I know Framework somewhat have their hands tied by modularity, but for the 13 I think the next thing they need to focus on more than anything is battery life. At this point it’s not just MacBooks that greatly outperform the FW13 on that front, but also several competing x86 laptops. For the price it’s difficult to justify taking such a steep hit, even with repairability factored in.
I guess this depends on how much you value repairability. To me it makes the difference between having a laptop or a desktop. A non-repairable laptop is inherently unreliable as you must return it to the manufacturer for any repairs and are without a computer for that time. That makes traditional, non-repairable, laptops something you get for mobility in addition to your desktop. Not only does this increase costs significantly but also that you need to worry about syncing when moving between the two.
Value of repairability depends on how often things break and how accessible shops are. In my case it’s pretty unusual, and I’m not often far from an Apple store.
It also depends on how much one values the qualities inherent to laptops. To me, a battery that burns through quickly is a major ding to its portability and extra overhead (needing to find outlets to sit near, having to carry a quick charging brick, etc) that I’d rather not deal with.
I think there’s probably a happy medium to be found in this situation. Power consumption could be reduced by switching RAM away from DIMM slots and over to CAMM, allowing usage of efficient LPDDR modules for example, and they could also offer an ultra-low-power soldered RAM mainboard option for those willing to trade off a little bit of repairability for battery life (I’d bet most people will never bump up their RAM before upgrading their mainboard anyway).
Whatever the case, repairability is something I value, but not so much that it overrules all the other qualities that make a laptop good. It’s one aspect of many.
I would like Framework to go the System76 route and adopt a Linux distro that they can very finely tune for the Framework laptop. Macbooks have the benefit of MacOS being completely vertically integrated. I don't see Framework putting this level of effort in, even on the OEM Windows side of things like other OEMs such as Dell do. Their software needs to be better.
I don't think this is what makes Apple jump ahead of the competition. M-series is a low-wattage SOC.
I recently bought a lunar lake laptop, which is a similar design but x86. I'm easily getting 12+ hours on Linux and performance is superb. Turns out all it takes is putting everything on the chip and then running it at 15 watts. Which, I think, also proves you do NOT need ARM for such low wattage with acceptable performance. You just need really new production lines and the right chip design.
Not entirely convinced its just hardware as we're seeing from the now direct comparisons between windows and steamos. [1] Linux based SteamOS is thrashing windows in both gaming performance and using less battery to do it.
I would agree it's not just hardware, software also plays a role. Tons of background services connected to the internet are going to drain battery like crazy, we learned this with mobile phones a decade ago.
However, I'm not convinced MacOS is much better than Windows here. I'm sure It's a little bit better though - I've never seen MacOS pinned at 100% CPU usage doing seemingly nothing.
Pennywise, pound foolish, unfortunately, but more than that, companies that do hardware and software well are few and far between, and the ones that do tend to be highly valued. AMD famously does value software much, or pay their software developers relatively well.
Is it pound foolish, or just a rational business decision?
Let's say they ship 100k laptops per year. Let's say they could meaningfully improve battery life with a team of half-dozen excellent software engineers, which would cost on the order of a few million a year. For the sake of argument, let's say ~$3M/yr. That increases the price per laptop by ~$30 on average. That's a premium I'd pay for improved efficiency, but judging by the comments here and elsewhere, the premium they're already charging above the raw component prices seems to be at the upper end of what most people are willing to pay.
It's fiendishly difficult to become the next Apple, Tesla, Nintendo, or Valve with thick enough margins on your hardware (or services) to afford excellent software engineering teams, so it makes sense that so few hardware companies attempt it, and many who try eventually give up.
It's a much more interesting question when framed with numbers! But let's say they ship 500k main boards (since it doesn't need to be a totally assembled laptop to benefit), and it only costs $300k, not $3 million (and a couple laptops) to the right eastern European software developer to perfectly tune some Linux config files. Then it's only $1.50 per laptop, and they could arguably just eat that cost.
Framework doesn't have to spend enough to be the next Apple (nor do they have the resources to be), they just need to spend enough to not be so desperately far behind Dell.
The explanation makes sense in isolation, it just seems like a local maxima if you zoom out.
Sure, it totally depends on specifics! Only Framework knows how many mainboards they sell per year and can make an educated guess at how many of them end up running Linux.
Also, note that Framework already employs at least one person[1] working full time on Linux compatibility and support, so at least some of the low-hanging fruit may have already been picked. I'm sure they could spend an additional $300k, $3M, $30M, or more on improving Linux efficiency. I can't estimate what the benefit would be at each of those levels, nor do I know what the price impact would be, nor the sales impact. I don't know what they currently spend on Linux support except that it's at least one FTE.
We don't have enough information to answer or even meaningfully estimate most of these questions. I'm not saying they're making good decisions or bad decisions with respect to Linux support, I'm just saying neither of us have enough information to know.
I think almost anyone on earth would pay an extra money per month for Apple level battery life on a modular, repairable, hardware up gradable, Linux box.
Apple at the moment has zero competitors for upscale laptops and this would make System76 the only other alternative for a quality hardware machine.
The rave reviews alone would be free marketing worth well above the money invested in the software engineers.
It would be the default goto box for a modern alternative to Mac much like Lenovo used to be a decade ago before MBA enshittification set in there.
Battery life is the ONE thing preventing myself and many others from pulling the trigger on a System76 and I would gladly pay much more above and beyond a macbook pro for an alternative to a macbook with equivalent battery life but linux.
Plus PopOs is open source so there could be cross pollination with the Linux team on battery life optimization which would reap massive benefits for the Linux ecosystem as a whole and push more people towards Linux.
Something like this would be myself and many other peoples literal dream computers and withing a year or two's time almost any Linux user would be on System76 laptops, guaranteed
Funny thing is, a lot of the System76 tweaks aren't specific to their system (or OS) at all. One popular example is the system76-scheduler, which you can install on pretty much any hardware or distro for the same responsiveness improvements: https://github.com/pop-os/system76-scheduler
Kinda leads me to believe the whole "vertically integrate my Framework" shtick is a snipe hunt.
You may be misunderstanding me or I'm not explaining myself properly. I don't want them to vertically integrate in the same way that Apple does, I want them to invest more on the software side by selecting a distro and building around it. If they can piggyback on popOS then great but they need to invest in software.
Realistic starting point: hire one Linux developer to daily drive Debian Testing + mainline Linux kernel on Framework hardware, then upstream integration/optimization fixes to mainline Linux and Debian unstable.
Upstream fixes would benefit multiple Linux distros, reduce Framework support burden and increase the usability of Linux on Framework hardware.
I don't see how that would help, and in a lot of ways I feel like PopOS is an example of how phyrric the effort is. They're spending hundreds of thousands of dollars to support a distro that most users will probably replace with something else. Really all they have to do is ship a Debian/Fedora image as default and test the hardware config before shipping it so people have a level of QC to depend on. Building and maintaining an OS from scratch is a baby+bathwater solution to this, at least from where I'm standing.
What kind of problems do you anticipate this would fix?
It’s technically also covered by Mint and kinda (not as well) Ubuntu, but one thing that pop gets right in my view is bundling in the Nvidia drivers that a huge chunk of people are going to need, as well as enabling non-free repos by default, neither of which Fedora or Debian do (not to mention, a lot of users will find Debian’s user-facing packages too old [yes, even with Testing]). I say this even as someone who generally doesn’t use pop and favors Fedora.
PopOS is not wasted effort. The goal of PopOS is to have an out of the box Linux aimed at people new to Linux that has everything working out of the box. Specifically, graphics drivers working out of the box which is notoriously hard if you are running an Nvidia card. According to the Steam Hardware survey, it is 10th on the top 10 Linux distros[0]. Realistically higher when you consider that the Steam Deck and SteamOS heavily tilt the survey. I'm not asking them to build an OS from scratch and that is a crazy way to interpret what I said which was "build around a distro"
There isn't anything they can do about it while maintaining their allegiance to the SO-DIMM. Fundamentally, memory on a stick guzzles energy. The reality of energy efficiency and the myth of upgradeability are conflicting.
Indeed, indeed it does solve that problem, but other actors need to align. Someone would have to ship a CPU that people want that supports off-world LPDDR (i.e. not Intel Lunar/Arrow Lake), and laptop makes would need to adopt it.
My suspicion is what will actually happen is that CAMM2 is going to make inroads in desktop systems.
Seems like you're not aware of AMD's recent APUs, which seem to be very capable and very popular. I'm not sure if a laptop model sporting one with CAMM2 ram is available yet, but one can safely assume it's just a matter of time.
I'm also not sure what you mean by other actors needing to align. JEDEC has standardized CAMM2 already. Which is how all concerned actors accomplish alignment.
I am aware of them but as you pointed out they don't exist with CAMM. The one that I want is the Ryzen AI Max and all the implementations of it so far — the Framework Desktop, the HP Z2 G1a — have the memory soldered down without CAMMs.
I seem to remember executives from Framework being active in the HN comments at the Framework Desktop launch in which CAMM2 support was a highly requested feature. I suspect they received the message.
Lovely thing about the PC industry which differs from our friends in Cupertino is that it tends to explore the full design space over time. All good things come to those who wait.
I suppose the main thing keeping me from being interested in this thing is that every 2-in-1 convertible I've tried in the past was *heavy* and *cumbersome* to use as an actual tablet. I wonder how this holds up, but I'm not sure it's possible to fix the main issue: the keyboard makes it real awkward to hold. It's like a whole product class that looks amazing in marketing shots and is kindof a pain to use in any of the non-laptop modes in practice. Am I just holding it wrong?
No, I don't think you're doing it wrong. I think the microsoft surfacebook design, where it could be used like a normal laptop with the keyboard attached, or like a tablet with the screen detached was the only design that could really do both well, and it had its own issues.
The regular surface devices (and ipads and android tablets) that are tablet-devices first with flimsy detachable keyboards are fine if you have a table to set them on, but difficult to use on your lap, and often have a mediocre typing experience.
And, on the flip side, you get devices like this where the keyboard stays attached and folds around behind the screen. It can be good for certain use cases, but it's clearly meant to be a laptop first, and it's "tablet mode" is inevitably going to be more cumbersome than a "real" tablet.
My favorite 2nd device is an old Dell Inspiron 11 2-in-1. By all accounts it's a terribly specced device, but it still runs modern Fedora + KDE well, the touchscreen works well, and the display quality isn't awful.
Many of the knocks against it (small size, cheap plastic case, small battery) work in its favor by keeping it light enough to use as a tablet. The battery, RAM, and SSD are all serviceable. I see them on eBay frequently for as little as $20-40 each in lots or $50-70 standalone.
The 10" Lenovo Duet Chromebook tablets are a close second but don't age as well and can't be realistically repaired or upgraded.
I think it's a shame they went with 2 year old 13th gen Intel processors for these. The newer Lunar Lake Intel chips have much better battery life and thermals, and you can find them in other similarly priced laptops. The main downside is that Lunar Lake chips have soldered on RAM, but given that Framework already sells a desktop with soldered RAM that shouldn't be a dealbreaker for them.
For their marketing it is. There is a very specific reason the desktop exists. On-premise AI workloads. You can't get the bus width on socketed ram that you do soldered ram at the current moment. 1 product is an exception, 2 would be a break from their replaceable component marketing.
It used to be that CPU and GPU were separate cards/chips, nowadays they are often integrated. There are good reasons to solder-in RAM. Yes it's a tradeoff, but it's reasonable. I don't expect FW to revert technical decisions that are not under their control. Mostly, I just don't want to be buying whole new laptop because my keyboard had a defect, or I cracked my screen. If the compute module comes with a soldered RAM, it's perfectly fine with me. For most most of my systems I buy it with RAM maxed-out, and when I get a new one, there's typically new generation of DDR, new speeds, etc. and I will not reuse my old RAM. I'm fine with buying and replacing a new CPU + RAM module as a unit, and I would buy such a module. The fact that there are modules with external memory is already enough good faith commitment to repairability.
From your careful word choice you know this but I'd like to draw attention to the specifics here. I'm not sure if it's actually soldered on but RAM is on-package like apple's M-series chips, unlike what used to be the case where soldering ram to the board was a choice made by the laptop mfgs.
Here's there's no choice to be made other than not using the chips. And unfortunately (although there are some benefits), it's probably not going to be just a few generations but a trend for high end processors going forward.
It is actually one of the few cases where I don't actually really care about independent upgradability. In my experience I find that I pretty much always upgrade my CPU and my RAM in tandem. New CPU architectures sometimes force it (e.g. need DDR5 instead of DDR4), and as long as you don't severely undersize your initial RAM choice I find that I run out of CPU headroom before I run out of RAM headroom.
So if there's performance gains to be had by co-locating RAM with the CPU in a single package, it makes sense to me to do so
> and as long as you don't severely undersize your initial RAM choice
That's the problem though. when dealing with used machines (because new ones are beyond your budget), you get cheaper hand me downs, and those are going to be of your undersized RAM variety. In the socketed days, you could get a five year old laptop, replace the existing RAM with the biggest sticks you cloud get your hands on, and get a few more years of life out of the machine. A laptop stuck at four gigs of ram these days isn't going to be great for much web browsing, but is also basically stuck at four gigs.
Unless Foundation always uses the current generation CPU, then this complaint is always valid. Or maybe invalid.
Thermals and energy consumption are almost always improving between generations. It's hard for me to think of 13th generation as old. Maybe I'm getting old!
The reason why I brought it up was that Lunar Lake was specifically designed to make an x86 offering that has comparable battery life with chips from Qualcomm and Apple. It's not a standard marginal generational improvement, for light office tasks (the type of use case this laptop is intended for) you get around 17 hours of battery life vs the previous generation's 10 hours (see benchmark results here: https://www.pcworld.com/article/2463714/tested-intels-lunar-... ).
I wanted to understand if any other cpu generations had comparable gains over the previous, but I need to work on an Intel ark scraper. Maybe I'll find some time to post it back here on HN. Thanks for added details
It's not surprising, Meteor Lake and onwards are all TSMC and unlike Intel they still actually know how to manufacture chips. Arc iGPUs are also a massive improvement over Iris.
It's not just a TSMC vs. Intel thing, Lunar Lake gets 7 more hours of battery life than Meteor Lake when doing the same "light office tasks" benchmark.
The numbers I've seen show that the peak power draw is similar to Meteor Lake, but it generally draws ~40% less than Meteor Lake when doing the same task. The SoC only idle power draw figures I've seen are 0.62W for Lunar Lake vs. 2.32W for Meteor Lake.
Yeah I complained about this on the framework subreddit when I first heard of the FW12, asking for reasons and expectedly got downvoted into oblivion. Apparently expecting them to use something mildly recent is too much to ask for and it's supposed to be a cheapshit student laptop aimed at schools or whatever. Wouldn't be as bad if it weren't Intel's worst CPU gen in recent memory, FW probably got a bargain box deal on them.
Meanwhile the HX370 is still warm from the oven and already in the FW13, linux drivers aren't even ready yet with compatibility complaints aplenty. Not to mention the FW Desktop with the AI Max which was the first launch of that chipset worldwide.
I ran XPS Developer Edition (Ubuntu) for 5+ years without any issues.
Using Framework for past year - every day I have to reboot because it freezes with 20+ Firefox tabs (Ubuntu 22.04, AMD). Tried all options (disable vGPU etc) but no luck.
That's frustrating, I'm sorry! I don't experience anything like that, and I have over a thousand tabs open in Firefox on NixOS.
Based on https://frame.work/linux Ubuntu 24.04 is the minimum supported version for the older AMD F13's, so I'd suggest updating and then reaching out to Framework if the problem persists on a supported distro.
on chrome based browsers with Linux I have to disable 'use graphic acceleration when available' in the browser settings or the browser with freeze intermittently. This has been a consistent issue across distros for years. Not sure if it also affects firefox.
Disabling this feature mostly works, but results in poor performance for some graphics heavy websites.
I also had a lot of issues with my AMD framework laptop and ended up reverting back to a older Intel Framework laptop. Top issues with the AMD laptop were the realtek wireless, and random AMD (integrated) GPU glitches. Intel hardware continues to be absolutely top tier for Linux support.
> on chrome based browsers with Linux I have to disable 'use graphic acceleration when available' in the browser settings or the browser with freeze intermittently. This has been a consistent issue across distros for years. Not sure if it also affects firefox.
Oddly enough works fine on all Chromebooks. It really is a matter of platform hardware/firmware qualification being suited to task.
Yep, my last laptop purchase was an xps13. Unfortunately the 8gb of soldered on ram is becoming woefully insufficient with how bloated the modern web has gotten.
I have a beefy desktop, but if I replace my laptop I think it will probably be a thinkpad.
People make fun of Chrome being a RAM hog but I'm having more issues with FF than I ever did with Chrome. For some reason on reddit specifically FF will randomly freeze for a few seconds, my cursor will stop rendering on top of it (like it's going behind the window) and it won't accept inputs. If I keep typing through it it all appears in the window after it unfreezes. Also FF will regularly take multiple gigs for random tabs like Youtube if I leave them open.
Tldr it might be more of a firefox problem than a framework problem.
I would hazard a guess that it’s more of a “devs only test against Chromium” problem than it is a Firefox problem. It’s a problem seen under WebKit-based browsers at times, too. Gecko and WebKit often behave differently and have different performance characteristics than Chromium/Blink does, but that’s often not accounted for at all. The extent of QA on non-Chromium browsers too often stops at “it technically runs”.
That's also probably part of the story but ultimately as an end user the fact is Firefox is a bad experience I'm suffering through only to not use Chrome. I can't force websites to patch whatever memory leak is causing Youtube tabs on FF to eat 5 GB of ram for example.
I have a ThinkPad T480s with Ubuntu and Firefox. Works like a dream except every now and again the whole thing locks up (have to power cycle). Happens with AWS Console and LinkedIn. Maybe it's doing me a favour.
I haven't had any issues with Wifi. A family member replaced theirs with an Intel chip to fix an issue; turns out reception was just bad and not the wifi chip.
That brings me to my second point: if the wifi chip's what's holding you back there's nothing stopping you from replacing it.
Wifi worked out of the box for me on FW13 both the AMD 7840U and the HX 370 mainboards, in Ubuntu 24.04 LTS and 25.04. Ethernet via USB-C dock worked out of the box as well.
I suspect it's actually quite difficult to build an entirely modular laptop to work with several operating systems while supporting the latest components, and be void of any driver issues.
I was seriously considering getting a Framework laptop (but 13") and what stopped me is the limited shipping. Right now I live in the area where I can order one, but I was looking at another country where I'd prefer to live. It won't happen next month or even this year. But if I order a Framework laptop I want to have access to the new and replacement parts and Framework doesn't ship there. I don't want to rely on forwarders. And that's a deal breaker. So I went with another Thinkpad instead.
And on the subject of 12" laptop, when it was teased I hoped for a MacBook Air killer. But that didn't happen. I like what Framework is doing, but common manufacturers are still looking like a better value.
I’ve been really curious about some of the hybrid handhelds / micro laptops from GPD lately. Framework should consider looking into this market niche as well.
I love my GPD Win mini, and I've had a few other GPD products before that, but I'm not sure if a framework-style modular design would work as well there.
For starters, they've always gone with soldered RAM for both physical size and performance reasons (RAM speed matters more with integrated graphics, and soldered RAM can go faster that socketed RAM.)
Additionally, I don't think there's enough space for a reasonable number of ports via expansion cards. Even in the larger 8-inch-screen models, I think 4 expansion slots would be very difficult, whereas the current 7-inch models have 5-8 ports. (I think some of the MicroPC models had a handful of fixed ports + one modular expansion slot, so maybe they could change that one to a framework-compatible design?)
To their credit, GPD has offered motherboard upgrades in the past when a refreshed design is otherwise compatible. But there are often changes in ports or layout or cooling design that make that impossible.
I'm glad Framework exists but I don't envy them. Selling laptops aimed at the most nitpicky never-satisfied group of people must make it hard to see through the noise. I wish them the best.
Is my framework perfect? No. But it's still the best laptop I've ever owned. Just like open source software, it's not about having the most shiny look and feel, but it's about knowing that the creator wants me to have the freedom and empowerment to open it up and adjust it to my needs and desires. Other companies will just say that I've voided the warranty when I try fixing or tweaking it myself.
Maybe these users are picky, but they are under-served, very influential, and very productive w.r.t. improving support for the laptop in the ecosystem. At this point FW became a go-to laptop for Linux hackers, and it gets great support in all avant-garde tech like e.g. NixOS.
There is a certain rich racingdriver/programmer/macuser who moved his whole company to framework laptops running linux. So they must be doing something right
I don't know who you are talking about but I'm not implying that Framework is doing anything wrong, just that they have chosen a very difficult user base to cater to.
I'm actually curious how much difficulty they have with their actual customers. There are always a ton of negative comments in these threads, but those seem to be from people who just fundamentally value different things than what framework is tyring to do (which is totally fine and valid!), and therefore very specifically aren't the customer base. I can't speak more broadly, but I've been super happy with my Framework16, warts and all (and there are a few).
My guess is that the core audience of people who want a repairable, upgradeable laptop tend to be tinkerers and don't mind a little bit of extra effort.
It's a combination of the compact, 12" plastic shell with the soft edges and rounded corners, near edge-to-edge keyboard and a soft two-tone colour palette. I only wish that it had a more square-ish, 4:3 screen like the iBook.
People have a lot of pretty valid issues with Framework. When you optimize as hard as they do for repairability, upgradeability, and modularity, you are going to make some sacrifices in other arenas that won't be to everyone's liking, and that's totally fine. But the one criticism that I haven't quite understood is price.
Yes, Framework is more expensive....the first time. But from then on (and for the 13, this is now 3 or 4 generations I believe), buying a new mainboard with a latest gen chip is going to be the cheapest you can get those specs for. Yeah, a used laptop might still be cheaper, but it won't be the same specs, and if used is what you want, you can probably get a used mainboard from the last gen that someone else is upgrading from for even cheaper.
Unless you are someone who thrashes their laptops so hard that you expect to need to replace not just the mainboard, but also the screen, keyboard and chasis, then framework represents by far the cheapest way to keep getting new laptop specs into the future.
I'm not even someone who upgrades laptops very often (every 5 years or so, historically), and even on that schedule, and I still thing that long term it will be cheaper. If you are someone who upgrades more often, then it makes an even bigger difference.
Again, to be clear, I think there are very valid reasons that a Framework might not be the right laptop for someone. I just thing that thinking about the one-time purchase price of the entire laptop is completely missing the entire point of Framework.
The more videos with the FW12 moving and used before the Camera I see, the less I can ignore the fat bezels. The design language at all is not made for „business“ which is refreshing and they obviously have a budget approach, but such ancient bezels don‘t do a contribution for anything. The lack of any Windows Hello enabling hardware was the final bit for my sad no-buy decision.
I know Framework somewhat have their hands tied by modularity, but for the 13 I think the next thing they need to focus on more than anything is battery life. At this point it’s not just MacBooks that greatly outperform the FW13 on that front, but also several competing x86 laptops. For the price it’s difficult to justify taking such a steep hit, even with repairability factored in.
I guess this depends on how much you value repairability. To me it makes the difference between having a laptop or a desktop. A non-repairable laptop is inherently unreliable as you must return it to the manufacturer for any repairs and are without a computer for that time. That makes traditional, non-repairable, laptops something you get for mobility in addition to your desktop. Not only does this increase costs significantly but also that you need to worry about syncing when moving between the two.
Value of repairability depends on how often things break and how accessible shops are. In my case it’s pretty unusual, and I’m not often far from an Apple store.
It also depends on how much one values the qualities inherent to laptops. To me, a battery that burns through quickly is a major ding to its portability and extra overhead (needing to find outlets to sit near, having to carry a quick charging brick, etc) that I’d rather not deal with.
I think there’s probably a happy medium to be found in this situation. Power consumption could be reduced by switching RAM away from DIMM slots and over to CAMM, allowing usage of efficient LPDDR modules for example, and they could also offer an ultra-low-power soldered RAM mainboard option for those willing to trade off a little bit of repairability for battery life (I’d bet most people will never bump up their RAM before upgrading their mainboard anyway).
Whatever the case, repairability is something I value, but not so much that it overrules all the other qualities that make a laptop good. It’s one aspect of many.
What if they let you swap your CPU for a M series chip?
"they" being Apple, presumably? I see no reason you couldn't attempt an M1 board swap aside from the irony of it.
Being Framework who offer a laptop where you can swap components out.
I would like Framework to go the System76 route and adopt a Linux distro that they can very finely tune for the Framework laptop. Macbooks have the benefit of MacOS being completely vertically integrated. I don't see Framework putting this level of effort in, even on the OEM Windows side of things like other OEMs such as Dell do. Their software needs to be better.
I don't think this is what makes Apple jump ahead of the competition. M-series is a low-wattage SOC.
I recently bought a lunar lake laptop, which is a similar design but x86. I'm easily getting 12+ hours on Linux and performance is superb. Turns out all it takes is putting everything on the chip and then running it at 15 watts. Which, I think, also proves you do NOT need ARM for such low wattage with acceptable performance. You just need really new production lines and the right chip design.
Not entirely convinced its just hardware as we're seeing from the now direct comparisons between windows and steamos. [1] Linux based SteamOS is thrashing windows in both gaming performance and using less battery to do it.
https://www.techspot.com/news/108059-steamos-significantly-i...
I would agree it's not just hardware, software also plays a role. Tons of background services connected to the internet are going to drain battery like crazy, we learned this with mobile phones a decade ago.
However, I'm not convinced MacOS is much better than Windows here. I'm sure It's a little bit better though - I've never seen MacOS pinned at 100% CPU usage doing seemingly nothing.
It’d be nice, but keeping a team of engineers capable of that sort of low-level work can’t be cheap, so it seems unlikely.
Pennywise, pound foolish, unfortunately, but more than that, companies that do hardware and software well are few and far between, and the ones that do tend to be highly valued. AMD famously does value software much, or pay their software developers relatively well.
Is it pound foolish, or just a rational business decision?
Let's say they ship 100k laptops per year. Let's say they could meaningfully improve battery life with a team of half-dozen excellent software engineers, which would cost on the order of a few million a year. For the sake of argument, let's say ~$3M/yr. That increases the price per laptop by ~$30 on average. That's a premium I'd pay for improved efficiency, but judging by the comments here and elsewhere, the premium they're already charging above the raw component prices seems to be at the upper end of what most people are willing to pay.
It's fiendishly difficult to become the next Apple, Tesla, Nintendo, or Valve with thick enough margins on your hardware (or services) to afford excellent software engineering teams, so it makes sense that so few hardware companies attempt it, and many who try eventually give up.
It's a much more interesting question when framed with numbers! But let's say they ship 500k main boards (since it doesn't need to be a totally assembled laptop to benefit), and it only costs $300k, not $3 million (and a couple laptops) to the right eastern European software developer to perfectly tune some Linux config files. Then it's only $1.50 per laptop, and they could arguably just eat that cost.
Framework doesn't have to spend enough to be the next Apple (nor do they have the resources to be), they just need to spend enough to not be so desperately far behind Dell.
The explanation makes sense in isolation, it just seems like a local maxima if you zoom out.
Sure, it totally depends on specifics! Only Framework knows how many mainboards they sell per year and can make an educated guess at how many of them end up running Linux.
Also, note that Framework already employs at least one person[1] working full time on Linux compatibility and support, so at least some of the low-hanging fruit may have already been picked. I'm sure they could spend an additional $300k, $3M, $30M, or more on improving Linux efficiency. I can't estimate what the benefit would be at each of those levels, nor do I know what the price impact would be, nor the sales impact. I don't know what they currently spend on Linux support except that it's at least one FTE.
We don't have enough information to answer or even meaningfully estimate most of these questions. I'm not saying they're making good decisions or bad decisions with respect to Linux support, I'm just saying neither of us have enough information to know.
[1] https://matthartley.com who was previously at System 76
I think almost anyone on earth would pay an extra money per month for Apple level battery life on a modular, repairable, hardware up gradable, Linux box.
Apple at the moment has zero competitors for upscale laptops and this would make System76 the only other alternative for a quality hardware machine.
The rave reviews alone would be free marketing worth well above the money invested in the software engineers.
It would be the default goto box for a modern alternative to Mac much like Lenovo used to be a decade ago before MBA enshittification set in there.
Battery life is the ONE thing preventing myself and many others from pulling the trigger on a System76 and I would gladly pay much more above and beyond a macbook pro for an alternative to a macbook with equivalent battery life but linux.
Plus PopOs is open source so there could be cross pollination with the Linux team on battery life optimization which would reap massive benefits for the Linux ecosystem as a whole and push more people towards Linux.
Something like this would be myself and many other peoples literal dream computers and withing a year or two's time almost any Linux user would be on System76 laptops, guaranteed
Funny thing is, a lot of the System76 tweaks aren't specific to their system (or OS) at all. One popular example is the system76-scheduler, which you can install on pretty much any hardware or distro for the same responsiveness improvements: https://github.com/pop-os/system76-scheduler
Kinda leads me to believe the whole "vertically integrate my Framework" shtick is a snipe hunt.
You may be misunderstanding me or I'm not explaining myself properly. I don't want them to vertically integrate in the same way that Apple does, I want them to invest more on the software side by selecting a distro and building around it. If they can piggyback on popOS then great but they need to invest in software.
Realistic starting point: hire one Linux developer to daily drive Debian Testing + mainline Linux kernel on Framework hardware, then upstream integration/optimization fixes to mainline Linux and Debian unstable.
Upstream fixes would benefit multiple Linux distros, reduce Framework support burden and increase the usability of Linux on Framework hardware.
I don't see how that would help, and in a lot of ways I feel like PopOS is an example of how phyrric the effort is. They're spending hundreds of thousands of dollars to support a distro that most users will probably replace with something else. Really all they have to do is ship a Debian/Fedora image as default and test the hardware config before shipping it so people have a level of QC to depend on. Building and maintaining an OS from scratch is a baby+bathwater solution to this, at least from where I'm standing.
What kind of problems do you anticipate this would fix?
It’s technically also covered by Mint and kinda (not as well) Ubuntu, but one thing that pop gets right in my view is bundling in the Nvidia drivers that a huge chunk of people are going to need, as well as enabling non-free repos by default, neither of which Fedora or Debian do (not to mention, a lot of users will find Debian’s user-facing packages too old [yes, even with Testing]). I say this even as someone who generally doesn’t use pop and favors Fedora.
Framework doesn’t use or support any nvidia hardware, yet.
PopOS is not wasted effort. The goal of PopOS is to have an out of the box Linux aimed at people new to Linux that has everything working out of the box. Specifically, graphics drivers working out of the box which is notoriously hard if you are running an Nvidia card. According to the Steam Hardware survey, it is 10th on the top 10 Linux distros[0]. Realistically higher when you consider that the Steam Deck and SteamOS heavily tilt the survey. I'm not asking them to build an OS from scratch and that is a crazy way to interpret what I said which was "build around a distro"
[0] https://store.steampowered.com/hwsurvey/steam-hardware-softw...
There isn't anything they can do about it while maintaining their allegiance to the SO-DIMM. Fundamentally, memory on a stick guzzles energy. The reality of energy efficiency and the myth of upgradeability are conflicting.
Just one of the issues addressed by https://en.m.wikipedia.org/wiki/CAMM_(memory_module)
Indeed, indeed it does solve that problem, but other actors need to align. Someone would have to ship a CPU that people want that supports off-world LPDDR (i.e. not Intel Lunar/Arrow Lake), and laptop makes would need to adopt it.
My suspicion is what will actually happen is that CAMM2 is going to make inroads in desktop systems.
Seems like you're not aware of AMD's recent APUs, which seem to be very capable and very popular. I'm not sure if a laptop model sporting one with CAMM2 ram is available yet, but one can safely assume it's just a matter of time.
I'm also not sure what you mean by other actors needing to align. JEDEC has standardized CAMM2 already. Which is how all concerned actors accomplish alignment.
I am aware of them but as you pointed out they don't exist with CAMM. The one that I want is the Ryzen AI Max and all the implementations of it so far — the Framework Desktop, the HP Z2 G1a — have the memory soldered down without CAMMs.
I seem to remember executives from Framework being active in the HN comments at the Framework Desktop launch in which CAMM2 support was a highly requested feature. I suspect they received the message.
Lovely thing about the PC industry which differs from our friends in Cupertino is that it tends to explore the full design space over time. All good things come to those who wait.
I suppose the main thing keeping me from being interested in this thing is that every 2-in-1 convertible I've tried in the past was *heavy* and *cumbersome* to use as an actual tablet. I wonder how this holds up, but I'm not sure it's possible to fix the main issue: the keyboard makes it real awkward to hold. It's like a whole product class that looks amazing in marketing shots and is kindof a pain to use in any of the non-laptop modes in practice. Am I just holding it wrong?
No, I don't think you're doing it wrong. I think the microsoft surfacebook design, where it could be used like a normal laptop with the keyboard attached, or like a tablet with the screen detached was the only design that could really do both well, and it had its own issues.
The regular surface devices (and ipads and android tablets) that are tablet-devices first with flimsy detachable keyboards are fine if you have a table to set them on, but difficult to use on your lap, and often have a mediocre typing experience.
And, on the flip side, you get devices like this where the keyboard stays attached and folds around behind the screen. It can be good for certain use cases, but it's clearly meant to be a laptop first, and it's "tablet mode" is inevitably going to be more cumbersome than a "real" tablet.
I doubt there's any perfect solution, TBH.
My favorite 2nd device is an old Dell Inspiron 11 2-in-1. By all accounts it's a terribly specced device, but it still runs modern Fedora + KDE well, the touchscreen works well, and the display quality isn't awful.
Many of the knocks against it (small size, cheap plastic case, small battery) work in its favor by keeping it light enough to use as a tablet. The battery, RAM, and SSD are all serviceable. I see them on eBay frequently for as little as $20-40 each in lots or $50-70 standalone.
The 10" Lenovo Duet Chromebook tablets are a close second but don't age as well and can't be realistically repaired or upgraded.
It's annoying how reviewers focus on the performance specs while mostly ignoring the big differentiator of the 2-in-1 feature.
I think it's a shame they went with 2 year old 13th gen Intel processors for these. The newer Lunar Lake Intel chips have much better battery life and thermals, and you can find them in other similarly priced laptops. The main downside is that Lunar Lake chips have soldered on RAM, but given that Framework already sells a desktop with soldered RAM that shouldn't be a dealbreaker for them.
For their marketing it is. There is a very specific reason the desktop exists. On-premise AI workloads. You can't get the bus width on socketed ram that you do soldered ram at the current moment. 1 product is an exception, 2 would be a break from their replaceable component marketing.
It used to be that CPU and GPU were separate cards/chips, nowadays they are often integrated. There are good reasons to solder-in RAM. Yes it's a tradeoff, but it's reasonable. I don't expect FW to revert technical decisions that are not under their control. Mostly, I just don't want to be buying whole new laptop because my keyboard had a defect, or I cracked my screen. If the compute module comes with a soldered RAM, it's perfectly fine with me. For most most of my systems I buy it with RAM maxed-out, and when I get a new one, there's typically new generation of DDR, new speeds, etc. and I will not reuse my old RAM. I'm fine with buying and replacing a new CPU + RAM module as a unit, and I would buy such a module. The fact that there are modules with external memory is already enough good faith commitment to repairability.
From your careful word choice you know this but I'd like to draw attention to the specifics here. I'm not sure if it's actually soldered on but RAM is on-package like apple's M-series chips, unlike what used to be the case where soldering ram to the board was a choice made by the laptop mfgs.
Here's there's no choice to be made other than not using the chips. And unfortunately (although there are some benefits), it's probably not going to be just a few generations but a trend for high end processors going forward.
It is actually one of the few cases where I don't actually really care about independent upgradability. In my experience I find that I pretty much always upgrade my CPU and my RAM in tandem. New CPU architectures sometimes force it (e.g. need DDR5 instead of DDR4), and as long as you don't severely undersize your initial RAM choice I find that I run out of CPU headroom before I run out of RAM headroom.
So if there's performance gains to be had by co-locating RAM with the CPU in a single package, it makes sense to me to do so
> and as long as you don't severely undersize your initial RAM choice
That's the problem though. when dealing with used machines (because new ones are beyond your budget), you get cheaper hand me downs, and those are going to be of your undersized RAM variety. In the socketed days, you could get a five year old laptop, replace the existing RAM with the biggest sticks you cloud get your hands on, and get a few more years of life out of the machine. A laptop stuck at four gigs of ram these days isn't going to be great for much web browsing, but is also basically stuck at four gigs.
Unless Foundation always uses the current generation CPU, then this complaint is always valid. Or maybe invalid.
Thermals and energy consumption are almost always improving between generations. It's hard for me to think of 13th generation as old. Maybe I'm getting old!
The reason why I brought it up was that Lunar Lake was specifically designed to make an x86 offering that has comparable battery life with chips from Qualcomm and Apple. It's not a standard marginal generational improvement, for light office tasks (the type of use case this laptop is intended for) you get around 17 hours of battery life vs the previous generation's 10 hours (see benchmark results here: https://www.pcworld.com/article/2463714/tested-intels-lunar-... ).
I wanted to understand if any other cpu generations had comparable gains over the previous, but I need to work on an Intel ark scraper. Maybe I'll find some time to post it back here on HN. Thanks for added details
It's not surprising, Meteor Lake and onwards are all TSMC and unlike Intel they still actually know how to manufacture chips. Arc iGPUs are also a massive improvement over Iris.
It's not just a TSMC vs. Intel thing, Lunar Lake gets 7 more hours of battery life than Meteor Lake when doing the same "light office tasks" benchmark.
Ok that part is surprising, Meteors already draw very little, like < 25W unless turbo boosting and 6W idling. What kind of wattage does Lunar have?
The numbers I've seen show that the peak power draw is similar to Meteor Lake, but it generally draws ~40% less than Meteor Lake when doing the same task. The SoC only idle power draw figures I've seen are 0.62W for Lunar Lake vs. 2.32W for Meteor Lake.
Do the math on "6W idling". That's terrible for SoC energy draw. For modern expectations of battery life the SoC has to be in the milliwatt range.
Selling a laptop with soldered ram might literally kill the company at this stage. Their entire brand and market is based on repairability.
Yeah I complained about this on the framework subreddit when I first heard of the FW12, asking for reasons and expectedly got downvoted into oblivion. Apparently expecting them to use something mildly recent is too much to ask for and it's supposed to be a cheapshit student laptop aimed at schools or whatever. Wouldn't be as bad if it weren't Intel's worst CPU gen in recent memory, FW probably got a bargain box deal on them.
Meanwhile the HX370 is still warm from the oven and already in the FW13, linux drivers aren't even ready yet with compatibility complaints aplenty. Not to mention the FW Desktop with the AI Max which was the first launch of that chipset worldwide.
I ran XPS Developer Edition (Ubuntu) for 5+ years without any issues.
Using Framework for past year - every day I have to reboot because it freezes with 20+ Firefox tabs (Ubuntu 22.04, AMD). Tried all options (disable vGPU etc) but no luck.
That's frustrating, I'm sorry! I don't experience anything like that, and I have over a thousand tabs open in Firefox on NixOS.
Based on https://frame.work/linux Ubuntu 24.04 is the minimum supported version for the older AMD F13's, so I'd suggest updating and then reaching out to Framework if the problem persists on a supported distro.
on chrome based browsers with Linux I have to disable 'use graphic acceleration when available' in the browser settings or the browser with freeze intermittently. This has been a consistent issue across distros for years. Not sure if it also affects firefox.
Disabling this feature mostly works, but results in poor performance for some graphics heavy websites.
I also had a lot of issues with my AMD framework laptop and ended up reverting back to a older Intel Framework laptop. Top issues with the AMD laptop were the realtek wireless, and random AMD (integrated) GPU glitches. Intel hardware continues to be absolutely top tier for Linux support.
> on chrome based browsers with Linux I have to disable 'use graphic acceleration when available' in the browser settings or the browser with freeze intermittently. This has been a consistent issue across distros for years. Not sure if it also affects firefox.
Oddly enough works fine on all Chromebooks. It really is a matter of platform hardware/firmware qualification being suited to task.
Yep, my last laptop purchase was an xps13. Unfortunately the 8gb of soldered on ram is becoming woefully insufficient with how bloated the modern web has gotten.
I have a beefy desktop, but if I replace my laptop I think it will probably be a thinkpad.
Sounds like you might have received some bad RAM. Have you tried replacing it?
People make fun of Chrome being a RAM hog but I'm having more issues with FF than I ever did with Chrome. For some reason on reddit specifically FF will randomly freeze for a few seconds, my cursor will stop rendering on top of it (like it's going behind the window) and it won't accept inputs. If I keep typing through it it all appears in the window after it unfreezes. Also FF will regularly take multiple gigs for random tabs like Youtube if I leave them open.
Tldr it might be more of a firefox problem than a framework problem.
I would hazard a guess that it’s more of a “devs only test against Chromium” problem than it is a Firefox problem. It’s a problem seen under WebKit-based browsers at times, too. Gecko and WebKit often behave differently and have different performance characteristics than Chromium/Blink does, but that’s often not accounted for at all. The extent of QA on non-Chromium browsers too often stops at “it technically runs”.
That's also probably part of the story but ultimately as an end user the fact is Firefox is a bad experience I'm suffering through only to not use Chrome. I can't force websites to patch whatever memory leak is causing Youtube tabs on FF to eat 5 GB of ram for example.
I have a ThinkPad T480s with Ubuntu and Firefox. Works like a dream except every now and again the whole thing locks up (have to power cycle). Happens with AWS Console and LinkedIn. Maybe it's doing me a favour.
I've never had any issues on my AMD framework 13 with way more tabs than that (on Void Linux).
Older laptops, especially all-Intel ones, tend to be way more reliable on Linux from my experience.
I would have bought a Framework laptop but here were the dealbreakers - Ancient Intel processors - AMD option seemed to be accompanied by broken wifi
It's not that hard. Just provide a single modern option that just works. I don't want to troubleshoot device drivers in my spare time.
What's wrong with the AMD version's wifi? I've got the original AMD F13 (and run NixOS) and I have no issues.
It's also a plain old M.2 WiFi/BT module. So if you really hate it, swap it out.
I haven't had any issues with Wifi. A family member replaced theirs with an Intel chip to fix an issue; turns out reception was just bad and not the wifi chip.
That brings me to my second point: if the wifi chip's what's holding you back there's nothing stopping you from replacing it.
Wifi worked out of the box for me on FW13 both the AMD 7840U and the HX 370 mainboards, in Ubuntu 24.04 LTS and 25.04. Ethernet via USB-C dock worked out of the box as well.
> It's not that hard
I suspect it's actually quite difficult to build an entirely modular laptop to work with several operating systems while supporting the latest components, and be void of any driver issues.
I use the Framework 16 AMD with the Mediatek MT7922 WiFi 6E wireless chip using the mt7921e driver and it works great
I was seriously considering getting a Framework laptop (but 13") and what stopped me is the limited shipping. Right now I live in the area where I can order one, but I was looking at another country where I'd prefer to live. It won't happen next month or even this year. But if I order a Framework laptop I want to have access to the new and replacement parts and Framework doesn't ship there. I don't want to rely on forwarders. And that's a deal breaker. So I went with another Thinkpad instead.
And on the subject of 12" laptop, when it was teased I hoped for a MacBook Air killer. But that didn't happen. I like what Framework is doing, but common manufacturers are still looking like a better value.
I’ve been really curious about some of the hybrid handhelds / micro laptops from GPD lately. Framework should consider looking into this market niche as well.
I love my GPD Win mini, and I've had a few other GPD products before that, but I'm not sure if a framework-style modular design would work as well there.
For starters, they've always gone with soldered RAM for both physical size and performance reasons (RAM speed matters more with integrated graphics, and soldered RAM can go faster that socketed RAM.)
Additionally, I don't think there's enough space for a reasonable number of ports via expansion cards. Even in the larger 8-inch-screen models, I think 4 expansion slots would be very difficult, whereas the current 7-inch models have 5-8 ports. (I think some of the MicroPC models had a handful of fixed ports + one modular expansion slot, so maybe they could change that one to a framework-compatible design?)
To their credit, GPD has offered motherboard upgrades in the past when a refreshed design is otherwise compatible. But there are often changes in ports or layout or cooling design that make that impossible.
I'm glad Framework exists but I don't envy them. Selling laptops aimed at the most nitpicky never-satisfied group of people must make it hard to see through the noise. I wish them the best.
Is my framework perfect? No. But it's still the best laptop I've ever owned. Just like open source software, it's not about having the most shiny look and feel, but it's about knowing that the creator wants me to have the freedom and empowerment to open it up and adjust it to my needs and desires. Other companies will just say that I've voided the warranty when I try fixing or tweaking it myself.
Maybe these users are picky, but they are under-served, very influential, and very productive w.r.t. improving support for the laptop in the ecosystem. At this point FW became a go-to laptop for Linux hackers, and it gets great support in all avant-garde tech like e.g. NixOS.
There is a certain rich racingdriver/programmer/macuser who moved his whole company to framework laptops running linux. So they must be doing something right
I don't know who you are talking about but I'm not implying that Framework is doing anything wrong, just that they have chosen a very difficult user base to cater to.
I'm actually curious how much difficulty they have with their actual customers. There are always a ton of negative comments in these threads, but those seem to be from people who just fundamentally value different things than what framework is tyring to do (which is totally fine and valid!), and therefore very specifically aren't the customer base. I can't speak more broadly, but I've been super happy with my Framework16, warts and all (and there are a few).
My guess is that the core audience of people who want a repairable, upgradeable laptop tend to be tinkerers and don't mind a little bit of extra effort.
I also want to know but I doubt they will come out and say "yeah our customers are gigantic pains in the arse"
https://en.wikipedia.org/wiki/David_Heinemeier_Hansson
https://world.hey.com/dhh/linux-as-the-new-developer-default...
For whatever reason the FW 12 is giving me iBook G4 vibes and I love that.
Is it just the colors? (I also really like that)
It's a combination of the compact, 12" plastic shell with the soft edges and rounded corners, near edge-to-edge keyboard and a soft two-tone colour palette. I only wish that it had a more square-ish, 4:3 screen like the iBook.
Basically it says buying a used laptop worth more than the Laptop 12
People have a lot of pretty valid issues with Framework. When you optimize as hard as they do for repairability, upgradeability, and modularity, you are going to make some sacrifices in other arenas that won't be to everyone's liking, and that's totally fine. But the one criticism that I haven't quite understood is price.
Yes, Framework is more expensive....the first time. But from then on (and for the 13, this is now 3 or 4 generations I believe), buying a new mainboard with a latest gen chip is going to be the cheapest you can get those specs for. Yeah, a used laptop might still be cheaper, but it won't be the same specs, and if used is what you want, you can probably get a used mainboard from the last gen that someone else is upgrading from for even cheaper.
Unless you are someone who thrashes their laptops so hard that you expect to need to replace not just the mainboard, but also the screen, keyboard and chasis, then framework represents by far the cheapest way to keep getting new laptop specs into the future.
I'm not even someone who upgrades laptops very often (every 5 years or so, historically), and even on that schedule, and I still thing that long term it will be cheaper. If you are someone who upgrades more often, then it makes an even bigger difference.
Again, to be clear, I think there are very valid reasons that a Framework might not be the right laptop for someone. I just thing that thinking about the one-time purchase price of the entire laptop is completely missing the entire point of Framework.
The more videos with the FW12 moving and used before the Camera I see, the less I can ignore the fat bezels. The design language at all is not made for „business“ which is refreshing and they obviously have a budget approach, but such ancient bezels don‘t do a contribution for anything. The lack of any Windows Hello enabling hardware was the final bit for my sad no-buy decision.
It's going to remain a niche token product for everyone but connoisseurs to ignore with that kind of a specs-to-cost imbalance.
I don't see anything wrong with that Framework is going after a specific but lucrative segment of the market
No objection from me either. It's just that it defeats their intended purpose of pushing a paradigm shift to high-sustainability devices.