The new wave of on-device AI laptops with dedicated NPUs - worth the upgrade or mostly marketing?

Mariam Al Mansoori · 0 likes

software catching up to hardware capability is basically the history of every new chip category, give it 18 months and the apps will actualy use it properly

Alma Nilsson · 0 likes

local background blur and noise cancellation on video calls runs entirely on my npu now and my fan doesnt spin up anymore during meetings, thats a real win for me

Gentle Coder · 0 likes

bought one specifically to run local llms offline and honestly the npu support in the inference frameworks is still rough around the edges, gpu still does most of the heavy lifting

Mango Coder · 0 likes

yeah same experience llama.cpp npu backends are improving but not there yet for me

Silver Noodle · 0 likes

marketing is definitely ahead of reality right now but i dont regret the purchase, battery life alone on ai-assisted tasks justified it for my use case

Haruto Sato · 0 likes

windows is adding more npu-accelerated os level features every update cycle, feels like the os vendors are driving adoption faster than third party apps are

Youssef Mahmoud · 0 likes

as a photographer the on-device raw processing filters using the npu have been genuinely fast, noticeably faster then the cpu fallback on my old laptop

Hamza Malik · 0 likes

feels like the npu is this generation's 'dedicated AI chip' the way physics accelerator cards were a thing in the 2000s before gpus absorbed that workload entirely