Framework Laptop: Repasting & Two Year Report

#device-review #computer-repairing

Updated 2023-11-27: Added stuff on the RTC battery replacement module installation and part swap.

“Repasting” with PTM 7950

I’ve been reading testimonials from people replacing thermal paste in their laptops with PTM 7950, mostly praising the amazing effectiveness of the material, with some reporting a drastic 20 degrees C reduction. I was skeptical of such claims. Such temperature reduction just seemed impossible to me, especially considering the conductivity of PTM 7950 is pretty much the same as most non-terrible thermal pastes on paper.

Anyway I decided to give it a try myself when I saw this mystic material on sale for barely above $10 the other day.

The repasting process went pretty smoothly. However I did notice that the screw marked “3” on the heatsink seemed a lot looser than the other two. Maybe that has something to do with the horrible thermal I had before…

Time for the results. Spoiler alert: it’s impressive. I cannot do a scientific before and after comparison, but I did run a few of my day-to-day tasks as tests. Idle temperatures are already much better: 50°C before vs 38°C after. Under a short code compilation load (around 5 minutes), the initial temperatures are similar (~100°C before vs 96°C after), but the CPU was able to boost much closer to its designated PL2 power (58W vs 45W, advertised PL2 being 60W). After the processor stabilizes at its PL1 power it goes down to below 80°C, while the under the old paste it was a constant 90+°C. The task also finishes ~10% faster than before.

While my results aren’t as insane as some others got, it was still huge. The biggest improvement is that merely browsing the Internet without setting CPU power policy to powersave is now bearable – I’ve mentioned in my previous post that the fan noise of the Framework laptop is a pretty big issue for me personally, as the noise it emits when the processor is above 55°C is extremely distracting in a quiet room. I also got ~10% more performance out of this CPU, which probably makes this the best $10 I spent on tech-related products for the past few years.

So would I recommend the PTM 7950? Yes, but only when you have to do a repaste. As I mention before, I was skeptical of claims about the magical power of PTM 7950. To be clear I wasn’t doubting the thermal conductivity of the material, but rather the cause of the temperature reduction people have been reporting – maybe their old paste had completely dried out, or the factory paste application is utter garbage. For my specific case – I have my reasons to suspect the same due to the aforementioned loose screw on the heatsink I had. While I have never touched the heatsink assembly myself since day one until today, and the thermal profile of the machine was fairly consistent for the past 2 years, I couldn’t rule out everything. So I don’t really know if it is a factory defect or something else. Also I’ve seen reports that certain newer laptops already use similar phase change material as its TIM. So if you’re considering to repaste your brand new laptop (which isn’t always a great idea in the first place) with PTM 7950, do remember to check if that’s the case.

Factory thermal paste application

Factory paste removed

PTM 7950 applied

Semi-long term review

Disassembly reveals …

First disassembly in 2 years.

A few things jumped out when I opened up the laptop for repasting. The first, very obvious and somewhat worrying one is the mildly swollen battery (also known as SpicyPillows). I did not run the laptop a lot on battery during the first year, but the second year saw more usage with it unplugged (still not as frequent as my ThinkPad X1 yoga 1st gen). The swelling isn’t anywhere near extreme, and there hasn’t been a significant reduction of battery life since I got the machine. For this reason I’m not yet very compelled to replace it. But I will definitely be monitoring it much more closely from this point on, as it’s very well known that SpicyPillows are potentially dangerous.

The second, less obvious one is also about a battery, but this time the CMOS battery. It was completely flat. I was aware of a design flaw in these 11th gen Intel machines that the CMOS battery only gets charged when the laptop is on AC power. However I’m not sure this is the cause in my case as my laptop spends most of its lifetime powered on and connected to AC power. While I do appreciate the original, extremely detailed makeshift solution (which is under the “Archived” section of the post linked above), I’m not comfortable with the idea of scratching a trace off an expensive motherboard. Anyway now that there’s a solution that doesn’t involve scratching a trace off the motherboard, I’m going to install the RTC battery substitute module and update this post when that’s done.

Other early adopter issues

The touchpad is flaky. Being a touchpad hater as I am (and a trackpoint user for more than half my life), I had to deal with it when I’m on the go. A few months ago I found that physical clicks on the touchpad no longer works, and apparently I’m not the only one. [1] There appears to be a simple fix which consists of a single step that sounds a little ridiculous (“Try pressing the bottom middle of the Touchpad firmly a few times”), but it worked for me.

I do venture out to become an early adopter of certain products sometimes, My ThinkPad X1 yoga 1st gen being another example where Lenovo experimented with the “yoga” form factor in their ThinkPad line-up for the first time. Being an early(-ish) adopter of the Framework laptop, I fully expected that there could be issues. So far I’ve come across the CMOS battery design flaw, touchpad issue, BIOS bugs, unreasonable level of CPU throttling and fingerprint reader driver overheating the power button. None of these are severe enough to be a deal breaker for me, and all of the listed issues have been more or less fixed. I guess my overall experience of being an early adopter has been overall positive.

How well did it hold up?

Performance wise, it’s still a beast in its class. These Willow Cove cores still eats the Rust compilation workload that I now have to deal with on a daily basis as breakfast. Gaming still is an option for me considering the only game that I play nowadays is Minecraft, but is likely a no-no for any heavier games. While it’s not the most power efficient solution available today, it still meets the requirements of my setup just fine.

Battery life isn’t too great in the first place. But the lifetime seems to hold up decently despite having a slightly swollen battery pack. Again, still fulfills my needs.

Internals… don’t look great (CW: filthy looking fan). The fan seems to attract dust much faster than the one in my ThinkPad X1 yoga.

Random rants

I still find the default fan curve of the laptop downright stupid. Ramping it up to 80% of full speed at 60°C doesn’t seem wise to me. Yes there are third party tools that can force a fan speed, but I’d rather program a more sane fan curve instead.

Intel’s Xe driver for Windows is still hot garbage for graphics development.

I do not like the forum software Framework uses (Discourse). The idea of infinite scrolling never clicked with me. You simultaneously get the worst from pagination and single-paged designs. And their page search is still crippled to this day. Okay I’ll stop here and keep my other 10 objections to Discourse with myself.

Contemplating upgrades

Framework promised to provide an upgrade path to owners of their laptops and they delivered. Now I have the choice of replacing the motherboard in my Framework with a 12th/13th gen Intel board, or a Zen 4 based AMD board. I’ve explained why I didn’t upgrade to the 12th gen Intel board in my previous post, and I’ll state the most important part here again: I don’t want to ditch my old board that’s perfectly functional. I’ll also say the silent part that was omitted in the previous post aloud: I don’t want to purchase the extra parts to build a complete computer out of the old board either.

There are additional reasons that I didn’t upgrade to specific available boards. 13th gen Intel boards are especially objectionable due to the fact that they still use DDR4 memory. This is year 2023 and DDR5 memory isn’t hard to find now. Yes I guess you get to keep your old memory sticks, but I don’t want an upgrade that seems half-baked thank you very much.

AMD boards seem way more compelling, especially considering the performance boost they could bring at the price point they are sold at. But I’ve heard that they produce pretty much the exact same noise as the original boards once installed, runs as hot under heavy load, and would result in increased pain just by the virtue of being AMD [2]. And I would need to reconfigure the entire kernel [3]. For this reason I’ve pretty much shelved the idea of upgrading the motherboard in the very near future. The only reason for a motherboard upgrade for me now would be the old one straight up failing or become unsuitable for my tasks.

There is one upgrade I want to make though. I would absolutely love to replace the glossy screen from factory with the new matte one. I was never a fan of glossy screens on any laptop and will take a matte replacement any day of the week. The price is a little steep though.

There’s another upgrade that I’d love to get. A trackpoint keyboard with proper mouse buttons. I’m sure Lenovo would do anything to sabotage it, but I can always dream.

RTC battery substitute module installation and upgrades

The battery module was sent out the day before US Thanksgiving. I also decided to treat myself with the aforementioned matte screen upgrade and a proper Chinese keyboard, which were sent out on the same day.

Following the official guide linked above, the installation was pretty straight forward. I don’t have stellar soldering skills or very stable hands, but here are a few photos from the installation plus my commentry.

RTC battery substitute module packing. Unassuming.

The module itself.

After soldering. I applied a little bit too much solder.

Final position.

Troubleshooting the modded laptop

Unfortunately, as things usually are for me, it didn’t go right after this point. Immediately after finishing the mod the machine still refused to turn on with the AC adaptor unplugged. It could boot with the adaptor plugged in and stay on after unplugging it while the laptop is running. Unplugging the machine with it turned off would erase all CMOS data. This was the exact same behavior exhibited before I performed the RTC battery modification.

So I had to start the troubleshooting process. The power wire to the RTC battery substitute module measured 17.5 V to ground when the machine was powered on, so I didn’t mess up the soldering [4]. The outer ring of the module measured 3.2 V to ground, so the module itself is good. Neither values changed when I turned the machine off or had it unplugged after that. At this point pretty much all hardware faults that I could think of have been ruled out.

Then I remembered that my machine is on BIOS version 3.17, while the latest release was 3.19 for this motherboard. So I decided to give that a shot. I quickly grabbed the update from their website and installed it, and voilà: the machine was able to turn on without the AC power attached once the BIOS has been updated.

I still felt a bit iffy with my installation though, so I decided to go through the official installation guide again to make sure I followed everything – and of course I did not. There is apparently a specific orientation requirement for the module (a working one is shown in the second picture of step 3 in the guide). So I adjusted mine to match the picture for good measure, which resulted in the final installation photo shown above. There was also a comment under the official mentioning the module orientation issue.

3.2 V present on outer ring of the module with the computer powered on and plugged in.

17.5 V going into the module with the computer powered on and plugged in.

3.2 V still present when the machine is turned off but remain plugged in.

3.2 V still present when the machine is both turned off and unplugged.

By now I have the RTC battery substitute module fully working as Framework intended. However I couldn’t help but notice that this still doesn’t give the laptop the normally expected behavior according to the schematic of the module and a statement from Framework themselves, both of which can be found here. For a normal laptop one would expect when its main battery runs out it would still be able to keep time and the BIOS configuration. However with this modification being essentially just a voltage regulator (NCP718) and having no significant energy storage at all, the machine will lose all CMOS data once the main battery cuts out, which functions more like a normal laptop with a dead CMOS battery. But I guess it’s probably better than losing CMOS data even with a fully charged main battery…

Other upgrades

I actually performed these part swaps between soldering the battery module thingy in and the troubleshooting process. But that didn’t really matter because these upgrades went pretty smoothly without any drama. There were a lot of screws to undo and screw back in though (25 black ones and 48 silver ones?? I couldn’t really keep track of the number.) [5] These screws are pretty annoying to handle because they have incredibly short shanks and are incredibly easy to strip. With the official screw driver they easily stick together and the ifixit screw driver couldn’t reliably pick them up. The sheer number of them only added to the trouble I had to go through to replace my perfectly working keyboard with one that has additional legends on the keys that I will barely look at. Is it really worth the trouble? (Yes… if it had a trackpoint goddammit!)

The screen. Wanky Apple-esque packaging.

Screen unboxed. Apparently the newer 13 inch AMD model comes with this type of screen.
(which is objectively superior btw)

Keyboard unboxed. I can type Chinese using bopomofo at like 3 ch/min and I barely remember anything about Cangjie.

National guild of screws. Nobody will get this reference.

The screens side by side. Original screen on the left. The matte screen actually looked very good for what it is.

Back of the screens. Original on the left. Both made by BOE.

The new keyboard with backlight on. Fun fact: practically all computers sold in mainland China come with a standard US ANSI keyboard.

[1]: [1] [2]
[2]: Their crappy GPU driver being a major PITA for me – my experience with AMD graphics seems very inconsistent with most Linux users. No I refuse to elaborate here because that would derail the entire post.
[3]: Because I run Gentoo. Now laugh at me as you wish.
[4]: Even admittedly it looked pretty bad as all my soldering jobs have been so far.
[5]: So many that they actually forgot to circle out two of them in their official guide. Multiple comments under the guide already pointed this out, but Framework hasn’t modified the guide at the time of writing.