NVIDIA has stopped being just a graphics card company. With RTX Spark, a new line of Arm-based Windows 11 PCs built around what the company calls the RTX Spark Superchip, NVIDIA is shipping its own processor into laptops and compact desktops for the first time at consumer scale. The rollout begins with laptops launching in October 2026 through ASUS, Dell, HP, Lenovo, Microsoft, and MSI, with entry pricing starting at $2,599, while compact RTX Spark desktops follow in November 2026, pairing NVIDIA silicon with an unusual partner: MediaTek, the Taiwanese chipmaker best known for budget smartphone processors.
The announcement matters beyond spec sheets. NVIDIA built its market value on GPUs sold into other companies’ PCs and servers. RTX Spark is the first time it has put its name directly on a Windows machine’s main processor, competing head-on with Qualcomm, AMD, Intel, and Apple in a market NVIDIA has never directly sold into. NVIDIA’s own product page frames RTX Spark as the start of “the next era of Windows PCs,” and Microsoft backed that framing in a Windows Experience blog post published back in May 2026, months ahead of the actual shipping window. NVIDIA and its partners made that window concrete on October 7, 2026, when preorders opened across six confirmed launch partners, ASUS, Dell, HP, Lenovo, Microsoft, and MSI, ahead of October 16, 2026, the date NVIDIA confirmed RTX Spark laptops became globally available, including Microsoft’s Surface Laptop Ultra and HP’s OmniBook Ultra 16.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
What Is NVIDIA RTX Spark, Exactly?
The RTX Spark Superchip, Defined
RTX Spark is NVIDIA’s branding for a family of processors built around Arm CPU cores paired with a Blackwell-generation GeForce RTX graphics engine, designed for Windows 11 laptops and small-form-factor desktops. The chip was developed with MediaTek, and MediaTek’s own product page describes its role as enabling the platform for mainstream Windows PCs rather than just workstation-class machines.
From N1X Leaks to an Official Name
Before NVIDIA settled on the RTX Spark name, the chip circulated in leaks and engineering reports under the codename N1X. Tom’s Hardware tracked those leaks for months, and in a report on NVIDIA’s market entry, the outlet confirmed that N1X and RTX Spark describe the same underlying silicon family, not two separate products. NVIDIA itself has not used the N1X label in any official material. Readers searching for “N1X” and “RTX Spark” are looking for the same chip under two different names, one leaked and one official.
Confirmed Specs vs. Leaked Numbers on the RTX Spark Superchip
NVIDIA’s official RTX Spark materials confirm the chip combines a Grace-based Arm CPU complex with Blackwell-based RTX graphics and targets Windows 11, with NVIDIA now listing two confirmed tiers: an 18-core Grace CPU paired with 5,120 Blackwell GPU cores, and a 20-core Grace CPU paired with 6,144 GPU cores, both topping out at 128GB of unified LPDDR5X memory. Microsoft has confirmed exact numbers for its top-tier configuration: the RTX Spark Dev Box, NVIDIA’s compact desktop form factor arriving in November 2026, ships with the 20-core Grace CPU, a 6,144-core GPU, 128GB of unified LPDDR5X memory, and a 2TB SSD for $5,999.99. NVIDIA still has not published one universal spec sheet covering every OEM’s exact clock speeds and power limits, so those cheaper-tier details from other partners still trace back to engineering-sample leaks reported by Tom’s Hardware and VideoCardz rather than official press materials.
The High-End Configuration
Microsoft’s RTX Spark Dev Box, the compact desktop variant NVIDIA has scheduled for November 2026, confirms the top-tier configuration: a 20-core Grace CPU and a 6,144-core GPU, paired with 128GB of unified LPDDR5X memory, a 2TB SSD, and a $5,999.99 price tag. That unified-memory approach mirrors the design NVIDIA already shipped in DGX Spark, whose own 128GB of unified LPDDR5X memory was confirmed by WindowsForum reporting in August 2026, letting large AI models and game assets, including 3D scenes up to 90GB, sit in one shared pool instead of splitting across separate system RAM and VRAM.
AI Performance: Up to 1 Petaflop FP4
Alongside the core and memory counts, NVIDIA lists the top RTX Spark configuration at up to 1 petaflop of FP4 AI performance, the same performance language NVIDIA uses to describe its data-center AI accelerators. That figure is tied to the 20-core Grace CPU paired with a Blackwell GPU carrying 6,144 CUDA cores, a core count Wikipedia’s entry on the chip corroborated in April 2026, rather than the lower tier, and it is the number NVIDIA is leaning on to argue that RTX Spark machines can run local AI models that would otherwise require a discrete workstation GPU or a cloud connection.
The Lower-Tier Configuration
A second, confirmed configuration pairs an 18-core Grace CPU with 5,120 Blackwell GPU cores, stepping down from the top tier’s 20-core Grace CPU and 6,144 GPU cores. Reporting from Tom’s Hardware and VideoCardz indicates some lower-tier laptops cap memory at 32GB rather than offering the full 128GB option. That is a steep cut, not a minor trim, and it suggests NVIDIA and its OEM partners plan to segment RTX Spark the same way they segment RTX desktop GPUs: one halo configuration for enthusiasts and creators, one budget tier for everyone else. Confirmed OEM pricing fits that pattern: Microsoft’s Surface Laptop Ultra starts at $2,599.99 and HP’s OmniBook Ultra 16 starts at $3,199.99, both well below the $5,999.99 price of Microsoft’s top-tier, 128GB RTX Spark Dev Box, and further below ASUS’s $6,999.99 ProArt P16 configuration, which also carries 128GB of memory and a 2TB SSD, according to The Verge. NVIDIA and its partners still have not published exact memory configurations for every model in between.
From DGX Spark to RTX Spark: Tracing the Silicon Lineage
A Shared Silicon Family
NVIDIA CEO Jensen Huang said in 2025 that the processor underpinning DGX Spark, NVIDIA’s personal AI supercomputer built around the GB10 Superchip, shared its lineage with the N1 family now sold as RTX Spark. Tom’s Hardware reported that connection directly, and The Verge later described RTX Spark as a broader product family built on the same GB10-derived silicon rather than a rebrand of DGX Spark itself.
Why the Distinction Matters for Buyers
The distinction matters for buyers. DGX Spark is a complete NVIDIA-branded mini workstation aimed at AI developers, already compared against rival desktop AI boxes in our DGX Spark vs. Framework Desktop vs. Mac Studio breakdown. RTX Spark is the consumer and prosumer branch of that same silicon tree, sold through traditional PC makers instead of directly by NVIDIA.
Why NVIDIA Needed MediaTek
Splitting the Engineering Work
NVIDIA has never shipped a high-volume Arm CPU for Windows PCs before, and building the surrounding chip infrastructure, memory controllers, I/O, power management, from scratch would have taken years. MediaTek already builds Arm-based system-on-chips at massive scale for phones and has experience getting silicon through Microsoft’s Windows-on-Arm certification process.
The partnership splits the work along familiar lines. NVIDIA supplies the GPU and AI acceleration blocks, MediaTek handles the CPU architecture licensing and system integration. It follows similar logic to NVIDIA’s data-center push, where it increasingly licenses Arm CPU designs rather than building everything in-house, a pattern also visible in Arm’s own push into custom data-center silicon against AMD. Arm Holdings sits at the center of both stories: its instruction set licenses the CPU side of RTX Spark just as they license the server chips NVIDIA and its rivals are racing to ship.
The Arm Acquisition That Never Happened
That dependency on Arm licensing also explains why NVIDIA’s earlier attempt to buy Arm outright carried so much strategic weight. Owning Arm would have let NVIDIA set licensing terms for every company building Arm chips, not just its own. Regulators blocked that path, so partnerships like the MediaTek deal became the fallback route into markets NVIDIA could not reach by building everything itself.
Windows 11 on Arm Gets Its Biggest Push Yet
A Different Pitch Than Snapdragon
Microsoft has tried to make Windows on Arm work for nearly a decade, with limited success against Qualcomm’s Snapdragon-powered laptops. RTX Spark changes the pitch. Instead of asking buyers to accept weaker gaming performance for better battery life, NVIDIA is marketing RTX Spark machines as capable of real GeForce RTX gaming and local AI workloads on the same Arm chip.
Microsoft’s Copilot+ Angle
Microsoft’s own blog post framed the launch as a turning point for Windows hardware, tying it explicitly to Copilot+ PC branding and on-device AI features. For Microsoft, RTX Spark gives Windows-on-Arm a far stronger halo product than anything Qualcomm has managed to deliver on its own.
RTX Spark vs. DGX Spark: What Actually Changes
| Attribute | DGX Spark | RTX Spark |
|---|---|---|
| Product type | Complete NVIDIA mini workstation | Processor platform sold via OEM laptops/desktops |
| Core silicon | GB10 Blackwell Superchip | Same silicon family, consumer-tier variants |
| Target buyer | AI developers and researchers | Mainstream and prosumer Windows 11 users |
| Operating system | NVIDIA DGX OS (Linux-based) | Windows 11 |
| Sold by | NVIDIA directly | ASUS, Microsoft, HP, and other PC makers |
| Memory (top config) | Up to 128GB unified | 128GB unified LPDDR5X, confirmed via Microsoft’s $5,999.99 RTX Spark Dev Box |
The two products share a GPU lineage but serve different shelves. DGX Spark competes with other AI developer boxes, while RTX Spark competes for the same shoppers currently weighing a Qualcomm Snapdragon laptop or an Apple MacBook, a dynamic that already shapes how NVIDIA and Apple compare on market value.
Who Is Actually Building RTX Spark PCs?
Six Confirmed Launch Partners
ASUS, Dell, HP, Lenovo, Microsoft, and MSI are now confirmed RTX Spark launch partners, with preorders open across all six as of October 7, 2026, according to NVIDIA and Tom’s Hardware. ASUS lists an RTX Spark-equipped ProArt laptop on its official site, with a 128GB, 2TB ProArt P16 configuration priced at $6,999.99, according to The Verge. Microsoft confirmed its Surface Laptop Ultra at $2,599.99, reaching buyers October 16, 2026, and a higher-end RTX Spark Dev Box at $5,999.99, carrying the full 128GB memory configuration, 20-core Grace CPU, and 6,144-core GPU, arriving separately as a compact desktop in November 2026. HP confirmed its OmniBook Ultra 16 at $3,199.99 on that same October 16 timeline. Dell, Lenovo, and MSI joined the same October 7 preorder wave, though none of the three had published specific configuration pricing as of this writing.
Acer and Gigabyte Still Missing
Acer and Gigabyte remain the only previously rumored names without any confirmed RTX Spark presence, no preorder listing and no product page. With ASUS, Dell, HP, Lenovo, Microsoft, and MSI now all taking preorders, the partner roster has moved from speculation to confirmation for six of the eight companies once linked to the platform, but treat Acer and Gigabyte as likely rather than locked until each publishes its own product page.
Pre-Orders So Far
October 7 to October 16: The Preorder Window
NVIDIA and Tom’s Hardware confirm that RTX Spark preorders opened October 7, 2026, across ASUS, Dell, HP, Lenovo, Microsoft, and MSI, nine days ahead of the October 16, 2026 on-sale date when the first machines actually begin shipping to buyers. That nine-day gap between preorder and ship date is short by PC industry standards, which suggests NVIDIA and its partners had working retail units ready well before the public announcement rather than opening reservations on hardware still finishing validation.
Laptops Ship October 16, Compact Desktops Wait Until November
Not every RTX Spark form factor arrives on the same date. NVIDIA has confirmed October 16, 2026 as first availability for RTX Spark laptops, including Microsoft’s Surface Laptop Ultra, HP’s OmniBook Ultra 16, and ASUS’s ProArt P16. Compact RTX Spark desktops, the category that includes Microsoft’s RTX Spark Dev Box, are scheduled to follow in November 2026. Shoppers specifically waiting on the 128GB Dev Box rather than a laptop configuration should plan for a launch roughly a month behind the rest of the lineup.
Pricing Still Pending for Three Partners
Not every preorder page tells the same story. ASUS, HP, and Microsoft have published exact configuration pricing, covered in full in the pricing section below. Dell, Lenovo, and MSI confirmed they are shipping RTX Spark hardware by joining the October 7 preorder wave, but had not listed specific model names or prices as of this writing. Shoppers who want a firm number before October 16 currently have three confirmed options: Microsoft’s Surface Laptop Ultra, HP’s OmniBook Ultra 16, or ASUS’s ProArt P16.
Pricing: Confirmed RTX Spark Prices
NVIDIA and its partners confirmed real-world RTX Spark pricing on October 7, 2026, ending months of analyst speculation. Microsoft’s Surface Laptop Ultra starts at $2,599.99, HP’s OmniBook Ultra 16 starts at $3,199.99, Microsoft’s top-tier RTX Spark Dev Box, with its full 128GB memory configuration, 20-core Grace CPU, 6,144-core GPU, and 2TB SSD, lists at $5,999.99, and ASUS’s 128GB, 2TB ProArt P16 lists at $6,999.99. Unlike the laptops, the Dev Box is a compact desktop scheduled to arrive in November 2026, about a month after the October 16 laptop wave. Morgan Stanley analysts had earlier estimated the high-end, 128GB configuration at roughly $2,899, so confirmed pricing for that top tier landed about $3,100 above the analyst projection for Microsoft’s Dev Box, and roughly $4,100 above it for ASUS’s higher-specced ProArt P16.
What Each Configuration Costs
| Device | Maker | Starting Price | Availability |
|---|---|---|---|
| Surface Laptop Ultra | Microsoft | $2,599.99 | October 16, 2026 |
| OmniBook Ultra 16 | HP | $3,199.99 | October 16, 2026 |
| RTX Spark Dev Box (128GB) | Microsoft | $5,999.99 | November 2026 |
| ProArt P16 (128GB) | ASUS | $6,999.99 | October 16, 2026 |
| RTX Spark laptop | Dell | Preorder open; pricing not yet published | October 16, 2026 |
| RTX Spark laptop | Lenovo | Preorder open; pricing not yet published | October 16, 2026 |
| RTX Spark laptop | MSI | Preorder open; pricing not yet published | October 16, 2026 |
How It Compares to Analyst Estimates
For context, confirmed RTX Spark pricing, from $2,599.99 up to $6,999.99 for ASUS’s top ProArt P16 configuration, sits well above typical Windows-on-Arm laptops running Qualcomm chips, which commonly start closer to $999. It positions RTX Spark closer to premium creator and workstation laptops than to mainstream consumer notebooks.
The 100-FPS Marketing Claim, Explained
Where the Number Comes From
NVIDIA has promoted RTX Spark with claims of around 100 frames per second in modern games, a figure meant to signal that Arm-based Windows PCs no longer have to compromise on gaming performance. That number comes from NVIDIA’s own marketing material, and it almost certainly depends on specific games, resolutions, and the use of DLSS frame generation rather than native rendering performance alone.
Why Independent Testing Still Matters
No independent outlet has yet published standardized third-party benchmarks across multiple RTX Spark configurations. Readers comparing this to NVIDIA’s desktop GPU lineup, where real-world numbers from cards like the RTX 5090 are well documented, should wait for independent reviews before treating the 100-fps figure as a guaranteed baseline.
Market Impact: Apple, Qualcomm, and AMD Feel the Heat
RTX Spark drops into a Windows-on-Arm market that Qualcomm has largely had to itself since 2024. It also lands as Apple keeps pushing its own Arm-based Mac lineup and as AMD defends its Copilot+ PC share with Ryzen AI chips. The table below lines up publicly documented specs for the main rival chips against what has been confirmed for RTX Spark’s top configuration.
How RTX Spark Stacks Up on Paper
| Chip | CPU cores | Max clock | NPU performance | Source |
|---|---|---|---|---|
| NVIDIA RTX Spark (top config, confirmed) | 20 | Not officially disclosed | Not officially disclosed | Microsoft RTX Spark Dev Box specs |
| Qualcomm Snapdragon X2 Elite Extreme | 18 | Up to 5.0GHz | Up to 80 TOPS | Qualcomm product brief |
| AMD Ryzen AI 9 HX 475 | 12 (24 threads) | Up to 5.2GHz | Up to 60 TOPS | AMD CES 2026 announcement |
| Apple M5 (base) | Not disclosed | Not disclosed | 16-core Neural Engine | Apple official specs |
NVIDIA’s confirmed 20-core figure would beat every rival in raw CPU count, but core count alone does not decide real-world performance, and NVIDIA has not disclosed clock speeds or NPU throughput the way Qualcomm and AMD already have. Our broader look at who actually makes GPUs besides NVIDIA and AMD shows just how thin the competitive bench already is on the graphics side, which is exactly why NVIDIA entering the CPU side as well has rivals paying close attention.
Historical Context: NVIDIA’s Decade-Long Push Into CPUs
From Grace to RTX Spark
NVIDIA’s interest in CPUs is not new. The company tried and failed to acquire Arm Holdings outright in a deal that collapsed in 2022 under regulatory pressure. It later built Grace, a data-center Arm CPU for AI servers, and has continued expanding server-side CPU ambitions through projects like the Vera CPU platform offered through CoreWeave. RTX Spark’s own 18-core and 20-core CPUs carry the Grace name into consumer hardware for the first time, extending that same lineage onto a Windows laptop sitting on a store shelf next to a MacBook.
Why NVIDIA Chose a Partner This Time
That history explains why NVIDIA chose a partner instead of going it alone again. Grace and Vera targeted data centers, where NVIDIA already controls the sale relationship. Consumer Windows PCs require retail certification, driver support across thousands of software titles, and OEM relationships NVIDIA does not have built from scratch, exactly the gap MediaTek fills.
Manufacturing: Who Actually Fabs the RTX Spark Superchip?
An Unconfirmed Process Node
Neither NVIDIA nor MediaTek has published the exact foundry process node used for the RTX Spark Superchip in official specifications, and buyers should treat any specific node number circulating online as unconfirmed until one of the two companies states it directly. What is clear is that demand for advanced packaging and leading-edge capacity keeps climbing across the entire AI PC category, the same pressure chipmakers are weighing as they evaluate next-generation nodes like Intel’s 14A process, which Apple and NVIDIA are both reportedly evaluating for future designs.
A Capacity Crunch Across the AI PC Category
That broader capacity crunch matters for RTX Spark’s launch timeline. A chip built on a constrained leading-edge node competes for the same wafer allocation as NVIDIA’s data-center GPUs, which remain the company’s largest revenue driver by a wide margin. Any supply squeeze on the RTX Spark Superchip would likely show up first as limited configurations or regional availability gaps rather than an outright delay, since NVIDIA has strong incentive to prioritize whichever product line carries the better margin in a given quarter.
What This Means for Developers and Gamers
For Developers: Unified Memory at Scale
For developers, RTX Spark’s unified memory architecture is the bigger story than gaming frame rates. A 128GB shared pool of LPDDR5X memory, large enough to hold 3D scenes up to 90GB entirely in memory, backed by NVIDIA’s listed figure of up to 1 petaflop of FP4 AI performance on the top configuration, lets a single laptop run larger local AI models than most Windows machines can handle today, without needing a discrete GPU with its own separate VRAM. That is a meaningfully different pitch than past Windows-on-Arm laptops, which mostly competed on battery life rather than raw compute.
For Gamers: The Compatibility Question
For gamers, the bigger question is software compatibility. Windows-on-Arm has historically struggled with games that rely on anti-cheat systems or low-level graphics calls never tested on Arm silicon. NVIDIA’s GeForce RTX branding carries weight with PC gamers, but that reputation was built entirely on x86 systems, and RTX Spark will need real third-party game testing to prove the translation layer holds up under demanding titles.
Predictions: Where RTX Spark Goes From Here
- NVIDIA and OEM partners confirmed official RTX Spark pricing on October 7, 2026, the same day preorders opened across ASUS, Dell, HP, Lenovo, Microsoft, and MSI, ahead of the October 16, 2026 on-sale date, with Microsoft and ASUS pricing landing well above Morgan Stanley’s earlier $2,899 estimate for a 128GB configuration. Expect Dell, Lenovo, and MSI to publish their own specific model names and pricing in the days around that October 16 launch, closing the remaining gap with ASUS, Microsoft, and HP.
- Independent gaming and productivity benchmarks from outlets like Tom’s Hardware should arrive within the first month after the October 16, 2026 launch, testing whether the 100-fps marketing claim holds across more than a handful of optimized titles.
- Acer and Gigabyte remain the only previously linked OEMs without a confirmed RTX Spark product; expect them to join by early 2027, following the pattern of past Windows-on-Arm launches where the first wave of hardware arrives narrow and expands over several quarters.
- Qualcomm and AMD will face pressure to refresh their own Copilot+ silicon roadmaps faster, given NVIDIA’s confirmed CPU core-count advantage and its GeForce brand recognition among gamers.
- Watch whether Microsoft expands Copilot+ PC feature requirements specifically around RTX Spark’s unified memory capacity, which could become a new tier above today’s Copilot+ baseline.
FAQ: NVIDIA RTX Spark Questions, Answered
Is NVIDIA RTX Spark the same thing as DGX Spark?
No. They share related GB10-derived silicon, but DGX Spark is a complete NVIDIA-branded AI workstation running DGX OS, while RTX Spark is a processor platform sold inside Windows 11 laptops and desktops built by partners like ASUS.
What happened to the N1X name?
N1X was a leaked internal codename reported by Tom’s Hardware and VideoCardz before NVIDIA’s official announcement. NVIDIA now markets the chip publicly as the RTX Spark Superchip, and has not used the N1X label in official materials.
How much does an RTX Spark PC cost?
NVIDIA and its partners confirmed official pricing on October 7, 2026. Microsoft’s Surface Laptop Ultra starts at $2,599.99, HP’s OmniBook Ultra 16 starts at $3,199.99, Microsoft’s top-tier, 128GB RTX Spark Dev Box, a compact desktop shipping in November 2026, lists at $5,999.99, and ASUS’s 128GB ProArt P16 laptop lists at $6,999.99. Those figures replace Morgan Stanley’s earlier estimate of roughly $2,899 for a 128GB configuration.
Can I preorder an RTX Spark PC right now?
Yes. Preorders opened October 7, 2026 across ASUS, Dell, HP, Lenovo, Microsoft, and MSI, ahead of the October 16, 2026 on-sale date for RTX Spark laptops. ASUS, HP, and Microsoft have published exact model pricing; Dell, Lenovo, and MSI confirmed participation but had not published specific pricing as of this writing. Microsoft’s compact RTX Spark Dev Box desktop is scheduled separately for November 2026.
How much is the ASUS RTX Spark ProArt P16?
ASUS’s ProArt P16 configuration with 128GB of unified memory and a 2TB SSD lists at $6,999.99, according to The Verge, making it the most expensive confirmed RTX Spark configuration to date, above Microsoft’s $5,999.99 RTX Spark Dev Box, which carries the same 128GB memory and 2TB storage.
Which companies are making RTX Spark laptops?
Six partners have confirmed RTX Spark products with preorders open as of October 7, 2026: ASUS, Dell, HP, Lenovo, Microsoft, and MSI. ASUS, HP, and Microsoft have published exact pricing, ASUS’s ProArt P16, Microsoft’s Surface Laptop Ultra and RTX Spark Dev Box, and HP’s OmniBook Ultra 16, while Dell, Lenovo, and MSI have not yet published specific model pricing. Acer and Gigabyte are the remaining names from earlier reporting that have not confirmed any RTX Spark product.
Can RTX Spark PCs actually play modern games well?
NVIDIA claims around 100 fps in its own marketing, but independent third-party benchmarks have not yet been published. Game compatibility on Windows-on-Arm has historically been inconsistent, particularly for titles with strict anti-cheat software.
When do RTX Spark PCs ship?
RTX Spark laptops begin reaching buyers October 16, 2026, after NVIDIA and partners opened preorders and confirmed pricing on October 7, 2026. Compact RTX Spark desktops, including Microsoft’s RTX Spark Dev Box, follow separately in November 2026.
What chip is actually inside RTX Spark?
An Arm-based Grace CPU co-developed with MediaTek, paired with a Blackwell-generation GeForce RTX GPU. Microsoft’s RTX Spark Dev Box confirms the top configuration at a 20-core Grace CPU and a 6,144-core GPU, with up to 1 petaflop of FP4 AI performance, though NVIDIA has not published full specs for every OEM’s lower-tier model.
Does RTX Spark replace Qualcomm as the main Windows-on-Arm chip option?
Not immediately. Qualcomm’s Snapdragon X2 Elite already ships broadly across many OEMs and price points. RTX Spark enters as a premium alternative rather than a wholesale replacement, at least in its first generation.
