NVIDIA RTX Spark: All you need to know

“iPhone made people think differently about smartphone, RTX Spark is going to make people think differently about the PC”: Nvidia.

Nvidia Spark is here to usher 'a new paradigm of personal computing' that puts agentic AI front and centre on a Windows PC. The RTX Spark technology puts a Blackwell RTX GPU and an Arm CPU on one chip, sharing up to 128GB of memory, in Windows lapto built for local AI agents. This could change the workflow of creatives in a host of different ways – including opening up how you interact with creative software and accelerating a spectrum of creative workflows.

I went to the launch day last week, and saw a room full of demos showcasing the tech's potential for creative workflows and gaming – including video editing, CAD, community game modding, productivity overhauls and more. It's clear Nvidia and its partners are committed to the future of agentic computing: but what does this mean for creatives?

Firstly, what is RTX Spark?

NVIDIA announced RTX Spark on 31 May at GTC Taipei, calling it a superchip for Windows PCs built around personal AI agents. It joins a Blackwell RTX GPU, with 6,144 CUDA cores and fifth-generation Tensor Cores, to a 20-core Arm-based Grace CPU over NVIDIA's NVLink-C2C link. MediaTek collaborated on the CPU design.

The CPU and GPU share up to 128GB of unified memory, and NVIDIA quotes up to 1 petaflop of AI compute. Laptops from ASUS, Acer, Dell, HP, Lenovo, Gigabyte, Microsoft Surface and MSI are all up for release, in 14 to 16-inch designs as thin as 14mm.

It runs Windows 11 on Arm. Apps with a native Arm version run directly, and everything else goes through Microsoft's Prism emulation layer. NVIDIA also says the laptops deliver equivalent performance on battery and plugged in – which is one of the biggest improvements compared to its older laptops.

Why this matters for creatives

A normal laptop GPU is stuck with its own small pool of video memory. On RTX Spark the GPU shares the CPU’s pool of up to 128GB, so big scenes and models aren’t capped. NVIDIA says that means rendering 90GB+ 3D scenes, editing 12K 4:2:2 video and running 120-billion-parameter language models locally. The GPU’s share is adjustable in the NVIDIA App, and NVIDIA’s largest local model, Qwen3.8-Flash-Next, needs a 128GB machine with at least 78GB given to the GPU.

The second important part for creators is privacy and cost. Local models mean footage and client assets don't have to leave the machine, and there are no per-generation cloud fees. The data example I saw at the launch is that the 61m tokens it would take to complete a heavy creative workflow will cost $20 in electricity, as opposed to the $1200 token cost it would take in the cloud.

As this is hybrid intelligence, and the local AI can't yet do everything, there's the NeMo Switchyard Verifier, which kicks in when the task is sent to the local LLM. This decides if the task can stay local, or needs to be sent up to the cloud.

The software

One of the most impressive parts of the launch is how many apps Nvidia has worked with to ensure optimisation with the Spark technology. More than 100 Windows software providers, including Adobe, Blackmagic Design, Blender, CapCut, ComfyUI and OTOY, are backing the platform. With Adobe actually rewriting its Premiere Pro and Photoshop apps from the ground up to suit Spark, it's clear many software companies see the potential here.

I saw a CAD workflow move from proposal to fully rendered 3D design in 26 minutes using AI. This is a process that usually takes weeks. For solo creatives, this kind of tech is potentially a good thing – it will allow you to create whole studio's worth of work individually. But as usual with AI advancements, the impact on team structure and the need for jobs will be immense. When we're talking about some creative workflows, the old Adobe-style argument is that that just means creatives can do more – but in this instance, how many buildings really need to be built?

Anyway, let's quickly look at the impacts on other creative apps:

• Adobe. NVIDIA says Adobe is rearchitecting Photoshop and Premiere around RTX Spark, promising up to 2x faster AI, editing, colour and effects. Premiere gets a new video pipeline built on unified memory, the GPU and TensorRT. Photoshop gets an engine built around GPU-accelerated compositing, for live filters, HDR and natural brushes. Substance 3D Painter and Stager will run natively, and updates are due as the laptops go on sale. Adobe's public betas already include TensorRT-accelerated Object Mask, Enhance Audio, Scene Edit Detection and Auto Reframe in Premiere, a GPU-accelerated colour mode, and on-device object removal in Photoshop.
• Blender. Version 5.3 adds DLSS 4.5 Ray Reconstruction as a denoiser for the Cycles viewport, aiming for a near-final rendered view with far less noise. It is currently in the 5.3 alpha builds.
• DaVinci Resolve. Blackmagic Design says it expects portable RTX Spark laptops to help on-the-go production. Resolve Studio 21 has a native Arm build.
• ComfyUI. New RTX Video nodes upscale resolution, interpolate frames (up to 4x frame generation) and convert SDR to HDR, so artists can generate at lower resolution and finish higher.

What can the AI agents do for creatives?

Nvidia frames the new PC tech as moving from "tool to teammate". Essentially, you can use a trio of AI agents as a team of assistants, who will prepare assets and set up workflows. Two open-source agents, Hermes Agent and OpenClaw, are getting native Windows apps that can carry out tasks in Windows applications, work across apps, generate images and video, and search local files. Adobe says it will extend Premiere and Photoshop so people can create, edit and design with Windows agents.

That means that the most powerful advancement is that these AI agents will interface with these creative apps for you. Yes, you read that right. You don't even have to go into the app to utilise its tools and features – you can describe to the agent what you want it to do, and it will open the app, individual file, and apply the edits or create the work natively within the application.

This is exciting and alarming, depending on how you look at it. In a world in which creatives are increasingly expected to be multi-disciplinary, this functionality will allow you to use apps you don't have specialist knowledge of. Alarming because, well, obviously it's another shortcut on human skill. However, in line with much of the creative AI narrative, the sell is that the human remains the director, the refiner, the ultimate creative – the technology is there to speed it up, make it simpler, enable more creativity and eliminate busy work. Where you sit on this probably fluctates, and it isn't a simple conversation.

The agents can also work as admin/productivity assistants. They can connect to your files, email account, calendar and more and stay on top of your daily workflow. This is probably appealing for the majority of creatives, as this is the true busy work we'd all like to eliminate.

Georgia has worked on Creative Bloq since 2018, and has been the site's Editor since 2023. With a specialism in branding and design, Georgia is also Programme Director of CB's award scheme – the Brand Impact Awards. As well as immersing herself with the industry through attending events like Adobe Max and the D&AD Awards and steering the site's content streams, Georgia has an eye on new commercial opportunities and ensuring they reflect the needs and interests of creatives.