NVIDIA N1X Confirmed: Two RTX Spark SoC Configurations
The rumored split between NVIDIA’s N1 and N1X chips has taken a different turn.
At IFA, NVIDIA officially confirmed that the N1X SoC powering its RTX Spark platform will be available in two hardware configurations. Both use the N1X name but offer different CPU, GPU, and unified-memory specifications.
The flagship configuration combines a 20-core Grace CPU with a 6,144-core Blackwell GPU and supports up to 128GB of LPDDR5X unified memory.
A second configuration reduces the CPU to 18 Grace cores and the GPU to 5,120 CUDA cores, with 24GB to 32GB of unified memory.
Both configurations target Windows 11 on Arm and fall under Microsoft’s Copilot+ PC ecosystem.
🧩 N1X Comes in Two Specification Tiers #
Earlier industry reports suggested that NVIDIA’s RTX Spark platform could use two different SoCs, referred to as N1 and N1X.
NVIDIA’s official positioning is different: both tiers are designated N1X, representing different configurations of the same underlying chip design.
The two versions can be summarized as follows:
| Specification | N1X High-End | N1X Second Tier |
|---|---|---|
| CPU | 20-core Grace | 18-core Grace |
| GPU architecture | Blackwell | Blackwell |
| CUDA cores | 6,144 | 5,120 |
| Unified memory | 24GB–128GB | 24GB–32GB |
| Memory type | LPDDR5X | LPDDR5X |
| Product targets | Laptops + mini PCs | Laptops |
| OS | Windows 11 on Arm | Windows 11 on Arm |
| AI platform | Copilot+ PC | Copilot+ PC |
This approach allows NVIDIA and its OEM partners to build products across different performance and price segments without introducing a separately named N1 platform.
🚀 Flagship N1X Combines Grace and Blackwell #
The highest-end N1X configuration is the most interesting version of the platform.
It integrates:
- 20-core NVIDIA Grace CPU
- 6,144 CUDA-core Blackwell GPU
- 24GB to 128GB LPDDR5X unified memory
The architecture combines CPU and GPU resources into a single SoC rather than relying on a conventional discrete CPU + GPU arrangement.
NVIDIA N1X
│
┌────────────┴────────────┐
│ │
20-Core Grace Blackwell GPU
CPU 6,144 CUDA
│ │
└────────────┬────────────┘
│
LPDDR5X Unified
Memory
24GB–128GB
The large unified-memory configuration is particularly relevant to local AI workloads, where model size and working-set capacity can become a limiting factor.
Instead of separating system RAM and GPU VRAM into distinct pools, the unified-memory design allows the CPU and GPU to operate from the same memory resource.
💻 Second N1X Targets More Compact Laptops #
The second N1X configuration is designed for a narrower product segment.
It reduces the CPU to 18 Grace cores and the Blackwell GPU to 5,120 CUDA cores.
Memory capacity is also significantly lower:
N1X Second Configuration
18-Core Grace CPU
+
5,120 CUDA-Core Blackwell GPU
+
24GB–32GB LPDDR5X Unified Memory
│
▼
Laptops
Unlike the high-end configuration, this version is currently positioned exclusively for laptops.
The reduced configuration gives OEMs another way to build RTX Spark systems while maintaining the same fundamental Grace + Blackwell SoC architecture.
🔬 Why Are There Two N1X Configurations? #
Industry reports have attributed the difference between the two configurations to silicon binning during manufacturing.
In semiconductor production, chips manufactured from the same design can exhibit different levels of achievable performance or functional yield. Manufacturers can then classify the resulting dies into different product tiers.
A simplified example would be:
N1X Silicon
│
Manufacturing
│
┌───────┴───────┐
│ │
Higher-bin Lower-bin
│ │
▼ ▼
20C / 6144 18C / 5120
CUDA CUDA
│ │
▼ ▼
High-end tier Laptop tier
However, NVIDIA has not officially confirmed that silicon binning is the reason for the two configurations.
Therefore, binning should currently be treated as an industry explanation rather than an officially documented architectural detail.
🧠 RTX Spark Targets Local AI and RTX Graphics #
NVIDIA positions RTX Spark as a single-chip computing platform that combines AI acceleration and RTX graphics for Windows PCs.
The platform is intended to span multiple form factors, including:
- Thin-and-light laptops
- High-performance laptops
- Compact desktop systems
- Mini PCs
The goal is broader than simply providing gaming performance.
NVIDIA is targeting workloads such as:
RTX Spark
│
├── Local AI development
├── AI inference
├── Creative applications
├── RTX graphics
├── Gaming
└── Personal AI agents
This positions RTX Spark closer to a general-purpose local AI platform than a conventional gaming laptop GPU.
🤖 Mini PCs Could Run Always-On AI Agents #
The mini PC version of the high-end N1X configuration is particularly interesting for local AI.
Because the system combines a high-core-count Grace CPU, a large Blackwell GPU, and substantial unified memory, compact desktop systems can potentially operate local AI models continuously.
NVIDIA specifically positions these systems for 24/7 personal AI-agent workloads.
The concept is different from using a cloud-based AI assistant:
Cloud AI
User
│
▼
Internet
│
▼
Remote GPU cluster
│
▼
AI response
Local AI Agent
User
│
▼
RTX Spark
│
├── Grace CPU
├── Blackwell GPU
└── Unified Memory
│
▼
Local execution
A local agent can operate closer to the user’s applications and data, potentially reducing dependence on cloud inference for suitable workloads.
🪟 Windows 11 on Arm and Copilot+ PC #
N1X-powered RTX Spark systems will run Windows 11 on Arm and belong to the Copilot+ PC category.
This makes NVIDIA’s platform part of the broader transition toward Arm-based Windows PCs.
The combination is notable because N1X brings NVIDIA’s CUDA and RTX software ecosystem into an Arm-based PC platform while retaining a unified CPU/GPU memory architecture.
For developers, this could create an interesting local AI environment combining:
- Arm CPU execution
- Blackwell GPU acceleration
- CUDA software
- RTX features
- Large unified-memory configurations
- Windows desktop applications
📅 RTX Spark Systems Are Expected in October #
According to the current schedule, the first N1X-powered RTX Spark systems are expected to launch in October.
The initial OEM partner lineup includes:
- Acer
- ASUS
- Dell
- MSI
- HP
- Microsoft
- Lenovo
NVIDIA and its partners have not yet disclosed final system pricing across the different configurations.
Actual pricing will depend on the specific OEM, memory capacity, display and chassis configuration, and regional availability.
🔮 N1X Could Become NVIDIA’s Local AI PC Foundation #
The official N1X specifications make NVIDIA’s RTX Spark strategy clearer.
Rather than creating a completely separate chip for every product class, NVIDIA is using the N1X name across multiple configurations:
N1X
│
┌───────────┴───────────┐
│ │
High-End N1X Second N1X
│ │
20C Grace + 6144 18C Grace + 5120
Blackwell Blackwell
│ │
24GB–128GB 24GB–32GB
│ │
┌────┴────┐ │
│ │ │
Laptops Mini PCs Laptops
The high-end variant’s combination of 6,144 CUDA cores and up to 128GB of unified memory gives RTX Spark a substantially different profile from traditional thin-and-light AI PCs.
Meanwhile, the smaller N1X configuration provides OEMs with a lower-tier option without introducing a separate N1 product name.
The bigger question is how these configurations will perform in real-world AI inference, creative workloads, and gaming once retail systems arrive.
For now, one point is clear: NVIDIA’s N1X is not a single fixed specification. It is a multi-tier SoC platform designed to bring Grace CPU, Blackwell GPU, and unified-memory computing into the next generation of Windows Arm PCs.