Adobe Rearchitects Photoshop, Premiere for NVIDIA RTX Spark

NVIDIA’s push to bring local artificial intelligence processing to mainstream Windows computers has secured its first major creative software overhaul. Adobe has rebuilt core components of Photoshop, Premiere and Substance 3D to run natively on the newly introduced RTX Spark system-on-chip, targeting a professional audience that has historically relied on cloud connectivity or bulky workstations for AI-assisted editing.

The optimization yields processing speeds twice as fast for AI tasks and standard editing workflows, according to Adobe Vice President Deepa Subramaniam. Rather than routing compute-heavy generative operations through remote servers, the updated applications execute them directly on the device.

For Photoshop users, the architectural shift translates to immediate visual feedback during complex operations. Applying effects such as Light Leaks or isolating subjects with the Remove tool now renders in real time. Previously, these generative fill and neural filter functions required either significant local GPU overhead or round trips to Adobe’s Firefly cloud infrastructure, introducing latency that disrupted iterative design processes.


Video editors working in Premiere gain comparable advantages across rendering, color grading and audio processing. High-resolution video timelines demand sustained throughput that traditional laptop processors often struggle to deliver without thermal throttling. By offloading AI-powered editing tasks to dedicated silicon, the software maintains performance stability during extended sessions.

Substance 3D Painter, widely used in game development and visual effects production, sees its heaviest texturing steps accelerated. Opening, saving and exporting large 3D projects — routines that routinely bottleneck artists working with high-polygon assets — now complete significantly faster.

The RTX Spark architecture occupies a distinct position within NVIDIA’s broader hardware strategy. Unlike the DGX Spark platform engineered for enterprise AI developers and corporate data centers, RTX Spark targets individual creators and independent studios operating on compact desktops or thin-and-light laptops. This segmentation mirrors the historical divide between NVIDIA’s Quadro workstation graphics cards and its consumer GeForce lineup, but applies the concept to integrated AI processing rather than discrete GPUs.

System-on-chip designs consolidate the central processor, graphics unit and neural processing components onto a single die. This integration reduces power consumption and physical footprint while increasing bandwidth between compute units — a critical requirement for generative AI models that constantly shuttle massive tensors between memory and processing cores. The approach parallels Apple’s M-series silicon strategy, which demonstrated that unified memory architectures could dramatically accelerate creative workloads on portable hardware.

Adobe’s decision to rearchitect its flagship suite signals confidence that local AI inference will become the default expectation for professional creative tools. As generative features expand beyond simple image manipulation into full scene composition and automated video assembly, the computational demands will only intensify. Hardware vendors and software developers are now locked into a cycle where each advancement forces the other to adapt.

The optimized versions of Photoshop, Premiere and Substance 3D are configured to activate their accelerated pathways automatically when an RTX Spark processor is detected by the host operating system.