The Framework laptop is 1mm thicker and 200 grams heavier than the 9th gen Carbon. We've gone long past the point of diminishing returns when it comes to size/weight vs repairability trade offs.

200g is 15%.

It is, however, I'm lucky enough to be in good health that carrying 200g extra in my backpack has no negative implications vs the added benefits.

But only about the weight of a box of teabags extra: https://weightofstuff.com/10-household-items-that-weigh-abou...

Is that a property of the components being soldered in, or just a property of the fact that Framework cannot get access to the highest quality components on the market?

Genuine question, I know very little about electrical components.

With RAM, the problem is more that high capacity LPDDR4 modules simply aren't available on SODIMMs.

However this is only because manufacturers don't make them. If Apple asked Micron or Samsung for them, I'm sure it would happen.

A better argument for non-replaceable RAM can be found in the Apple Silicon chips. Building the memory into the SoC provides very tangible performance and efficiency benefits.

>With RAM, the problem is more that high capacity LPDDR4 modules simply aren't available on SODIMMs.

And I don't think they ever will be. From the little bit that I've read, the higher voltage that SODIMMs have to use has to do with noise in transmission. LPDDR4s have been connection through the direct solder, so are able to use lower voltages.

I'm not an electrical engineer but I have a hard time believing we can't design a socket that provides a connection as stable and free from interference as a solder joint. It could mean making a PGA or LGA socket similar to what we use for CPUs, but it should still be perfectly doable.

You'll note that the Framework does not have a socketed CPU, either. Because that would be a disaster.

the mounting hardware needed for replaceable dimms inherently takes more space on and above the board. the solder approach also gives more flexibility for board layout, since you don't have to reserve space for the exact size and shape of a standard module. see a teardown of the new blade 14 to see how this can be beneficial.

there is an inherent tradeoff between size, battery life, and modularity. if you can make a modular laptop with good specs and battery life, a competitor will always be able to offer the same thing in a smaller chassis or with a bigger battery.

They have to use DDR4-3200 to be able to put the RAM on a stick. Integrated systems can use LPDDR4x-4267. You can't put that on a stick. It's a trade-off. It turns out it takes extra power to drive high-speed signals across long traces with connectors.

I thought the power advantage mostly came from the total bus being half as wide (32-bit LPDDR4X versus 64-bit DDR4).

Do you have a reference for why it can’t be put on a removable module?

I'm fairly sure the power advantage comes from VDDQ being much lower: .6V for LPDDR4x and 1.2V for DDR4. LPDDR4x is or at least can be 64 bits wide, just like DDR4.

That's fair enough. I guess I just hope that we figure out a way to overcome this I suppose.

What performance/cost impact are we talking about?

Judging by reviews, the Framework loses 10% of CPU+GPU performance versus reference designs. That could be due to their memory subsystem or thermal design, or both. Ars Technica said the battery life was "mediocre" but I would have gone with "terrible". Compared to the Dell XPS 13 the Framework has only 60% the life.

As arstechnica noted, the XPS 13 is not a great comparison point because the display is darker and has a lower resolution.