Re:Framework Laptop 16 Dual M.2 Adapter Expansion Bay Module
by sottol on Monday December 16, 2024 @04:01PM
(#42435398)
Great idea - I'm currently using a barebones Framework 16 mainboard with the 7940HS as my homeserver and local-AI playground in a 3D-printed case.
UMA with the iGPU is nice though not much faster than CPU (20-30%? don't recall).
But installing 64 or even 96GB RAM is more attainable though slower (60GB/s) than Apple's $$$-mem-upgraded M* Pro (200), Max (400) and Ultras (800). Although DDR5 is comparable to a regular M1 (70) afair. Most LLMs in the 20GB range run at around 7-12tok/s which is fine for playing around with them.
Re:Framework Laptop 16 Dual M.2 Adapter Expansion Bay Module
by MostlyStable on Monday December 16, 2024 @03:11PM
(#42434887)
oooh, I was just looking at their refurbished M2 drives (1TB for $50, 2TB for $115). Adding 4 TB of storage for ~$275 looks pretty good. I doubt I need quite that much, but I may get it anyways, haha.
Re:Framework Laptop 16 Dual M.2 Adapter Expansion Bay Module
by nrp on Tuesday December 17, 2024 @11:12AM
(#42442559)
We expect most Framework Laptop 16 users won't need more than the two M.2 storage slots that are on the Mainboard itself, so we don't want to end up with a wasted module.
Or you can also test RAID 0 speed.
UMA with the iGPU is nice though not much faster than CPU (20-30%? don't recall).
But installing 64 or even 96GB RAM is more attainable though slower (60GB/s) than Apple's $$$-mem-upgraded M* Pro (200), Max (400) and Ultras (800). Although DDR5 is comparable to a regular M1 (70) afair. Most LLMs in the 20GB range run at around 7-12tok/s which is fine for playing around with them.
https://frame.work/blog/our-first-new-framework-laptop-16-ex...