My First 15 Minutes with RTX Spark
Linus Tech Tips YouTube 535,376 views as of 15 October 2026
Why it is here
Linus Tech Tips at Nvidia’s RTX Spark launch event (14:36): a press conference on local AI and hands-on demos with Hermes, ComfyUI and OpenClaw. About 535,376 views on 2026-10-15.
Description
Description written by Gemini from the videoGemini 3.8 Flash, 15 October 2026
Summary
Linus Sebastian from Linus Tech Tips attends an NVIDIA press event and tests the newly announced NVIDIA RTX Spark platform on an MSI Prestige N16 Flip AI+ laptop. He provides a first-look overview of the hardware architecture, local agent onboarding and performance in Windows, gaming with DLSS 5, anti-cheat compatibility on Arm, and local multi-device routing.
What is shown
- [00:19] Architecture diagram of the RTX Spark SoC, highlighting the 20-core Grace CPU, 128 GB LPDDR5X unified memory (300 GB/s), and Blackwell GPU tile built in collaboration with MediaTek.
- [01:42] Floor demo of a mobile robot utilizing onboard physical AI to pick up socks and drop them into a hamper.
- [03:21] Hardware overview of the MSI Prestige N16 Flip AI+, demonstrating its thin profile, port layout (dual USB4 40Gbps, HDMI 2.1), 360-degree convertible hinge, and 99.9 Wh battery.
- [04:04] Setup and execution of Hermes Agent in Windows running purely offline on local Qwen3.8 27B; given a single prompt, it builds a fully operational real-time system monitoring dashboard plugin.
- [06:56] Live demo of NBA 2K27 at 1440p showing DLSS 5 3D-Guided Neural Rendering toggled on and off.
- [07:42] Hands-on gameplay of Marvel Rivals running natively on Windows on Arm with functional anti-cheat.
- [08:58] Real-time path-traced ray tracing running in Java Minecraft.
- [09:56] Side-by-side rendering viewport comparison in Blender between OptiX Denoiser and DLSS Ray Reconstruction.
- [10:36] Demonstration of NVIDIA PAIR (Personal AI Router) distributing compute across multiple local devices on a LAN.
- [12:42] OpenClaw smart-home assistant demo adjusting room ambiance from conversational input, inspecting security camera feeds to diagnose a broken lawn sprinkler, and drafting service requests via Discord.
Claims & numbers
- The flagship RTX Spark N1X configuration features a 20-core Grace CPU, 6,144 Blackwell GPU cores, 1 Petaflop of FP8 AI performance, and 128 GB of LPDDR5X unified memory (256-bit, 300 GB/s); the entry-level SKU has 18 CPU cores, 5,120 GPU cores, and is capped at 32 GB RAM.
- NVIDIA compares RTX Spark gaming performance to the discrete laptop RTX 5070 Mobile.
- Retail mockups show ASUS ProArt P14 and P16 laptops with RTX Spark priced at $6,499.99 and $6,999.99.
- NVIDIA claims running 61 million output tokens locally on Qwen3.8-Flash-Next costs approximately $20 in electricity, compared to roughly $1,200 per month in API fees for cloud models like Claude Sonnet 5.5 Medium.
- Hermes Agent completed the autonomous multi-step dashboard coding prompt in 1 hour and 16 minutes (5:21 PM to 6:37 PM) using local Qwen3.8 27B.
- DLSS 5 features a model claimed to run twice as fast as DLSS 4, alongside an optional performance mode scaling frame generation up to 6X.
- Anti-cheat compatibility on Windows on Arm covers over a dozen competitive titles starting October 16th.
Notable quotes
- [00:11] “This is the first credible Arm-based gaming chip for Windows...”
- [03:50] “It’s a convertible! As cool as all that is, can the setup possibly be as simple as what Nvidia showed on stage?”
- [05:07] “That is pretty cool. Very powerful, and we’re just scratching the surface.”
Assessment
This is a hands-on preview and event impressions video filmed at an NVIDIA press showcase, interspersed with home testing of Windows agent setup. While Linus personally verifies the local Hermes agent dashboard generation and tests demo floor systems, comprehensive benchmarks, thermal testing, and battery runtime validation are deferred to the full review.
Described by gemini-3.8-flash on 2026-10-15 from the video’s audio and frames.