NVIDIA's Arm chip effort just got confirmed by a Cinebench benchmark. On July 18, a processor called NVIDIA N1X quietly appeared in the Cinebench 2026 benchmark database. Benchmark scores, honestly, usually aren't worth much discussion—today A beats B, tomorrow C overtakes A, the day after D surpasses C, and the cycle repeats. But this time is different.

💡 What You Will Learn

NVIDIA's Arm chip effort just got confirmed by a Cinebench benchmark. On July 18, a processor called NVIDIA N1X quietly appeared in the Cinebench 2026 benchmark database. Benchmark scores, honestly, u

📜 Table of Contents

• 20-Core Arm Chip Debuts: Benchmark Score Matches M3 Max
• 10 Cortex-X925 Big Cores + 10 Cortex-A725 Little Cores: How NVIDIA Chose Its Arm Configuration
• It's Not Just About the Benchmark Score—The GPU Is NVIDIA's Real Ace
• Achieving This Score at 50W—Power Efficiency Is the Real Game-Changer
• Starting at $1,800—What Makes NVIDIA's Arm PC So Expensive
• Launching This Fall, Led by the Surface Laptop Ultra
• Windows on Arm Finally Has a Real Contender

NVIDIA's Arm Chip Effort Just Got Validated by a Cinebench Score

On July 18th, a processor called the NVIDIA N1X quietly appeared in the Cinebench 2026 benchmark database.

Benchmark scores usually aren't worth much discussion—today A beats B, tomorrow C beats A, the day after D beats C, and so on. But this time is different, because this is the first time an NVIDIA self-developed Arm PC chip has appeared in a public database.

The score: 5,771 points. The 14-core Apple M3 Max scores 5,722. They're practically neck and neck.

20-Core Arm Chip Debuts: Benchmark Score Matches M3 Max

This chip was found inside a Microsoft Surface Laptop Ultra engineering unit. Cinebench couldn't even identify it properly, showing "JMJWOA-Generic-CPU," and only detected a frequency of 1.01 GHz—clearly a recognition issue with the pre-release platform, not the actual clock speed.

The leaker uncovered a hidden high-performance mode. When enabled, it allows PL1 80W / PL2 95W, with both CPU clusters running at 2.8 GHz simultaneously, drawing around 50W of actual power.

Under a 10-minute sustained load test, the N1X scored 5,771 in multi-core, about 1% higher than the M3 Max's 5,722; in single-core, it scored 540, about 4% lower than the M3 Max's 562. The 16-core M3 Max still leads by 14%, but that's the top-tier version with a 40-core GPU, and it costs significantly more.

10 Cortex-X925 Big Cores + 10 Cortex-A725 Little Cores: How NVIDIA Chose Its Arm Configuration

The N1X's configuration is 10 Cortex-X925 cores + 10 Cortex-A725 cores. The X925 is Arm's current most powerful off-the-shelf big core, with a 36% IPC improvement over the previous generation X4. NVIDIA went straight for 10 of them, plus 10 A725 efficiency cores, for a total of 20 cores.

This differs from Apple's approach. The 14-core M3 Max has 10 performance cores + 4 efficiency cores, while the N1X has more than double the efficiency cores. Its ability to match the M3 Max in multi-core scores comes from the sustained load capability stacked by more little cores.

However, the N1X's CPU was designed with help from MediaTek—NVIDIA and MediaTek announced their collaboration back in 2024, with the N1 series being a jointly developed Arm PC chip. NVIDIA's strength lies in GPU and AI acceleration, while the CPU side is handed to MediaTek—a clear division of labor. MediaTek has over a decade of experience in mobile Arm chip design, and NVIDIA leveraging that to enter the market is far faster than starting from scratch. From another angle, this collaboration also signals that NVIDIA isn't planning to follow Apple's fully in-house approach, but rather using a combination of Arm's public architecture + MediaTek's design + its own GPU to enter the market quickly.

It's Not Just About the Benchmark Score—The GPU Is NVIDIA's Real Ace

The GPU part of the N1X holds the real killer feature: 48 SMs, 6,144 CUDA cores, matching the scale of the desktop RTX 5070. NVIDIA officially states this chip can run AAA games at 1440p with over 100fps.

Combined with 128GB of unified memory (LPDDR5X, 16 channels) and 1 petaflop of FP4 AI compute, NVIDIA's play is clear: not to go head-to-head with Apple on CPU benchmarks, but to differentiate through GPU + AI ecosystem. When NVIDIA announced RTX Spark at Computex 2026, CEO Jensen Huang's exact words were "this is a personal AI supercomputer," positioning it in direct competition with Apple's Pro series.

Achieving This Score at 50W—Power Efficiency Is the Real Game-Changer

The N1X's default power settings are conservative—PL1 is only 37.5W, and PL2 is only 64W. In the hidden high-performance mode, it achieved 5,771 points at around 50W actual power draw.

For comparison, the Apple M3 Max 14-core has a TDP between 40-60W, putting both on the same level of power efficiency. On the x86 side, Intel Lunar Lake at 28W scores around 4,000 points, while AMD Strix Point at 45W scores around 4,500-5,000. The N1X, at 50W, matches x86 processors that require higher power draw.

Behind this is the natural advantage of the Arm architecture, and it's also the confidence behind NVIDIA's claim at Computex that "this is the most efficient PC platform ever." More critically, the N1X is still an engineering sample—the final version has room for optimization in both power and frequency. If the production version can push frequencies a bit higher or slightly loosen the power limits, the lead could widen further.

Starting at $1,800—What Makes NVIDIA's Arm PC So Expensive

The N1X's biggest challenge isn't performance—it's price.

Morgan Stanley predicts the N1 base model will start at $1,800, with the N1X top-tier potentially reaching $2,900. As a flagship device, the Surface Laptop Ultra will likely be priced above $2,500.

To quote PCWorld directly: the RTX Spark's price "will make your wallet hurt." But this time, NVIDIA has three cards to play that differ from the Snapdragon X Elite:

• CUDA Ecosystem: Not reinventing the wheel, but bringing the desktop GPU software ecosystem directly to Arm laptops. Developers' existing CUDA toolchains, AI frameworks, and rendering pipelines can migrate seamlessly. This is something neither Snapdragon nor Apple can offer—Snapdragon lacks a GPU compute ecosystem, and Apple's Metal is closed and niche
• Local AI Compute: 1 petaflop of FP4 compute lets AI developers truly run large models on a laptop without needing to connect to the cloud every time. NVIDIA has also partnered with Microsoft on the OpenShell framework, including a security framework for local AI agents
• Deep Microsoft Integration: Windows 11 26H1 is specifically optimized for the N1X—Cinebench previously froze at 76% loading due to compatibility issues, but after the update, it ran to completion in one go. Microsoft is also positioning the Surface Laptop Ultra as its flagship, coordinating from the system layer down to the hardware layer

Launching This Fall, Led by the Surface Laptop Ultra

NVIDIA has confirmed that RTX Spark laptops will launch this fall. Leading the charge is the Microsoft Surface Laptop Ultra, featuring a 15-inch mini-LED display, up to 128GB of unified memory, and an enhanced cooling system. It was already showcased at Computex 2026, and ZDNet's hands-on review concluded that "Microsoft is getting serious."

The partner list includes Dell, Asus, MSI, HP, and Lenovo—essentially all major PC makers are on board. NVIDIA has also revealed the RTX Spark roadmap, committing to three generations of continuous iteration. This signals it's not a one-off experiment but a long-term strategy—NVIDIA is gearing up for a sustained battle against Intel, AMD, and Apple in the PC market.

Windows on Arm Finally Has a Real Contender

Looking back at the history of Windows on Arm: the first-generation Surface RT in 2012 ran on Tegra 3, with abysmal performance; then the Snapdragon 8cx series arrived, with adequate performance but still poor software compatibility; in 2024, the Snapdragon X Elite finally brought performance close to Apple's M-series level, but GPU and AI compute were still far behind.

NVIDIA's entry this time is different from all previous players. It's not a mobile chip company testing the PC waters—it's the world's most valuable chip company, armed with its GPU ecosystem, AI ecosystem, and the deepest partnerships in the PC industry, coming to fight a serious battle.

At the end of the day, NVIDIA making Arm chips itself signals a structural shift in the PC industry. A GPU company, at 50W, has caught up to the moat Apple spent five years building with its own silicon. The next question is: once it actually launches this fall, how many users can it take from MacBook? At a starting price of $1,800, would you consider buying one?

【Image Suggestions】fig1-specs-table.png (N1X vs N1 spec comparison table) placed above the "It's Not Just About the Benchmark Score" table; fig2-power-efficiency.png (power comparison chart: N1X 50W vs M3 Max vs Lunar Lake vs Strix Point) placed above the "Achieving This Score at 50W" subheading; fig3-ecosystem.png (NVIDIA's three cards: CUDA/AI/Microsoft integration) placed above the "Starting at $1,800" subheading.