Independent editorial site covering NVIDIA RTX Spark. Not affiliated with NVIDIA Corporation.
NVIDIA RTX Spark fuses a Grace Arm CPU, a Blackwell RTX GPU, and up to 128 GB of unified memory into a single superchip — delivering 1 petaflop of AI on a slim Windows laptop or a palm-sized desktop.
RTX Spark unites a 20-core Grace CPU and a 6,144-core Blackwell RTX GPU through NVIDIA NVLink-C2C, sharing a single 128 GB memory pool. MediaTek co-designed the Arm CPU for industry-leading efficiency.
5th-gen Tensor Cores, FP4 precision
Co-designed with MediaTek
≈ 1,000 TOPS with sparsity
Shared CPU + GPU via NVLink-C2C
1M-token context, on-device
Ray tracing, DLSS 4.5, Reflex
Agents like Hermes Agent and OpenClaw can now run securely on your primary device — executing tasks in Windows apps, reasoning across workflows, and semantically searching your local files.
NVIDIA OpenShell runtime routes requests to local models based on your privacy policy and can mask personal info before queries hit the cloud.
Microsoft and NVIDIA collaborated on identity, containment, and policy controls so agents run isolated under user control.
Premiere and Photoshop are being rearchitected for RTX Spark — up to 2× faster AI, editing, coloring, and effects.
The Arm-on-Windows race has a new entrant. Here's how the top SKU stacks up against Apple Silicon, Qualcomm, and the incumbents.
| Chip | Architecture | Memory | AI peak | GPU | Stack |
|---|---|---|---|---|---|
| NVIDIA RTX Spark | Grace Blackwell (Arm) | Up to 128 GB unified | ≈1,000 TOPS (FP4, sparse) | 6,144 CUDA · 5th-gen Tensor | CUDA · RTX · DLSS · TensorRT |
| Apple M4 Max | Apple Silicon (Arm) | Up to 128 GB unified · 546 GB/s | 38 TOPS (Neural Engine, INT8) | 40-core Apple GPU | Metal · MLX · Core ML |
| Qualcomm Snapdragon X2 Elite | Oryon (Arm) | Up to 64 GB LPDDR5X | 80 TOPS (NPU) | Adreno integrated | DirectML · Copilot+ runtime |
| Intel Core Ultra 200 / AMD Ryzen AI | x86 | Discrete RAM + dGPU VRAM | 40–50 TOPS (NPU) | Discrete GeForce / Radeon | DirectML · CUDA (on dGPU) |
Peak AI figures are not apples-to-apples: NVIDIA's 1,000 TOPS is FP4 with sparsity; Apple's 38 TOPS is INT8 on the Neural Engine; Qualcomm's 80 TOPS is the NPU only. CUDA + TensorRT remains NVIDIA's structural advantage.
"The PC is being reinvented. For forty years, you launched apps. Click. Type. With RTX Spark and Microsoft Windows, you ask — and the PC does the work. This is the new PC. The personal AI computer."
"Our goal is to deliver unmetered intelligence to every home and every desk with Windows. RTX Spark marks a real breakthrough towards that vision."
"The best creative work in the world happens in Adobe tools — and our expanded partnership with NVIDIA and Microsoft will make those experiences faster and more powerful than ever."
"Creators shouldn't have to choose between portability and performance. With RTX Spark, Dell is delivering RTX performance and massive unified memory in the XPS 16 Creator Edition."
"Highly optimized models running locally through llama.cpp with RTX Spark's AI performance will unleash the next wave of personal, private agents."