest core. Warped heatsink may actually promote pump-out with low-viscosity pastes.
7.) Post #14: If the CPU gets cooled down quick enough it won't shut off the laptop, but that's not often. In hindsight this makes a lot of sense that the crash was thermal related.
Also post #27: If I don't put a limit in OCCT it will hit 100c (or more, I don't know; it immediately shuts down). Consistent with 2.).
8.) Thermal pads make good contact (corroborated by pics in post #29), so it couldn't be VRM failing. Also undervolting + underclocking shouldn't anger the VRM and cause computer to shut down. Since VRM is not the case, issue is related to CPU itself. CPU is pristine, so it's the thermal paste and/or heat sink.
9.) Post #34: More or less complete coverage with a tiny amount not covering the edge of the corners. It's best I could do without it spilling over the edges of the heatspreader.
At this point yea I think it's the paste alright.

Some inconsistencies that I noticed:
1) A bit weird that the the system didn't thermal throttle at high CPU temps.
2) OP stated to have repasted multiple times trying to get the temp down prior to posting. Apparently those attempts didn't work. Probably ineffective application method/amount.

 

ICD used ASTM D-5470 to test their thermal conductivity. How other brands test theirs is pretty much not stated upfront (or at least I can't find it easily on their sites). Still, there are other things to consider about when comparison thermal pastes such as viscosity and mounting pressure. Great liquidy pastes fail eventually at low mounting pressure and uneven contact surfaces.
I think you (OP) should take a look at these thermal paste comparisons 1 ( older but still good)

 

Yes, those are interesting and useful. Thank you for the links as well. But, as usual, their did their testing on desktops. Many more wildcards come into play on laptops. IC Diamond is a good example. The thermal conductivity is not great, but it is the only paste that will work well on laptops with poorly fitting heat sinks. It is more like putty than paste. YMMV if you have a laptop. Always best to do your own testing if a laptop is involved, and do not be surprised if the results are not similar to testing with a desktop.

Nothing works nearly as well as liquid metal, but that cannot be used if the heat sink fit is poor. It will be worse than regular paste.

 

Tests done on test bench or desktops are not particularly useful for laptops who have *castrated* cooling vs. desktops.

Don't look elsewhere (laptops and desktops). Especially if you have notebooks. Jokebooks or proper laptops doesn't matter.

1.) Thermal Grizzly Conductonaut
Thermal conductivity: 73.0 W/mK
2.) Coollaboratory Liquid Ultra
Thermal conductivity: 38.4 W/mK
3.) Phobya Liquid Metal
Thermal conductivity: 40.0 W/mK

4.) Phobya NanoGrease Extreme (A very good option if you lost in the HS lottery)
Thermal conductivity: 16.0 W/mK

5.) IC Diamond (Maybe best option if you lost in the HS lottery)
Thermal conductivity: 4.5 W/mK
5.) Thermal Grizzly Kryonaut
Thermal conductivity: 12.5 W/mK
5.) Gelid GC Extreme
Thermal conductivity: 8.5 W/mK

6.) Prolimatech PK-3
Thermal conductivity: 11.2 W/mK

 

I knew I forgot about something before I hit "Post Reply"
I forgot to recommend IC Diamond given how well it served most people.

 

That post was to get people talking more than anything else. IC Diamond is better for laptops than most thermal pastes with higher thermal conductivity. MX-4, PK-1 and NT-H1 are the worst TIM products I have ever used for laptops. This is because of sloppy engineering and poor quality control in laptop manufacturing more than the quality or thermal conductance of the TIM. Those same pastes do an OK job for desktops with heat sinks that fit correctly and have firm contact pressure.

Yes, definitely. The heat sink fit was so unbelievably horrible on my P570WM that IC Diamond was literally the only thermal paste (should be called putty because it is so thick) that worked well. Everything else did a horrible job because of the horrendously poor fit of the heat sink. The replacement heat sink was an improvement, but it was still pretty bad.

 

I know... You know this. But a lot persons who don't know... As long people read review sites with tests of thermal paste - or whatever hardware (There are several of them on the web) as the devil read the Bible... Or listening on people with no knowledge. We will continue see people who will run into boiled hardware - Aka they use thermal paste who is not well suited for notebooks. This is pure fact. We are taught from childhood that we shouldn't trust 100% on strangers. The same should apply to any tech sites we visit on the web who publish own or others reviews.

 

Indeed. People that like to think of themselves as "experts" and "professionals" are often wrong. The yellow highlight is totally wrong. It might be a typographical error of some sort, but definitely not accurate. I do not know what the last bullet point is supposed to mean. It is not "thin" if that is what they meant, and it does apply uniformly every time I use it.

 

If you ignore hardware degradation you would be right, but since OP probbably ran his CPU on 90s c day after day the CPU is simply now half dead. Otherwise nobody would care about temps. , gtfo troll.

Only thing OP can do is buy new CPU, repaste it, go lapp and heatsink mod, delidd. liquid metal etc. and run it on stock or slightly overclocked otherwise it will die on him again.

 

It does thermal throttle and watching XTU stress test it thermal throttles at 95-99c, drops to 90c, turns off thermal throttle, and repeats this all over again. It gets stuck in that loop and eventually fails. Basically it can't seam to cool it quick enough.

Probably, will know for sure when I can redo it with ICD and go back to stock values.

Only started hitting 90c as of like last week when these issues started. Before that it'd never get past 80c. Based off what I've been reading this is all pretty consistent with pump-out. Won't really know more until I've repasted with ICD on Thursday when it'll be arriving.

 

It takes a whole lot more thermal abuse to harm an Intel CPU than most people think. You would have a difficult time damaging one from overheating even if you tried to do so on purpose. It is very unlikely running in the high 90's has hurt anything other than temporarily impairing performance and stability, and then only while it is overheating. Thankfully, they are extremely resilient, durable and not anywhere even remotely as fragile as GPUs and older AMD CPUs.

 

It's the surrounding components that become the worry long term like capacitors. If the heat is being drawn away by the heatsink this impact is lessened.

 

Ok, the ICD paste showed up early. I've applied it exactly as they've instructed. I've done 4 10 minute tests at 100% load. Hottest it got was 91c, which it was able to successfully thermal throttle back to 70c using OCCT (2 tests). Using XTU hottest it hit was around 81c and on average always stayed at around 70c (2 tests). Guessing OCCT is a bit more stressful on it. All 4 tests were at stock x40 speeds (4GHZ) with a -0.110v undervolt.

I'll be testing it out today in games to see if still crashing ingame, but so far I don't get any crashing or thermal throttle looping issues now in stress tests. Thank you everyone for your help. I think you may have just saved me $300-400, lol.

 

OCCT stresses the CPU cache as well as the logical processors, so I think it's definitely more taxing on the power consumption and heat generation side. ICD definitely gave you a clear improvement in stability, which is great. If nothing else goes wrong, then I think your problem is pretty much solved.