DIY heatsinks on a notebook? (#666.010)
Intro
I bought my ASUS Strix G15 during the pandemic when I needed a new computer and there was a GPU shortage (scalpers). Even though the hardware of laptops is power limited, in part to avoid issues with heat, laptops are still very limited by cooling potential (thermal mass, air flow). Additionally, fans have to be small, due to restraints in chassis size, but also spin relatively quickly to make up for the smaller surface area while still outputting a somewhat acceptable airflow to carry away heat. Unfortunately, the faster spin means that they’re more audible, and in my case produce two annoying spikes around the mid range (approx. 500Hz, see green graph). Instead of a gentle, low frequency whoosh of a traditional 140mm case fan, these smaller fans are more of a whine.
Why is it stood up? To give the back less restriction to fresh air.
I have anecdotally noticed lower temps and less fan noise in this position - Yes, I should’ve tested this while I was at it, but didn’t think of it.
CPU: AMD Ryzen 7 5800H
GPU: NVIDIA GeForce RTX 3060 Laptop
Drives: Sabrent Rocket 4 Plus 2TB (C:), Toshiba XG3 1TB (M:)
The fix
So, I thought to bandaid this gash in my day-to-day life by increasing the thermal mass and airflow with a heatsink directly on the heatpipes that cover the CPU, and another over the GPU. Additionally, since they’re very inexpensive, I thought to slap on some heatsinks on the SSDs, too, just as a test.
There are 13K/W thermal pads under heatsinks, and the fan is a BeQuiet Pure Wings 2 140mm modded to work with a USB-A header. According to its spec sheet, it will spin at approx 500rpm at 5V, which is USB’s max output voltage. It’s utterly inaudible from my desk position.
Methodology
I clicked on “Start Logger” in HWiNFO and immediately started either FurMark, CPU Burner (part of FurMark) or CrystalDiskMark. I also took screenshots as a precaution after every test. These screenshots and the raw data are available from this websites download central: TinyURL.com/666TonsOfDownloads, the mentioned files are in folder: “Graphics - Spectrograms, charts, etc”, then folder “666.010 - Notebook Heatsink Testing”.
I allocated limited time to this project, so used HWiNFO’s chart extremes as testing length. When the test start reached the left side I stopped testing. This brought test lengths to just under 200 seconds.
Results
NB: Utilization of target component was pinned at 100%, so this was not graphed.
CPU
While stock stagnated down to 50 Watts out of the rated 65 Watts package power at which it should be performing, the mod stagnated to approx 58 Watts. That’s a 10-15% improvement.
However, the package temperature stayed the same (while the power increased, remember), with less fan usage in the second half of the test. Win. Kind of win. Win.
Keep in mind that this is an isolated test. Because the cooling system is partly shared, if the GPU were processing as well, the CPU wouldn’t stand a chance. Thankfully, in gaming at least, the CPU is never at 100% like this in my experience.
GPU
Overall, a win. Though, the very minor delay and then equal stagnation in fan speed suggests that there’s still some physical limitation. Plus, there are still some dips in power towards the end, suggesting that if the test were run longer, it might have eventually dipped further in power. 130Watt is a lot for a medium-sized laptop, and it shows.
Drives
Night and day.
Though, even without a heatsink, the Sabrent (C:) at max usage is well below the 70°C at which that drive would start throttling (source).
The quite old Toshiba (M:) with 40TB in total host reads is still safe at stock, but could definitely benefit with AT LEAST a copper foil/graphene sticker-style heatsink - like the one Sabrent use. (Scroll down a bit in this article)
Overall, these are welcome improvements and should hypothetically improve the very long term reliability.
Discussion / Improvements
I was very happy seeing a measurable increase in power in/output for processing, meaning there would be better spent power going into actual performance. But…
Since my device is never at 100% utilization for any more than a few seconds (gaming, audio, not even video/3D does for my specific usage), any additional cooling performance that THIS method provides is too inconvenient to recommend for average notebook users. I’ve already disassembled it, as I’d have to have the device exposed 24/7 otherwise. That’s something with which I’m really not comfortable.
However, I’ll for sure be ordering a copper/graphene sticker for the Toshiba drive like the ones Sabrent use (I fully believe that is why that drive ran so cool, comparatively). That little fix should last me until I can build my own tower again. 😈 Speaking of which, how does this parts list look?
If you attempt this, know that a copper/graphene sticker on a warm SSD can be really helpful, but, for anything else that isn’t just delaying the thermal throttling by a minute or so, a more permanent solution would be needed such as one of the videos below. The former is quite time intensive and seems more like a fun project with an old laptop as it involved a full custom chassis swap and a lot of modding. The two latter are a little more quick-fix, but have their own issues… At this point, we really need to ask, why did I get a laptop? 😂
Thanks if you’ve made it this far.
This was my first computer-themed experiment, and one I’ve always wondered about as a student with a notebook and now a functioning (!?!?) adult also with a notebook (kind of by choice).
And please get in touch if you saw I missed something or have suggestions 😁
2024 10 05 UPDATE
Three days later, my graphene/copper (2280-sized) stickers arrived for the old (hot-running) Toshiba drive:
What is “Cut Heatsink” on the graph?
The surface mount components on the PCB have different heights. The sticker in the Mod test (green) didn’t stick onto the (square, top left on PCB) storage controller, as it’s shorter than the larger rectangular storage components on the PCB, and I thought that the terrible performance was due to hot air being trapped in that space between the sticker and controller. So, I wanted to ensure it made proper contact and cut it to fit the storage controller. This mad things worse somehow…
Damn. I really had high hopes for this one, especially because it would increase the surface area significantly while me still being able to close the notebook backing. Sadly, I think that the heatsink somehow blocks airflow to essential components on the PCB and the sticker just isn’t as thermally conductive as the component’s own thermal conductivity relative to the air.
This doesn’t mean that it won’t work on your SSD as this particular situation may be due to the design of the components on this PCB, but it also might be an issue with the copper/graphene sticker; It’s a brand I don’t recognize (this alone means nothing, though). Anecdotally, it does feel cool to the touch and rapidly warms in your hand - which is a very good sign.
I did try an IcyBox and fnateck external enclosure for 2280 NVME, only to find out that these don’t report temperature.
As with anything, more testing is necessary, but we’re out of time, so, good night. Come back again.
