NVIDIA’s chip for Windows laptops, RTX Spark, has launched in Korea

Sales began on October 16, 2026

It combines the CPU and GPU into one chip that shares up to 128GB of memory

In Korea, the 32GB model starts in the ₩5 million range

Hero image from the NVIDIA RTX Spark press release Hero image from the NVIDIA RTX Spark press release — Source: NVIDIA

What launched

NVIDIA logo NVIDIA logo

On October 16, NVIDIA announced that RTX Spark laptops had gone on sale worldwide

It’s four months after the chip was unveiled at Computex 2026 in June

The first wave is nine models from six manufacturers

Manufacturer Model
ASUS ProArt P14, ProArt P16
Dell XPS 16
HP OmniBook Ultra, OmniBook X 14
Lenovo Yoga Pro 9n
Microsoft Surface Laptop Ultra
MSI Prestige N16 Flip AI+, Prestige N16

In Korea, ASUS and MSI products launched first

The US starting price is $2,599 (₩3.64 million)

The specs, explained

RTX Spark laptop and the small desktop due in November RTX Spark laptop and the small desktop due in November — Source: NVIDIA

The chip is called N1X

It’s a version of the GB10 from the DGX Spark AI workstation, with specs adjusted for Windows laptops

Item N1X
CPU Arm 20 cores (10 X925 + 10 A725)
GPU Blackwell, 6,144 CUDA cores
Memory LPDDR5X up to 128GB
Memory bandwidth 300GB/s
AI compute 1 petaflop FP4
Power draw About 80W

The CPU was co-designed with MediaTek

It’s a unified memory architecture in which the CPU and GPU share memory

The weights of large AI models can all be loaded into the laptop’s memory

NVIDIA gave the 74-billion-parameter model Nemotron 3.5 Bolt as an example

It’s shrunk to fit in 24GB of memory and run directly on the laptop

There’s also a lower-tier N1 chip with 18 cores, but at launch it ships only in laptops

Prices in Korea

Bars showing Korean price ranges for RTX Spark laptops by memory Bars showing Korean price ranges for RTX Spark laptops by memory — Source: self-rendered based on reports by ZDNet Korea and The Elec

32GB is in the ₩5 million range, 64GB in the ₩7 million range, and 128GB in the ₩9 million range

According to The Elec, the MSI Prestige N16 is the lowest price in Korea at ₩5,799,000

That’s more than ₩2 million above the US starting price of ₩3.64 million!!

The US starting price is for a 24GB configuration, so it’s hard to compare directly with the Korean 32GB model

NVIDIA cited rising memory prices as the reason for the high prices

One official said they didn't expect memory prices to rise this much

Performance in reviews

Bar chart comparing 3DMark Time Spy Extreme scores Bar chart comparing 3DMark Time Spy Extreme scores — Source: self-rendered based on the Tech Times review roundup

Overseas reviews came out all at once on October 15–16

The verdict converged on strong GPU performance but relatively weak CPU performance

[Graphics] 3DMark Time Spy Extreme 6,060 points, 29% higher than Ryzen AI Max+ 395

[Games] Cyberpunk 2077 at about 70fps at 1080p Ultra, about 47fps at QHD

[CPU] Multi-core at Apple M4 Pro level, single-core below the M5 generation (HotHardware)

[Battery] 14 hours 35 minutes of video playback (HotHardware)

NVIDIA touted 100fps or more in the latest games at 1440p, but review figures are lower than that

The 300GB/s memory bandwidth falls short of the up to 672GB/s of the RTX 5070 laptop GPU

Graphics scores are higher than the integrated chips compared, but discrete GPUs still lead in gaming performance

Compatibility and what’s still to come

Microsoft logo Microsoft logo

RTX Spark laptops run on Windows on Arm

Programs for x86 run through Microsoft’s emulator Prism, with some loss of performance

The main factor limiting games is anti-cheat compatibility

Games with Arm anti-cheat on launch day include Counter-Strike 2 and Fortnite

Kernel anti-cheat games without an Arm build won’t run even through Prism!!

EA’s anti-cheat doesn’t support Arm yet either, and Call of Duty is slated for support in 2027

League of Legends and PUBG: Battlegrounds aren’t on the announced support list

If there’s a game you play often, be sure to check before buying~

For development, a PyTorch-CUDA build for Windows on Arm came out for the first time

97% of the 15,000 most popular Python packages have Arm versions

A small desktop with the same chip launches in November

Sources