> On the opening day of COMPUTEX 2026 in June, Jensen Huang officially unveiled the N1X. But this chip, dubbed the "x86 killer," benchmarks on par with Apple's 2023 M3 Max. Multiple delays, performance controversies, and ecosystem challenges make this gamble increasingly uncertain.
---
1. What Is the N1X?
The N1X is NVIDIA's first consumer-grade Arm-architecture PC processor SoC, co-developed with MediaTek. It is essentially a mobile adaptation of the GB10 "Grace Blackwell" superchip used in the DGX Spark mini PC.
Core Specifications
| Component | Specification | |------|------| | CPU | 20-core heterogeneous (10x Cortex-X925 + 10x Cortex-A725), up to 4.0GHz | | GPU | Blackwell architecture, 6144 CUDA cores (scale comparable to RTX 5070) | | Memory | 128GB LPDDR5X unified memory, 256-bit, 301GB/s bandwidth | | AI compute | 180-200 TOPS | | Process | TSMC 3nm | | TDP | 45-80W | | Partners | MediaTek (CPU design) + Microsoft (Windows on Arm ecosystem) |
Key design decision: Unified Memory Architecture (UMA) — CPU and GPU share a 128GB memory pool. CUDA cores can access system memory directly without going through PCIe to discrete VRAM — a major architectural advantage for AI inference and development workloads.
---
2. Benchmark Reality: Outdated Before Launch?
Geekbench 6 Results
| Processor | Single-core | Multi-core | Cores | Released | |--------|------|--------|------|----------| | NVIDIA N1X | 3096 | 18837 | 20 | 2026 | | Apple M3 Max | 3124 | 18920 | 14 | Nov 2023 | | Apple M4 | ~3800 | ~15000 | 10 | 2024 | | Apple M4 Pro | ~4000 | ~22000 | 14 | 2024 | | Apple M4 Max | ~4000 | ~25000+ | 16 | 2024 | | Qualcomm X Elite | ~2800 | ~15000 | 12 | 2024 | | Intel Core Ultra 9 285HX | ~2800 | ~16000 | 24 | 2025 | | AMD Ryzen AI Max+ 395 | ~2900 | ~17000 | 16 | 2025 |
The awkward facts:
- N1X (2026) single/multi-core performance is roughly on par with M3 Max (2023)
- 20 cores barely matching 14 cores — Apple's architectural efficiency negates the core count advantage
- ~30% single-core gap vs. the M4 series (2024)
- ~35% multi-core gap vs. M4 Max
- Game compatibility: Most Windows games are x86 and run through emulation with 10-30% performance loss
- Adobe suite: Arm versions of Photoshop/Premiere exist but lag x86 in feature completeness and plugins
- Professional software: Native Arm support for AutoCAD, MATLAB, SolidWorks is slow
- Drivers: Low Arm coverage for peripherals (printers, scanners, pro audio interfaces)
- Apple has proven Arm PCs viable (market share nearly doubled in three years)
- Windows on Arm is maturing fast under Microsoft's push
- CUDA is irreplaceable in the AI era
- Unified memory gives structural advantages for AI workloads
- Repeated delays expose newcomer inexperience in CPU design
- Benchmarks match a three-year-old M3 Max; clear architectural efficiency gap
- Windows on Arm still has many compatibility gaps
- Qualcomm X Elite already proved "Arm PCs are good but don't sell"
- At late-2026 launch, it faces Intel Nova Lake and AMD Zen 6
- SemiAccurate: Inside the repeated NVIDIA N1/N1X delays
- TechPowerUp: NVIDIA N1X Geekbench score analysis
- Tweaktown: N1X vs AMD Strix Halo performance comparison
- ITC.ua: Detailed N1X specs and competitor comparison
- UDN: N1X gaming laptop positioning and Alienware partnership
Important Caveats
1. Pre-release data: Scores collected June 2025 on unoptimized Ubuntu 24.04; hardware and drivers untuned 2. Capped clocks: Test ran at 2.81GHz base, but data shows the chip can boost to 4.05GHz; production units may run faster 3. Windows optimization: Final product runs Windows on Arm; Microsoft's tuning may improve performance 4. GPU is the star: Geekbench is CPU-only and doesn't reflect the N1X's biggest selling point — the Blackwell GPU and CUDA ecosystem
---
3. Timeline of Delays
| Date | Event | |--------|------| | Oct 2023 | Reuters first reports NVIDIA and AMD developing Arm PC chips, expected 2025 | | Early 2025 | Subtle defects found after tape-out; NVIDIA engineers fixed without redesign, slipping to early 2026 | | Jun 2025 | Geekbench scores leak (3096/18837), chip identified as "NVIDIA N1x" | | Jul 2025 | SemiAccurate reports new chip issues may require silicon changes, pushing launch to late 2026 | | Jun 2026 | Official unveiling at COMPUTEX 2026; mass products still due late 2026 or CES 2027 |
Root cause of delays: Architectural defects found during engineering validation required silicon layout changes — typically meaning a re-tape-out that directly affects mass production. For a GPU-first company like NVIDIA, CPU design complexity clearly exceeded expectations.
---
4. Market Positioning: Who Does It Compete With?
Direct Competitors
| Competitor | Positioning | Strengths | Weaknesses | |------|------|------|------| | Qualcomm X Elite | Windows on Arm laptops | First-mover, long battery life, Copilot+ certification | Poor gaming compatibility, few native apps, mediocre CPU performance | | Apple M4 series | MacBook/iPad Pro | Dominant single-core, closed ecosystem loop, excellent efficiency | Closed ecosystem, no Windows | | Intel Core Ultra 200HX | High-performance gaming laptops | x86 compatibility, mature gaming ecosystem | High power, weak iGPU, lagging AI compute | | AMD Ryzen AI Max | High-performance APU | x86 compatibility, strong iGPU | High power, lagging AI ecosystem |
N1X Differentiators
1. CUDA ecosystem at the edge: 6144 CUDA cores mean full CUDA applications run locally — PyTorch, TensorFlow, Blender, CUDA-accelerated video encoding. Neither Qualcomm X Elite nor Apple M-series offers this. 2. AI development workstation: 128GB unified memory + 200 TOPS could make N1X laptops mobile AI workstations capable of running 200B-parameter models. DGX Spark has already proven this path. 3. Gaming potential: Blackwell + DLSS 4 could deliver near-120W RTX 4070 laptop gaming performance at ~65W. 4. Launch partners: Dell (Alienware) and Lenovo confirmed; Alienware 16 Aurora may debut first.
---
5. Ecosystem Challenge: The Windows on Arm Dilemma
The N1X's biggest challenge isn't hardware — it's software:
Known Issues
Microsoft's Exclusivity Deal
Key context: Microsoft and Qualcomm had an exclusivity agreement for Windows on Arm that expired at the end of 2024. That's why NVIDIA and AMD began planning Arm PC chips in 2023 — waiting for that window to open.
---
6. NVIDIA's Real Motives
1. Pushing AI compute to the edge: Datacenter GPU growth is slowing; NVIDIA needs consumer AI. The N1X is essentially a "portable DGX Spark." 2. CUDA lock-in: Getting consumers accustomed to CUDA on laptops reinforces NVIDIA's AI development dominance. 3. Countering Apple M-series: Apple proved Arm PCs viable and won over creators and developers; NVIDIA wants them back via CUDA. 4. The MediaTek partnership: NVIDIA lacks CPU design expertise; MediaTek brings Arm core design and manufacturing experience — a "GPU giant + mobile SoC giant" joint strike.
---
7. x86 Killer — or a Different Story?
Case for "Revolution"
Case for "Just a Story"
Most Likely Outcome
The N1X won't kill x86, but it could be a key catalyst for the Windows on Arm ecosystem. Its success depends not on beating Apple's M-series, but on convincing AI developers, creators, and gamers to accept a "CUDA-first" Arm laptop.
For NVIDIA, the N1X is more of an ecosystem placeholder — even with modest initial sales, establishing the perception that "NVIDIA can make PC processors" paves the way for future iterations (N2, N3...).
---