[HARDWARE MOD] P157SM cooling overhaul

[MC-Sammer]

Over the last month or so, I've been planning on overhauling my (slightly dated) P157SM to give it better cooling, and possibly higher performance.

 

There were a number of sources I used for inspiration, and I'll attempt to document all of them later on, but first, pictures!

 

This is the first modification I made. I increased the opening of the CPU fan to allow more air to go into the fan. As I recall, it helped with idle temperatures but I don't remember by how much.

This picture shows how much I increased the opening by, I didn't do just that bit and stop

 

Next up, I increased the width of the intakes directly over the GPU and CPU to reduce ambient heat buildup in those areas.

 

 

I ultimately decided that increasing the width was not the best route to go (and it was quite ugly) so I decided to simply remove the vents

 

I used some diamond mesh to cover those areas, but I don't have any pictures of that at the moment.

 

 

 

Next I used some HVAC tape to direct the air flow from the CPU fan into the heatsink and not out the tiny gaps in the area.

 

Then I swapped out the heatsinks (originally I ordered this from XoticPC and went with their copper cooling upgrade (I was young(er)) for some bigger ones.

 

 

 

Everything is all nice and snug

 

 

And finally with the lid back on.

 

With the changes made between cutting the vents, and the last picture, I was able to drop load CPU temps ~5C which I thought was pretty nice.

 

My goal is to also get rid of the vent over the fan, but I haven't found time to do it. I'm also planning on lapping the GPU heatsink, which will, again, take time.

 

I'll post more updates as I make them (and post the benchmarks I've already done).

 

Update:

GPU Work:

So I have a 780M at the moment (I'm planning on going to a 980M, though I may just wait for the Pascal GPU's to come out) and I decided to lap the heatsink, since I'm planning on overclocking it.

 

Fresh of the block

400 Grit (10 minutes)

600 Grit (10 minutes)

800 Grit (10 minutes)

1000 grit (7 Minutes)

1200 Grit (7 Minutes)

1500 grit (about 25 minutes for this stage)

 

And I thought it looked pretty smooth. One of my references mentioned cutting off the little tabs on the block, since they're mostly there for adding additional pressure (since the plate is usually warped)  and I wanted to just lap the area directly over the GPU die (plus I don't really have a good place for grinding copper with my rotary tool).

 

 

For the paste I used http://www.coolermaster.com/cooling/thermal-compound/mastergel-maker/ which I went with based on the favorable review posted at http://www.play3r.net/reviews/cooling/cooler-master-master-gel-maker-nano-thermal-paste-performance/ and the pads are Fujipoly high end (didn't think I'd need to go ultra).

 

 

I also added some little sinks on to the heatpipes, though they might be suspect (more on that in the results below).

 

 

And I sealed up a gap in the fan to direct more airflow through the larger fins.

 

 

I also used from more HVAC tape to try to prevent ambient air from being drawn in to the fans, but this too might be suspect (it might be needed to draw it away, since where else is the heat from all those fins supposed to go?)

 

And the results are... interesting...

pre upgrade

Spoiler

Lakes of Titan x32
Preset 1080
5033 points, 83 Fps
70C max

 

Plasma (1M-particle)
Preset 1080
5090 points, 84Fps
59C

 

PhysX 3 Fluids (GPU)
Preset: 1080
2249 Points 37 Fps
63C Max

 

Stress test: GPU core burner V2 (furry donut)
Preset: 1080
after 400 seconds
86C

post upgrade

Spoiler

Lakes of Titan x32
Preset 1080
5057 points, 84 Fps
64C max

 

Plasma (1M-particle)
Preset 1080
4852 points, 84Fps
57C

 

PhysX 3 Fluids (GPU)
Preset: 1080
2158Points 35 Fps
64C Max

 

Stress test: GPU core burner V2 (furry donut)
Preset: 1080
after 300 seconds
85C

 

So the initial benchmark had a 5C drop in temperature, but the following results aren't as promising.

In fact, the stress test results appears to be higher (I'm speculating that it'd be over 86C by 400 seconds).

 

I'm thinking that either the added heatsinks on the heatpipes are dispersing the heat before it can get to the fins, my lapping job wasn't so great, or I added too much TIM. I may try the washer trick from the p150 cooling threat mentioned below, but I'm spent on energy for this weekend.

 

June 3rd update

I finally got back around to working on my clevo some more and after a third re-pasting my temps are down (taking into account that today is hotter than when I first tested)

Spoiler

Stock
780M, 850MHz core, 2500MHz ram

Lakes of Titan x32
Preset 1080
5033 points, 83 Fps
73C max

 

Plasma (1M-particle)
Preset 1080
5020 points, 83Fps
58C

 

PhysX 3 Fluids (GPU)
Preset: 1080
2180 Points 36 Fps
65C Max

 

Stress test: GPU core burner V2 (furry donut)
Preset: 1080
after 400 seconds
82C

 

I suspect that my older paste jobs were either too thick, or I bent the heatpipe and kinda threw the plate out of alignment.

 

Inspiration:

Modding my clevo in general:

http://null-bin.blogspot.com/2015/04/extensive-clevo-p150sm-cooling-and.html

Sealing up the gaps in the fins:

Additional cooling ideas:

 

 

 

 

Edited June 3, 2016 by MC-Sammer
Updates

[tmash]

Would be nice to see temps with CPU OC, I need more fins for increased "virtual" cooling capacity 

Edited April 14, 2016 by tmash

[MC-Sammer]

20 hours ago, tmash said:

Would be nice to see temps with CPU OC, I need more fins for increased "virtual" cooling capacity 

 

Yeah I'm thinking that too, though I won't have a 'before' to compare against

 

I'm back up to uC 17 and I installed XTU and throttlestop is all unhappy and won't hold the multiplier settings

 

But I'm gonna be working on the GPU side of things this weekend, so maybe I'll play around some more with uC and try getting rid of XTU and see if that helps

Edited April 15, 2016 by MC-Sammer

[Brian]

Looking forward to your final tweaks to get temps down.

[MC-Sammer]

Howdy all,

 

So I've been having some issues with my P157SM overheating while gaming (particularly with fallout 4). It seems to me that the fans aren't kicking on early enough to keep the system cool, though I know from logging at least the CPU and GPU are within reasonable limits. It's sort of strange that I can't seem to pinpoint one particular thing that is overheating, My current guesses are either the caps/resistors above the CPU (covered by the little copper edge of the CPU plate) or some of the thermal pads on the GPU plate aren't lined up correctly. It's also possible that the EC firmware from @Prema isn't kicking in quite early enough, but that'd be surprising for him to not have already addressed this at some point.

 

The only time I've been able to re-create it outside of Fallout is one time I was running both the CPU and the fuzzy donut GPU burner from MSI Kombuster. I haven't been able to get it to over heat just running one or the other.

 

I'll have to re-examine the pads one of these upcoming weekends, though it'd be a shame to screw up the good paste job I did on the GPU this last time lol

 

PS: I sent some financial love Prema's way, because he really deserves it!

Edited June 8, 2016 by MC-Sammer

[senso]

Have you ever repasted you laptop?

[MC-Sammer]

@senso Yup, both the CPU and GPU.

[MC-Sammer]

And now it appears that I've fried some portion of the integrated GPU

 

I'm able to get windows loading correctly with a generic driver from M$, but I had to use gpedit to block the driver from reinstalling itself.

 

And now to debate if I should fork up the dough for a new 4700MQ, or if I should invest in a new system; decisions, decisions...

[senso]

19 hours ago, MC-Sammer said:

@senso Yup, both the CPU and GPU.

 

How long ago?

It should be done annually, and if you have an air compressor blasting(1-3 secs is more than enough) the vents every month so they won't build up dust will keep it nice and cool.