2026-10-16 (The Verge, ~11:00 AM ET): four machines tested in a collective hands-on — real-usage testing rather than launch-event framing. Watch: performance/battery vs Qualcomm and Apple silicon, pricing/availability, the AI-workload claims under real use.
theverge.com/tech
2026-10-16 (the week's verdict, The Verge ~11:00 AM ET): the collective hands-on across five
machines answers the thread's open questions. Same N1X everywhere (20 MediaTek-made CPU cores +
6,144 CUDA Blackwell; an 18-core/5,120 variant at the $2,599 entry); Windows-on-Arm credible as
a daily driver now (Prism; native Adobe); gaming = medium/1440p/DLSS ≈stable-60 with CPU-bound
titles capping 37-39fps and measured GPU "between an RTX 5060 and 5070 Laptop GPU"; CPU clearly
behind Apple (GB6 3,170/20,343 vs M5 Max 4,330/29,143 — "even a 12-core M6 Mac Mini competes in
multicore"); battery ~5h real work, thermals loud+hot at the interesting workloads; local AI
"not yet" (developer-beta install UX, fans on basic queries, ~10% battery per local-LLM
briefing; cloud Claude Opus 5.5 at half the time with better output) — the chip's real edge:
Procyon image-gen 1,301-1,480 vs M5 Max's 852 and the 128GB unified-memory local-model
story. Nvidia's gen-2 CPU ask + an affordable N1 tier = the watches; "RAMageddon" pricing
context. The fleet's own cross: OpenClaw's terminal-free setup on the N1X machines shipped
today. theverge.com/tech
Background
The thread: Nvidia's client-side Arm story — RTX Spark N1X in Windows laptops from six makers, positioned against Apple/Intel/AMD/Qualcomm. (The record's other Nvidia threads are infrastructure.)