Laptop is getting hot, how do I get it to run cooler?

Laptop: Thinkpad E470

Specs:
i7 7500u (dual core 3.5GHz)
940MX 2GB DDR3
8GB DDR4 SODIMM single channel


So my laptop gets hot, really hot. Like 97 degrees on GPU and like 85 degrees on GPU when gaming. I was wondering how to tackle this problem

I tried undervolting, still hits 97 degrees. Performance in games is crippled, from 120FPS average to below 30 with huge drops

I tried manual fan control, still hits 97 degrees after a while


This is what I am working with. CPU is up top, GPU is the small square

I will try cleaning that fan and see if it helps

I was thinking a repaste is now in order. I am stuck between liquid metal, normal thermal paste, or a graphite pad.

I am a student so this laptop is going to get thrown around a lot, I’m not too sure if liquid metal would be a reliable solution.

At the same time, I’m not sure even a high end thermal paste like kryonaut would help. Both undervolting and cranking the fans haven’t even made a dent in temps, im not sure a paste has the punch I need

I’m also worried the paste or liquid metal will get messed up when I go outside. Walking through -25 degree weather with wind means that the laptop gets really cool and pastes might get hardened and softened. That’s why I was also considering a graphite pad

I really can’t figure this out, so I thought I might have a better chance if I ask around.

Any help would be appreciated

 

Hello,
From your description, I would try changing thermal paste, and making sure it is doing a good contact on CPU and GPU before anything.

 

What should I change it with? Liduid metal? Paste? Pad

 

Any good thermal paste will work unless you have defective heatsink. I used CoolerMaster Mastergel maker nano, Gelid GC Extreme is also a good paste.

 

Gotcha, I’ll give that a shot

 

Undervolt it,even if its a ULV CPU

 

I did try that already, still nada

 

I'd recommend Thermal Grizzly Kryonaut, it's the best non-conductive thermal paste in my opinion. Their Carbonaut pad should be a good alternative. How long do you typically walk in -25?

 

I usually stroll around half an hour or so when it's cold. I'm worried that that low temp will make liquid metals and most pastes dry/harden and that would really impact performance. I know carbon pads don't get affected by the cold

 

@Richard Zheng Conductonaut review. There should be alternatives from other brands as well.

 

So I just had my laptop apart 3 separate times to test the IC Graphite Thermal Pad. In short; I do not recommend it for direct-die style CPU scenarios (your laptop).

The main issue, an an issue I have seen frequently, is the heatsink on the laptop is shared between the CPU & GPU. It uses only 3 mounting locations over the GPU, and 4 across the CPU. The distribution of pressure on these 2 chips is further altered by a thin black adhesive on the underside of the heatsink that directly outlines the processor die. As a result of all of this, getting even and proper mounting pressure onto the chips is not easy. The thermal pad is so thin, it simply floats between the processor die and the heatsink.

I first tried a single pad. Within the first 60 seconds of benchmarking I was hitting 90C on my GPU which has never happened before. The CPU was only slightly worse off, around 96C almost immediately. These are temps I've never hit before. I immediately shut down and regrouped.

I went in and scraped off the plastic black adhesive around the heatsink surface to ensure it was completely flat and could not impede contact. After cleaning the heatsink, I applied TWO sheets of graphite per processor die. Effectively doubling the thickness and ensuring the heatsink 'squished' it down. I made sure each mounting screw was fully tightened.
On initial boot up, temps were very good. But again, once I fired up a benchmark the temps immediately shot up to 80C. This time though, they settled around 85C and hovered. An interesting finding, but still a lot hotter than my previous thermal paste application.

Defeated, I once again disassembled the laptop and then re-applied my original thermal paste; Noctua NT-H1. After the default spread of paste and reassembly I booted to verify my computer was back to normal.

After an hour of stress tests I can confirm the thermal paste is FAR superior, at least in this case. GPU/CPU idle = 34C/40C GPU/CPU load = 68C/80C (extended).

I really think these graphite pads are intended for CPU's with 'lids' and much more accurate heatsink mounting. Laptop heatsinks are very tricky and the small die size leaves almost no room for error. It's either not tight and there's a gap or you over-tighten and crush the processor die.

 

Thanks for this brilliant write up! I probably won’t go with graphite pads now knowing how temperamental they can be. I think I will go with the “safe” option and go with kryonaut for CPU and GPU. I might try liquid metal if that still can’t get it cool

 

So, how did it go?

 

I decided to scrap that idea. I found that just limiting CPU to 15W or so did the trick. Throttling was all but eliminated. I'm pretty sure that repaste would have been better, but just a TDP limit worked well enough for my needs

 

So you replaced throttling... with throttling? Alright my dude whatever works for you.

 

I replaced wild thermal throttling that would cripple performance with a stable throttled system. Rather than HUGE frequency spikes, I get a smooth line. It works for me, but obviously isn't the best solution

 

That's a good point! I didn't know your ThinkPad also has lousy implementation of thermal throttling. My Yoga 720 is super stable and stays right at the equilibrium. I wish more laptops had a PID style algorithm instead of bang-bang control for temperature modulation.

 

Damn, the Yoga 720 is a brilliant machine. I wish Lenovo would make a 2 in 1 version of the X1 Extreme, that would be my dream laptop.

I wish a lot more laptops took the "slow but steady" approach rather than the "gun it till we hit 100" approach since wild throttling kills performance. I guess the "gun it" approach is better for short loads, but longer loads struggle to keep stable clocks