Nvidia’s RTX Spark Laptops Hit Shelves With Adobe Backing and a $2,600 Floor

Nvidia’s first RTX Spark laptops reach stores today, Friday, October 16, backed by the kind of software endorsement that often decides whether a new silicon platform sinks or swims: Adobe has reworked Photoshop, Premiere and Substance 3D to exploit the chip. The promise is a Windows notebook that runs large AI models on-device, handles professional video edits and plays games without a discrete graphics card. Prices start at $2,600.

• Arm meets Blackwell: up to 20 Arm CPU cores, a Blackwell GPU and as much as 128GB of unified memory on one package.
• Adobe on board: optimisations the two companies claim deliver up to 2x faster AI processing.
• Costly configurations: entry systems start at $2,600, while 128GB machines run past $5,900.
• Caveats: several AI features remain in beta, and x86 software still depends on emulation.

Arm Meets Blackwell: The Hardware Story

RTX Spark is a system-on-a-chip that fuses CPU, GPU and memory into a single piece of silicon, following the template Apple established with its M-series processors. The flagship tier combines 20 Grace Arm cores — ten performance and ten efficiency — with a Blackwell GPU carrying 6,144 CUDA cores and up to 128GB of LPDDR5x memory at 9,400 Mbps. A second tier drops to 18 CPU cores and 5,120 GPU cores. The CPU was co-designed with MediaTek, and the chip is built by TSMC on a 3-nanometre process.

Memory bandwidth reaches 300 GB/s, just shy of the 307 GB/s associated with Apple’s M5 Pro. Nvidia also claims the platform is twice as efficient as a laptop pairing an Intel Core Ultra 9 285H with a dedicated RTX 5070 while gaming at 60 watts. That is a vendor figure. Machines with discrete graphics can draw far more power to go faster, so the real gain is battery life rather than peak performance.

Pricing, Memory and the MacBook Yardstick

Unified memory drives the bill. Systems using the 18-core part start at $2,600, while the 20-core tier begins at $3,700. Lower configurations top out at 24GB or 32GB, so reaching the headline 128GB requires the most expensive models: the Surface Laptop Ultra with 128GB costs $5,900, and the Asus ProArt P16 starts at $2,800.

The obvious yardstick is Apple’s 16-inch MacBook Pro with M5 Max and 128GB, priced at $7,000. Launch partners include Asus, Dell, HP, Lenovo, MSI and Microsoft.

Adobe’s Bet, Local AI and the Emulation Question

Adobe’s participation is the clearest sign RTX Spark has staying power. Photoshop has been rebuilt around GPU-accelerated compositing, visible in the Remove tool and on-canvas effects such as Light Leaks. Premiere gains acceleration for playback, rendering, colour work and AI features including Enhance Speech and Object Mask, while Substance 3D Painter speeds up texturing and the handling of large projects.

“AI agents are helping creators unlock deeper and richer workflows using the Adobe applications they know and love. Powered by RTX Spark, Adobe’s optimised tools deliver up to 2x faster performance to accelerate these new workflows.” — Jason Paul, vice president of local AI at Nvidia

Deepa Subramaniam, Adobe’s vice president of product marketing for creative professionals, frames the effort as aimed at users for whom performance is critical, noting that projects now routinely span images, video, audio, motion and 3D. Not everything arrives on day one: the AI enhancements in Photoshop and Premiere are in beta and require sign-up, and the Substance 3D improvements are promised “soon”.

Nvidia is also releasing Nemotron 3.5 Bolt 74B, a 74-billion-parameter model that activates nine billion parameters per step and fits within 24GB through 2-bit quantisation, and taking PAIR, an open-source tool that spreads AI workloads across machines on a local network, out of beta. Downloads of local models are now five times last year’s total, the company says. Its own demo — Qwen3.8-Flash-Next scoring 40 points on the Artificial Analysis index on a 128GB RTX Spark, against 41 for cloud-based Claude Opus 4.7 Max six months earlier — is a hand-picked comparison, but it shows why memory capacity matters here.

Gaming remains awkward for Arm. Titles lacking a native build run through Prism, Microsoft’s x86 emulator, and Nvidia says it is working with anti-cheat vendors whose software has long blocked online play on Windows on Arm.

RTX Spark matters even to buyers who will never own one. AMD’s Ryzen AI Max 400 chips already offer generous unified memory without emulation, and Nvidia’s arrival pushes AMD and Intel to compete harder on efficiency. Still, much early coverage came from events hosted by Nvidia and Microsoft, and independent battery and emulation testing is thin. At $2,600 to start, with the best configurations reserved for 128GB models, this is a tool for AI developers and well-funded creators rather than a mainstream laptop. If endurance and emulation hold up in third-party reviews, the case for a MacBook narrows for the first time in six years.