Honestly, it was much cheaper than I had assumed it would be. A pre-built "performance" version starts at $1,699 (Ryzen 7 7840HS) with 16GB RAM and 512GB SSD. Compare that to the entry 16 inch MacBook Pro at $2499.
The "overkill" system with a Ryzen 9 7940HS, 32GB RAM and 1TB SSD is $2099. A pseudo comparison to the 16 inch MacBook Pro with M2 Pro, 32GB RAM and a 1TB SSD is $3099.
Also these systems are cheaper if done DIY and no Windows tax. Framework really might have a winner if this system can live up to their promises.
Not being snarky or anything, but do you have easy access to broadband internet where you are, or do you live in a locale that prohibits VPN?
It turned out more effective for me to buy a Chromebook and build a remote workstation to do actual work from, with the added benefit that I don't care if my Chromebook is lost or stolen because it doesn't store any data and costs less than a screen replacement (<$179).
I also wasn't impressed by the highly externalized "there's an adapter for that" dongle-shuffle for basic ports when Apple did it 2017, and if I wanted to bring it all with me, I'd still need to buy and lug around a eGPU if I wanted to do any compute. Plus I get like 14 hours of full-brightness work time off of one charge, similar to an M1 MacBook Air.
That's just my use case, of course. I've also built an SFF system with eDP when I have to travel and generate a lot of data and an Alienware m17 isn't enough.
Remoting into a beefy workstation works for plenty of command-line tasks, but is going to yield a subpar experience for GUI tasks. As someone whose daily driver for several lean years was a $180 netbook (remember netbooks?), anything that I could do on my remote machine was grand (at the time, a $5 VPS), and anything that I had to do locally was a slog.
> "there's an adapter for that" dongle-shuffle
> lug around a eGPU
One of the unique selling points of the Framework is that you can customize the ports to your needs, so dongles are unnecessary unless you need a whole lot of ports. And the Framework 16 doesn't need an eGPU, you can plug in an actual GPU that integrates with the body of the laptop.
No worries habibi. Your post isn't snarky and I get where you are coming from with what you wrote. I was more commenting on my expectations of the price given that this is a (relatively) new company on a larger form factor that has a lot of interchangeable parts, external GPU, etc.
This laptop isn't personally for me but I can see why people would like it. Also my workflow requires a lot of GUI applications for simulating and visualizing robotic movement and that is incredibly hard with ChromeOS or macOS.
admittedly, my primary role hasn't been development in over a decade, so nothing fancy. primarily vim with plugins, good integrations, and native compilers/utilities, and Microcenter or Costco special Chromebooks (14" 1080p matte screen, plenty of ports, and uSD). besides some native Linux utilities, the Chromebooks just run a browser for guacamole and some other hosted services.
I'm very interested by the Framework Chromebook, but it's a bit excessive for my needs. I'm also provided a lot of Apple hardware, which I'll use in the office.
That’s a great idea. Thanks, I’ll seriously consider it. I’m skeptical of the network usage though. I might need to use cellular with this approach, does this result in high usage?
that's based on your compression and what you're transferring, so it depends on your definition of "high" (ie- cost, likely to exceed an "unlimited" plan, etc).
I haven't exceeded 1GB/mo on cellular in over a year, but I also have easy access to wifi.
Thanks! I was skeptical, but I took the time to actually compare Framework, Macbook and custom remote workstation on these metrics: battery life, security against theft and accidents, high speed networking, upgradability, repairability, compatibility (eg. external Nvidia GPU, coreboot), offline use, performance and weight. Remote won on every metric (has no single drawback) except offline use, which is not a problem when you are at home. So I decided on doing this. If you don’t mind, how do you connect? What is your setup? Are you able to eg. forward USB devices to the workstation?
Closed the tab on the original, better response. Sigh.
I don't use usbip, but it's an easy setup. I do use kvm/libvirt. Tailscale didn't exist when I built everything, but the documentation is excellent, and I could replace most of this with similar convenience.
Everything I'm mentioning here is locally hosted.
I connect via either wireguard or Unifi Teleport, depending on which network is needed. Wireguard connects to guacamole which manages connections to more secure core services, controllers, and shares. Unifi Teleport allows easy access to surface services like a book server, IOT, NVR, webuis, APIs, etc.
Chrome remote desktop is much more convenient than I expected, and I use it to connect to friends' systems or to hop onto graphical instances that I don't care to set up lasting connections to.
My three primary nodes are for GPGPU compute, "big" jobs, and storage. I've removed most IPMI/IPKVM to allow more flexible consolidation and upgrades. My "OOBM" is now UPS with remote power plugs with always-on BIOS settings. Once things gracefully halt, I can kill power if/until needed (my downtimes are all planned, but I reserve the capability).
I've greatly simplified over the past half decade so I can focus more on using than administrating, and while I haven't removed all of it, I try hard not to create environments more complicated than I can understand after not being touched for a few months.
Thanks, that was one deep rabbit hole of a reply! I did not think things would be this complicated. I'm happy to hear some things for which I've intended to roll my own solution with my RPis already exist. I took my notes, I appreciate you!
No compare it to the MBA 15 w 512GB SSD and 16GB RAM upgrades and the 70W charger from their educational store + $150 gift card offer. $1575 inc Cali tax
Yes but the AMD design draws double the power even though it was fabbed on a lower process node. I mean it defeats the purpose of a portable. Now you need double the battery weight or lose half the runtime. No, the M2 still wins. If processing is the goal at the expense of portability then you can buy a nice 64 core EPYC system for $1700
The battery life is still fine. Plugging in once in a while isn't a huge deal. I'd rather save time compiling than not having to plug in once every 8 hours.
Yeah I was just referring to default prices. The only one I modified was the comparison to the Overkill system because the base Apple configuration does not have 32GB of RAM and 1TB SSD.
Ok, but that same spec is 1699 plus tax direct from Apple without a discount so the same as frameworks pre built 16 inch option that ships in Q4. If you compare their “build it yourself and bring your own OS” version without memory or storage shipping in late Q4 the base MBA 15 +70W charger w/o discount is still $100 cheaper, shipping now and very likely will outperform w double the battery life
What's the comparison look like with 2TB of storage? And maybe a bit more RAM (I don't think MBAs let you upgrade after purchase so buying a 512GB machine seems shortsighted for a lot of use cases unless all your development is in the cloud)
The storage and memory upgrades are expensive and also have a penalty in terms of power. Memory draws the same current regardless of if it’s being used and that is significant. Unified memory is also very different, has much higher bandwidth than RAM and lower latency. You can’t compare unified memory with RAM.
As far as the SSD, the base option comes with one 256GB NAND flash IC, where if you upgrade to 512GB they have to add a second. That matters because adding a second means you double the IO bandwidth using their current design. It’s something like 3500 vs 7500 MB/s. If you add more NAND modules you also scale up the IO speed further but the difference between 3500 and 7500 is significant for all workloads where above that you probably won’t see a gain except in niche applications like video editing, assuming you have the complimentary processing power and unified memory to utilize that throughput. For storage, Apple provides 2TB cloud service for $10/mo and these have wifi 6E so it’s cheaper to use the cloud than buy the storage, assuming it’s not working storage, and safer too.
Also, upgrades on Apple products hold their market value very poorly. The M2 studio just came out and maxed out M1 units are on sale NIB at 1/4 price of retail from 3 months ago, which is not at all a true quantification of performance since the M2 is only what 15-20% faster.
So after quite some research for the MBA 15; 16GB + 512GB + 70W adapter is the optimum for power draw, performance, cost and value preservation.
Overall, it seems the best value strategy with Apple products is to take the first upgrade for memory and storage and leave it at that. You can then resell every year and recoup 60% and upgrade to the latest w a new battery. Apple is actually very cheap after the first investment considering the productivity and time saving benefits, and you can’t steal an Apple device. Just don’t drop it
Also if you are considering big upgrades, compare the cost to just buying two base units as the base units are way closer to the manufacturing cost. Often there is more benefit to having two expendable units and they will hold their resale value better.
Unified memory literally is RAM. It's just LPDDR5. It might have better timings, due to being located closer, that's it. The framework 16 also does DDR5-5600, so you get 100GB/s of bandwidth with a dual rank memory setup.
In theory you'd take a latency hit, but in reality (at least the M1) was limited by it's memory controller and had DRAM latencies of ~100ns, which is actually worse than Zen 4 at ~70ns. This is because before querying RAM, you first need to query every level of cache, and then you hit the latency of the DRAM itself, so the actual distance is not a dominant factor.
So no, unified memory doesn't give any actual performance advantage. It could in theory (and marginally), in practice it's just marketing.
Apple doesn’t verify that purchases from the educational store are from students it’s honor system. Those prices are available to everyone, in small quantities.
Sit the 15 inch MBA next to the 16 inch MBP and the screens look identical. Performance wise the M2 w unified memory crushes X86 systems across the benchmarks at half the power budget.
The "overkill" system with a Ryzen 9 7940HS, 32GB RAM and 1TB SSD is $2099. A pseudo comparison to the 16 inch MacBook Pro with M2 Pro, 32GB RAM and a 1TB SSD is $3099.
Also these systems are cheaper if done DIY and no Windows tax. Framework really might have a winner if this system can live up to their promises.
EDIT: The verge also has a deep dive into the system https://www.theverge.com/22665800/framework-laptop-16-hands-...