Lunar Lake is probably not coming to Framework. Arrow Lake is more likely.
Lunar Lake is targeted at laptops with half as much cooling capacity as the Framework Laptop 13. Lunar Lake has much lower core counts (Lunar Lake has at most 8 cores and 8 threads, whereas the Meteor Lake that Framework currently offers have 16 cores and 22 threads) and has ram and Wi-Fi built directly into the CPU (meaning they cannot be upgraded).
Arrow Lake (which is also expected to be sold under the Ultra 200 naming scheme) is the actual successor to the Meteor Lake CPUs that Framework currently offers and has specifications and features more in line with that.
That has been an option since the very first Framework Laptop. The Intel based Framework laptops all have 4 USB4 ports (with 12th gen and 13th gen having Thunderbolt 4 certification and I expect the Core Ultra 100 models are undergoing that certification process) while the AMD based Framework laptops have 2 USB4 ports (other ports are limited to USB 3 10 Gbps).
All models are compatible with external dGPUs.
That would require a meaningful design change, which would hurt compatibility between parts. I think dual port modules are probably the more likely way for Framework to try to squeeze in additional IO.
So will there be a passively cooled Intel core Ultra 200 (Lunarlake) laptop version? (or with the “solid state” peltier element fororesystems element fans)
As for ram upgradebility, it is already convinient for me to upgrade mainboards (since they can be just swapped) in framework laptops, ram clearence or sticks are not needed for me.
lower corecount and lower powerenvelope are not a problem for me, just means that fans wont run loud under use.
I just like frameworks open design, where its possible to replace things within the shell, battery, keyboard, mainboard, display and if it means, that lunarlake would come, Id think id would be pretty cool.
Ofc if the CPUs would be socketable, would be also nice, but laptop CPUs are BGA so if Memory to, so shall be it if its more efficient.
I think when aiming for Lunar Lake the lowerpowered CPU would bottleneck any worthy GPU you connect to it.
We need to wait for either a thunderbolt 5 CPU which is within the normal power range so connecting a GPU makes sense, or for someone to create a Oculink connection which can be put into the current Framework laptops, while reaching the top speed of Oculink. There are some topics on this forum working towards an Oculink solution.
nah man , I think it might be mighty fine with a modern DGPU.
for example people paired i7 8700K’s with GTX 1080TI’s.
cpu z score of the i7 8700k (single ~570 and multi ~4350 points [as reference]) for a pairing with a gtx 1080ti (fp 32 (float) of 11.33 Tflops)
meaning with the CPU (ultra 200) having [P core lioncove about singlecore ~ expected 800 and E Skymont ~400 (since intel N200 uses gracemontE cores ( 2 versions older than Skymont//Gracemont → Cresmont → Skymont) + newer lithography) meaning its extremely fast at that low of a powerconsumption]
So that 17 - 30w cpu is about as fast as a 98w 8th gen cpu and could easily push some modern midlevel DGPU’s (maybe arc battlemage ones as well), as for the Igpu it is quite extremely competetive in itself, vs AMDs ones, but at a lower powerenvelope.
(has extremely low latency Ram on chip which also uses less power.)
(this gen (core ultra 200) and AMDs use similar TSMC nodes (amd 4nm finfet and intel uses their N3B[newer and better]))
so yea.
and if intel makes them cheap to aquire, than its a W, but if not its lost potential.
I saw the first reviews coming in, only the Intel numbers though, but indeed, looks really good, and would pair nicely with a eGPU. A shame the ram is not upgradeable.
Only thing we now need is Thunderbolt 5. Then my preferred setup of Laptop and eGPU can become reality!
To answer the original question: Lunar Lake is supposed to have only 3 TB ports instead of 4.
One more way in which it is targeted at smaller devices.
But it looks like it might be more 9000-gen TB controller probably implementing USB4v2. Because Intel’s official slides quote the same 3 DP tunnels per port that the new Barlow Ridge / 9000 controllers do (of which there are 40Gbit/s and 80Gbit/s variants. But both are USB4v2 with the 3 DP tunnels).
USB4v2 allows for full PCIe payload sizes. Will make PCIe for eGPUs more efficient.
To be fair the amount on devices that actually used all of the previous 4 is very small, not sure I have even seen any outside the framework, so can’t really fault intel for that too much.
, but yea it is indeed a shame, that the ram is not upgradable, maybe next gen with camm2 or camm3 (on chip ramconnectors) it could be replaceble and performant.
That was a deliberate choice to make it more power efficient. I do not think we fully know how much it does in % compared to all the other things (like monolithic, new TSMC node, single die). But I think the on-package RAM is supposed to have a large contribution to the power efficiency at the GPU performance it has. Would be interesting to know how close one could get to that with modular memory with all other advantages.
That was actually one of my considerations when I got my X280 from an e-Waste recycler. Although the 16GB RAM is soldered and it is an Intel 8th gen i7 with a 1080p touchscreen. It was really cheap (250+ USD) without an NVMe drive. It probably will run until it fails or I need better spec hardware, which by then maybe I’ll snag a Lunar Lake one. Buying it used is helpful to the environment while being at a great value for the potentially short lifespan it has.