2.5 Years review of Framework Laptop 13 7840U as a daily driver

1. The good

1.1 Linux support

Framework Computer is one of the brands that have official Linux support, with more than one distros. More community supported distros on discussion topics.

1.2 Official documentations

In the knowledgebase, there are tons of official documentations with videos about upgrading and installing software and hardware such as drivers, BIOS firmware installation, replacing RAM, SSD, battery and keyboards.

1.3 Customizability and openness.

In the “Configure Now” page, you can select numerous combinations of hardware parts to build your computers. However this advantage is only present on the DIY version, the pre-builts are the same “normal, pro, more pro, max pro” versions like most other brands.

Some (not all due to NDA) schematics are available and the embedded controller (EC) is open-source so advanced users can make their own modules and/or reprogram the EC to enhance performance or solve problems.

1.4 Transparent BIOS updates (important, I think)

Framework Computer has an official “channel” or category for BIOS announcement which listed highlights and upgrade process with words that are easier to understand. Linux Vendor Firmware Service (LVFS) are supported on all models. In comparison, most “mainstream” brands such as Dell, ASUS and HP, are lack of transparency and sometimes done in bad faith. They often incorporate their BIOS updates into Windows Updates, especially in recent times these updates are shoved in the “important” rather than the “optional” channel, making them nearly unavoidable for laypeople. Unsolicited BIOS updates on these brands often cause performance degradation or even render them bricks

Examples



Some links: bricking, function removal, more bricking, planned TDP reduction.

1.5 Balanced performance and good cost (7840U model, somewhat opinionated)

8c16t makes this model a good balance on performance and efficiency, using TLP or TuneD doesn’t affect battery runtime much. Since all cores are equal there’s less concern of which core to use under various tasks. Newer models with performance and efficiency cores have better runtime in theory but in practice the deviation is much larger, some much better some much worse than the 7840U. I think the requirement of software, operating system and programs, are higher for those CPUs to utilized the cores correctly.
The 7840U is also more cost-efficient, better than Intel 13th gen and Core Ultra 1. The AI 300 Series are expensive sidegrades.

1.6 USB-C PD charging

All four ports have USB-C charging supports with 9V 18W to 20V 100W. There’s very little performance impact on battery or a low wattage charger.

2. The neutral

2.1 Expansion cards

These “expansion cards” are hot-swappable but most premium laptops (minus Apple) have more than 4 ports and some expansion cards have more standby drains. My main complaint is the lack of space efficiency, it has enough space but why don’t they made the micro-SD card fully in when installed so the user can put the laptop in a backpack both ways?

2.2 Screen

The 2.2k matte 60Hz 100%sRGB screen is not the best among premium laptops but I don’t have much complaint as well.

2.3 Laptop disassembly

You can remove the cover with just 5 screws and fully dissemble the laptop with the only screwdriver Framework supplies, which is pretty nice, but what’s wrong with the bottom-right screw??? The other four screws are really nice, they hang around when unscrewed so you won’t lose them but the bottom-right one is pressing against the chassis and won’t loosen itself. This is pretty problematic because due to mechanical pressure, the bottom-right screw often screw in half a turn by itself. When I finish unscrewing these 5 screws I usually lift the top-left (near the Framework logo) and top-right (near the power button), then the bottom-right screw grabbed and almost damaged the screw and probably the input cover. I often need to place it on my lap, use my left hand to lift the cover and the right hand to unscrew the screw again.

Comparison

3. The bad

3.1 Speakers

The sound from the speakers feels quite a bit muffled, likely caused by downward-firing.

3.2 Durability on some components (non-critical)

The bezel started to break after 2 years of use.

Photos

There are also reports of other parts such as the chassis. I think the card slots could use some fillets.

4. The Ugly

4.1 Bad battery longevity

There are numerous reports of battery swelling even after the “battery extender” BIOS update. My battery health has declined to 82% after 2.5 years of use despite I seldom charge it over 80%. Both Texas Instruments (the BMS manufacturer) and ATL (battery cells manufacturer) are reputable brands so what went wrong? I concluded that the battery calibration could use some work.

Problem 1: Insufficient accuracy of battery model.
Modern BMSes use Kalman filter to continuously tracking the battery’s state-of-charge (SOC) and state-of-health (SOH). However, a detailed model is required. If the model loaded into the BMS is not accurate enough, errors may accumulate.

Technical details


This is a typical lumped-element equivalent circuit model. The V_oc (SoC) is the open circuit voltage at a given SOC, commonly referred as “discharge curve” (under a very low load). This curve is given by the battery chemistry and doesn’t change as the battery ages. The R0 is the “AC impedance” with polarization ignored. The RC components R1 and C1 are the model of “mass transport” or “polarization” voltage drops. When the battery is heavily loaded, additional electrical potential is caused by ions not moving fast enough to catch up, causing a difference in concentration.
The R0 and R1 changes according to battery age and operating temperature. The internal resistance may also increase slightly if the battery is near empty or near full. Either there’s a multi variable function that the BMS updates its coefficients or a [temp,soh,soc] vector is loaded to the BMS. This makes the BMS being able to differentiate between voltage drop caused by battery discharge (SOC change) and voltage drop caused by internal resistance/impedance.

I have witnessed up to 4% “jumps” of battery % and battery health after computer off overnight. Other brands only have this when stored for a long time or used under extreme temperatures.

Problem 2: Charge termination current is too low.
Unlike Lead-Acid or NiMH, Li-ion batteries can’t be trickle charged. The charging method is typically constant-current constant voltage (CC-CV) some models have multi-staged constant current (MSCC-CV). The CV phase are the same, when the voltage has reached the limit, a constant voltage is applied until the current reaches the termination current. Then the current is turned off and the battery is charged. Continue holding that voltage, even if the limit voltage is not exceeded, may cause premature wear. On the Framework laptop, it could take 30 minutes to up to 2 hours to charge from 99% to 100%. At first I thought it was battery calibration, but after a discharge-charge cycle, the charging from 99% to 100% is still very long.


The “Consumption” shows charging power when the charger is plugged in. As shown, the termination current is too low, most other brands finish charging at around 5W.

Problem 3: Inaccurate(?) temperature bracket.
The charging speed slows way down when the temperature is below 20C, but the battery charges at full speed to up to 55C. The bracket could be moved down to 10-45C since charging at high temperatures is not good for battery health.

Other models

I currently prefer FL13 because:
FL12: 69% sRGB. FL13Pro: RAMpocalypse. FL16: power management issues causing CPU locked to 544MHz, GPU TGP to 30W, battery dischargeing even with 240W charger. Desktop: I’m not a heavy AI user.

Rating: Generally good

Despite the wall of text in “4. The ugly” section, it’s NOT a dealbraker. I only happened to write it because I somewhat understand some battery charging techniques. Please consider all the points equal weight i.e. 6 goods, 3 neutrals, 2 bads and 1 ugly. I may consider 1.4 double weight but you are entitied to your own opinions.

5 Likes

Very interesting review, thank you for sharing.

Based on your information on the battery issues, it sounds like you think there is a reasonable chance that battery lifetime could be improved with a firmware update to fix the BMS? Or would it required a replacement of the BMS?

It’s a BMS, not EC issue.
The EC is only responsible for reading the battery % and battery health from the BMS, the BMS is responsible of estimating the battery % (do the Kalman filter calculation) and reports to the EC.
The EC can’t even read the “charge termination current”. The BMS tells the EC its status, desired voltage and desired current. When the BMS deems the battery is full, both the desired voltage and desired current will change to 0.
Unlike the EC mentioned in “The good” paragraph 1.3, the BMS can’t be tuned. Parameters like the voltage curve, impedance curve, temperature range and charge termination current are locked behind a password that only the manufacturer knows. My best bet is Framework recognizing the problem and editing the BMS parameters for future battery batches. Framework might already did that since my computer was bought in 2024 but I’m not 100% sure.

1 Like

The ec is the one controlling the charge controller which is what could/would do the charge termination current thing.

Definitely agree on the batteries estimated state of charge being very jumpy but that may just be because I usually don’t get as much insight and multiple readjustments per second even when not being charged/discharged are normal.

Yeah can’t easily ota update the batteries bms, that would be a safety nightmare. Charge termination could be done on the ec though, especially when going anywhere near 100% soc that would probably be a pretty good idea to back off a bit early when charging to such high cell-voltages. At lower charge limits float charging is much less of a problem.

Sorry I guess I spent too much words on the battery so it went to another battery discussion thread again, probably also caused by recent “battery bloated” topics being posted by other users. Should’ve spent more time appreciating the BIOS transparency.

The EC controls the power electronics i.e. the buck-boost converter, but it’s the BMS who tells the EC how much the voltage and current should be supplied to the battery. You can terminate the current early by, well, setting a charge limit. For example if you set the limit to 80% the voltage will still be very close to 17.6V but the current will suddenly turn off, making the voltage to drop to 16.6V. If you set the same voltage limit but termination current as high as 0.4C to the BMS, the result would be the same. However a very high termination current can also cause inaccurate battery charge and health reading because of larger voltage drop caused by internal resistance.

Kinda

I think you have your tail wagging the dog here, the charge limit is achieved by varying the output voltage which in turn affects the charge current. The charge controller really only has one big lever to modulate charging, and input current limiting and that is controlling the output voltage (at can also turn of charging entirely but not discharging and if we had that could bypass the buck/boost converter). It can raise vbus above vbat which would cause charging proportional to the voltage difference or let vbus drop below vbat which would cause discharging proportional to the voltage difference.

The battery itself has no way to control termination current, yes the bms can disconnect both charging and discharging but that is not something it does on a whim.