QuoteLPDDR5x-7467, Dual-Channel
AFAIK: DDR5 SO-DIMM dual-channel (2*64-bit per channel) = LPDDR5X octa-channel (8*16-bit). Calling it a "128-bit system memory bus width" (the fast majority of laptops, mini-PCs and desktop PCs has a 128-bit RAM bus width), would be clear regardless.
119 GB/s = 128-bit * 7467 MT/s / 1000 / 8.
And yes, 7467 MT/s is not top of the line, that would be 9600 MT/s, that Panther Lake also supports. Maybe there is a 9600 MT/s X14 Carbon config.
Running AI LLM models locally and privatelyIf you care about using this laptop for AI extensively, know that with current SOTA, for its size, AI LLM Qwen3.6-35B-A3B model and its lowest usable quant, you will be restricted to about this amount of context:
Quote from: reddit.com/r/LocalLLaMA/comments/1sq94qx/is_anyone_getting_real_coding_work_done_with.. I've come to the conclusion that (1) 32768 is the biggest context I can get away with in an adequately smart model, and (2) it just ain't enough.
More infos here: notebookchat.com/index.php?topic=315954.0 ("Your own ChatGPT, offline: AI without the cloud on your laptop").
So, should you require more context for AI, get a 48 GB RAM config (if it exits) or the 64 GB RAM config.
A 9600 MT/s config will run AI token generation 9600/7467 = 28.5% faster. The 28.5% higher memory bandwidth should also allow for a faster iGPU, which in turn will result in faster prompt processing. (see the mentioned topic=315954.0, if you don't know what these terms mean)