NVIDIA put RTX Spark laptops on sale worldwide on October 16, 2026, with a starting price of $2,599 and six launch partners: ASUS, Dell, HP, Lenovo, Microsoft and MSI. Ten days later, the spec that buyers argue about is not core count or clock speed. It is memory. The top RTX Spark N1X laptop tops out at 128 GB of unified memory, while AMD pitches its Ryzen AI Max+ PRO 495 platform at 192 GB.
That 64 GB gap decides which local AI models a thin laptop can hold in memory, and it drives most of the RTX Spark vs Ryzen AI Max+ PRO 495 debate. This analysis separates what NVIDIA and AMD have confirmed from what only third-party reports claim. It also covers the Arm versus x86 split, the real price ladder, and what each camp still has to prove. For the raw numbers on day one, start with the launch pricing for RTX Spark laptops.
- Confirmed: two N1X laptop tiers, up to 128 GB and up to 32 GB of LPDDR5X, a 45 to 80 W power range, and a $2,599 starting price.
- Confirmed: AMD lists the Ryzen AI Max+ PRO 495 as a 16-core, 32-thread Zen 5 part and positions it with up to 192 GB of memory.
- Not confirmed: an official AMD retail price, and a shipping date for every 495-based system.
- Open question: whether Windows on Arm software support catches up before buyers lose patience.
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 NVIDIA Shipped on October 16
NVIDIA unveiled RTX Spark on October 7, 2026 as a platform that pairs an Arm-based Grace CPU with a Blackwell GPU in a single unified-memory design. Pre-orders opened a day later with the same two laptop configurations and a planned October 16 availability date, according to PC Guide’s pre-order roundup. NVIDIA then confirmed global retail availability on schedule.
The pitch is simple. NVIDIA wants a Windows 11 laptop that can run large AI models locally, edit 8K 4:2:2 video, render big 3D scenes and play current games, all from one chip with one memory pool. The company’s own product page says 128 GB of unified memory handles 3D scenes up to 90 GB and models up to 120B parameters with a 1M-token context. It also lists launch-window game support that includes Cyberpunk 2077, Fortnite, Gears of War: E-Day, Marvel Rivals, World of Warcraft and NBA 2K27.
Retail status matters here because older coverage still describes RTX Spark as pre-order only. That framing expired on October 16. As of October 26, no source in this review reports a rename, a recall, a discontinuation or a price change since launch. The only fresh item in the past week is a GeForce driver update that reportedly added a separate Windows 11 package for RTX Spark.
The Two N1X Laptop Tiers in Plain Numbers
NVIDIA’s specification page lists two laptop configurations. The higher tier combines a 20-core Grace CPU with a 6,144-core Blackwell RTX GPU, up to 128 GB of LPDDR5X unified memory and a stated 1 petaflop of AI performance. The lower tier pairs an 18-core Grace CPU with a 5,120-core GPU, up to 32 GB of LPDDR5X and 827 teraflops. Both run in a 45 to 80 W range and both ship with Windows 11.
Some outlets list the lower tier with a 64 GB ceiling. NVIDIA’s own page says 32 GB, so this article uses the official figure. A manufacturer could still configure a specific laptop differently, which is why you should read the exact memory line on any product page before you pay. A separate look at 6,144 cores versus 5,120 cores breaks down the gap between the two silicon tiers.
NVIDIA’s Windows on Arm porting guide adds a detail the marketing pages skip. It reportedly describes the high-end N1X as 10 performance cores plus 10 efficiency cores, and the 5,120-core variant as nine plus nine. A third NVIDIA configuration, a 6,144-core desktop part, appears on the spec page too. A compact RTX Spark desktop is reportedly set for November 2026, and DGX Station for Windows is planned for the fourth quarter. Neither product was available on October 26.
AMD’s Counter: Ryzen AI Max+ PRO 495 and the 192 GB Pitch
AMD did not wait for NVIDIA’s launch to answer. Its product page for the Ryzen AI Max+ PRO 495 describes a 16-core, 32-thread Zen 5 processor in the Ryzen AI Max PRO 400 Series. According to AMD’s materials, the chip also carries integrated Radeon 8065S graphics with 40 graphics compute units and an XDNA 2 NPU rated up to 55 TOPS. The same materials list support for up to 192 GB of memory, clocks up to 5.2 GHz and a 45 to 120 W range.
On October 12, AMD introduced the next Ryzen AI Halo Developer Platform, built around the 495. Reports say up to 160 GB of the 192 GB pool can be assigned to the GPU as video memory. AMD also named HP’s ZBook Ultra G3a 16 as a commercial system pairing, covered in more detail in the piece on HP’s 192 GB ZBook G3a. The sources reviewed give no official price or shipping date for that laptop.
AMD also played the timing game. Tom’s Hardware reported on October 5 that AMD published Gorgon Halo benchmark material ahead of the expected RTX Spark launch and claimed it had shipped more than half a million agentic PCs. “Gorgon Halo” is a reported platform name and “Strix Halo” a reported code name, so treat neither as a retail product name. One report also calls the 495 the successor to the Ryzen AI Max+ 395.
RTX Spark vs Ryzen AI Max+ PRO 495: Spec Table
The table below uses only specifications that NVIDIA and AMD publish themselves. Where a figure depends on a third-party report, it stays out of the table and appears in the text instead.
| Spec | RTX Spark N1X (high tier) | RTX Spark N1X (lower tier) | Ryzen AI Max+ PRO 495 |
|---|---|---|---|
| CPU | 20-core Grace (Arm) | 18-core Grace (Arm) | 16-core, 32-thread Zen 5 (x86) |
| GPU | 6,144-core Blackwell RTX | 5,120-core Blackwell RTX | Integrated GPU |
| Maximum memory | 128 GB LPDDR5X | 32 GB LPDDR5X | 192 GB |
| Stated AI performance | 1 petaflop | 827 teraflops | Rated separately (NPU) |
| Power range | 45 to 80 W | 45 to 80 W | Set by each system maker |
| Instruction set | Arm, Windows 11 | Arm, Windows 11 | x86 |
| Official starting price | $2,599 (platform floor) | $2,599 (platform floor) | No official price published |
Read the AI row carefully. NVIDIA’s petaflop figure comes from FP4 math on the GPU, while AMD rates its NPU in TOPS and runs larger models through the GPU and shared memory. The two numbers describe different workloads, so you cannot divide one by the other and call it a winner.
Why Memory Capacity Became the Headline Spec
Discrete laptop GPUs cap out at a fixed pool of video memory, and the model has to fit inside it. Unified memory removes that wall. The CPU and GPU draw from one pool, so a laptop with 128 GB can load a model that would never fit on a card with a fraction of that. Both camps now sell that idea, and both lean on the same argument: run it locally, skip the cloud bill, keep your data on the machine.
The capacity ladder looks like this. The lower N1X tier stops at 32 GB. The high tier reaches 128 GB. AMD’s platform reaches 192 GB. That puts the AMD ceiling 64 GB above NVIDIA’s best laptop and 160 GB above NVIDIA’s entry tier. For a buyer who plans to run one mid-size model at a time, the 128 GB tier is plenty. For someone who wants a large model plus a long context window plus other apps open, the extra 64 GB starts to matter.
Capacity is only half the story. Speed depends on memory bandwidth, and neither company has published bandwidth numbers for its laptops in the sources reviewed. A single third-party report attributes a 256-bit LPDDR5X-8533 interface and roughly 273 GB/s of bandwidth to the AMD platform. NVIDIA has not given a comparable figure for N1X. Until independent reviewers measure tokens per second on the same model, a bigger pool tells you what fits, not how fast it runs.
What Fits in 128 GB and What Needs 192 GB
Models up to 120B on the 128 GB N1X
NVIDIA says the 128 GB configuration runs models up to 120B parameters with a 1M-token context. That is a vendor claim for a best-case setup, and it will not hold on the 32 GB tier. NVIDIA also points to its own Nemotron 3.5 Bolt 74B-A9B 24GB, a 74-billion-parameter model tuned for tool calling and multistep tasks, as a model that brings more capable local AI to supported RTX Spark configurations. The 24GB in the name hints at a footprint that fits even the smaller memory tier, though NVIDIA does not spell out which configurations qualify.
Where AMD’s extra 64 GB earns its keep
The 192 GB platform matters most for people who push past the 120B class or who keep a very long context loaded next to other work. If 160 GB of the pool can go to the GPU, as reports say, a developer can leave a big model resident and still run an editor, a browser and a container stack. That scenario is the real selling point behind AMD’s Ryzen AI Halo platform. It is also the scenario where a Windows on Arm laptop with a 128 GB ceiling runs out of room first.
Most buyers will not live at either ceiling. A typical working set of a mid-size model, a coding assistant and a few apps fits comfortably in 128 GB. The practical question is whether you ever need the next tier up, and how much you will pay for headroom you may use twice a month.
Arm vs x86: The Software Question
The CPU instruction set is the other fault line. RTX Spark laptops run Windows 11 on Arm with NVIDIA’s Grace CPU cores. The Ryzen AI Max+ PRO 495 runs standard x86 software, and one comparison in the sources says it supports Windows or Linux. That difference decides whether your tools run natively, run through emulation or do not run at all.
NVIDIA clearly knows this. It publishes a Windows on Arm guide that covers both porting and emulation, and it reportedly shipped a GeForce 621.01 driver on October 21 that added a separate Windows 11 package for RTX Spark. Developers who depend on x86-only drivers, anti-cheat software or niche enterprise tools should test their stack before ordering. Gamers face the same risk, which a report on doubts about Arm gaming on RTX Spark examines in detail.
AMD holds a clear edge in compatibility, and it will keep that edge until Arm app coverage closes the gap. NVIDIA holds an edge in the CUDA and RTX software ecosystem, which its spec page lists next to Windows 11 support. For AI developers, CUDA support on a laptop is a bigger draw than any single benchmark. For IT departments with decades of x86 software, AMD’s pitch is easier to approve.
The Pricing Ladder: From $2,599 Upward
NVIDIA’s official launch price is $2,599, and the company notes that regional and configuration differences apply. Microsoft lists the Surface Laptop Ultra from $2,599.99, with the entry configuration reportedly carrying 24 GB of memory and a 512 GB SSD. PC Guide’s pre-order list put the MSI Prestige N16 Flip AI+ at $3,299.
Prices climb quickly after that. Secondary coverage reports an HP OmniBook Ultra at $3,199.99, a Dell XPS 16 Creator Edition at $3,799.99 for a 32 GB and 512 GB build, and an ASUS ProArt P16 starting at $2,799.99. The 128 GB Surface configuration is reported at about $5,899, and one tested ProArt P16 configuration reportedly cost $6,999.99. Those are configuration prices, not a single MSRP. The Surface Laptop Ultra review roundup shows how the top tier performs for the money.
AMD’s side has less clarity. The sources reviewed contain no official retail price for the Ryzen AI Max+ PRO 495, and one third-party report puts a Ryzen AI Halo workstation at roughly $3,999. An Amazon listing dated October 25 exists for a 495-based product, but a retailer page does not establish AMD’s official price or availability. Before the launch, analyst and vendor estimates for RTX Spark ran from roughly $1,799 to $2,899. Those were forecasts, and the real floor landed above the middle of that range.
Launch Timeline: Oct. 7 to Oct. 16
The sequence shows how closely the two companies tracked each other. AMD moved first on benchmarks, NVIDIA moved first on retail, and both made their major announcements within the same two weeks.
| Date (2026) | Event | Status on Oct. 26 |
|---|---|---|
| Oct. 7 | NVIDIA unveils RTX Spark (N1X CPU plus Blackwell graphics) | Done |
| Oct. 8 | Pre-orders open with two laptop configurations | Done |
| Oct. 12 | AMD introduces Ryzen AI Halo Developer Platform with the Ryzen AI Max+ PRO 495 | Done |
| Oct. 16 | RTX Spark laptops available globally from six OEMs, from $2,599 | On sale |
| Q4 | DGX Station for Windows planned | Not available |
NVIDIA first promised the fall window on May 31, when it said N1X would debut across laptops from Microsoft, Dell, HP, ASUS, Lenovo and MSI. According to that report, the company planned more than 30 laptops and 10 desktops over time. The October 16 launch delivered the first wave of that promise, not the whole lineup.
Market Impact: OEMs, Chipmakers and Apple
Six OEMs shipping the same NVIDIA platform is a bigger shift than any single laptop. Dell, HP, Lenovo, ASUS and MSI built their premium lines on x86 for decades, and Microsoft has its own Surface push. Now each of them sells an Arm machine with an NVIDIA logo on it, which gives NVIDIA leverage it never had in the laptop CPU slot. The market reaction showed up in chip stocks, as covered in the report on how RTX Spark hit Intel and AMD shares.
AMD’s response also tells you something. Its best answer is not a faster core but a bigger memory pool and a broader software base. That is a defensible strategy, since x86 compatibility is a real moat. But it also concedes that NVIDIA owns the GPU and CUDA story on the laptop, which used to be AMD’s argument against Intel.
Apple sits in the background of this fight. Its own pages list the MacBook Air with M5 and the MacBook Pro with M5 as the current Macs, and Apple has sold unified memory for years. NVIDIA is effectively offering Windows users the same architecture idea with CUDA attached. AMD is offering it with x86 attached. Apple is the only one of the three that controls the entire software stack.
Historical Context: From Project DIGITS to Windows on Arm
RTX Spark did not appear from nowhere. A third-party report says NVIDIA’s Project DIGITS became DGX Spark, a small desktop built around the GB10 Grace Blackwell chip with 128 GB of unified memory and 4 TB of storage, rated at 1 petaflop of FP4 performance. That box proved there was an audience for compact local AI hardware. RTX Spark takes the same memory-first design and puts it in a laptop that runs Windows. For a side-by-side of that desktop class, see the DGX Spark, Framework Desktop and Mac Studio comparison.
AMD took a parallel route. Its large-memory Ryzen AI Max line started with the Ryzen AI Max+ 395, and the 495 extends that formula to 192 GB. The Ryzen AI Halo developer platform turns it into a mini workstation for the same crowd.
Windows on Arm carries the heaviest history. Microsoft has pushed Arm versions of Windows for years, and each wave stalled on app compatibility more than on hardware speed. RTX Spark is the first Windows on Arm launch where the GPU vendor itself ships the platform, writes the porting guide and funds the driver work. That does not guarantee success, but it changes the odds compared with earlier attempts.
Who Should Buy Which: Decision Table
| Buyer | Better fit | Why |
|---|---|---|
| Local LLM user running models up to 120B | RTX Spark N1X (128 GB tier) | NVIDIA states the 128 GB tier handles that class with a 1M-token context |
| Developer needing a model plus a heavy toolchain resident | Ryzen AI Max+ PRO 495 systems | The 192 GB pool leaves room for more concurrent work |
| 3D artist or video editor on Windows | RTX Spark N1X | RTX features such as OptiX and DLSS, plus 8K 4:2:2 editing support |
| Enterprise IT with legacy x86 software | Ryzen AI Max+ PRO 495 systems | Native x86 avoids emulation risk |
| Buyer on a tight budget | Lower N1X tier or wait | Entry price starts at $2,599, and AMD pricing is unpublished |
| Gamer who wants a thin laptop | RTX Spark N1X, after checking your titles | RTX graphics help, but Arm compatibility needs a per-game check |
The shortcut: pick NVIDIA if CUDA and RTX features drive your work, and pick AMD if x86 compatibility or a memory pool above 128 GB drives it. If neither applies, the cheaper N1X tier or a wait for reviews is the safer call.
Five Predictions for the Next 90 Days
- Independent speed tests will matter more than capacity. Once reviewers publish tokens-per-second results on identical models, the memory-bandwidth question will replace the capacity question in most buying guides.
- AMD will push 495 systems into more OEM catalogs. HP’s ZBook pairing is the first named design, and the company has every reason to announce more before holiday shopping starts.
- NVIDIA will answer with software, not specs. Expect more driver packages, more optimized models like Nemotron 3.5 Bolt and more porting help for developers, since app coverage is the weakest link.
- The compact desktop and DGX Station for Windows will reset the comparison. If the November desktop and the fourth-quarter workstation arrive on time, NVIDIA gets a higher-memory story than any laptop can offer.
- More mid-range builds will fill the gap. The $2,599 entry point looks set for now, but expect OEMs to add configurations between the entry tier and the reported $5,899 128 GB Surface build.
These are analyst judgments, not confirmed plans. For ongoing launches in this space, browse the hardware section.
FAQ: RTX Spark vs Ryzen AI Max+ PRO 495
When did RTX Spark laptops go on sale?
NVIDIA announced global availability on October 16, 2026, through ASUS, Dell, HP, Lenovo, Microsoft and MSI. Pre-orders had opened on October 8.
How much do RTX Spark laptops cost?
NVIDIA gives $2,599 as the starting price, with regional and configuration variation. Microsoft lists the Surface Laptop Ultra from $2,599.99. Reported prices for higher configurations climb to about $7,000.
How much memory does RTX Spark support?
The 20-core, 6,144-core tier supports up to 128 GB of LPDDR5X. NVIDIA lists the 18-core, 5,120-core tier with up to 32 GB.
How much memory does the Ryzen AI Max+ PRO 495 support?
AMD positions the platform with up to 192 GB of unified memory. Reports say up to 160 GB can be assigned to the GPU.
Does the Ryzen AI Max+ PRO 495 have an official price?
Not in the sources reviewed. AMD lists the processor and the Ryzen AI Halo platform, but a retail listing on Amazon does not set an official price or launch date.
Does RTX Spark run on Arm or x86?
Arm. The laptops use NVIDIA Grace CPU cores and run Windows 11 on Arm, so x86-only software needs emulation or a native port. The Ryzen AI Max+ PRO 495 uses x86.
Is a desktop version of RTX Spark available?
Not yet. NVIDIA lists a 6,144-core desktop configuration, and a compact desktop is reportedly planned for November 2026. DGX Station for Windows is planned for the fourth quarter.
Which is better for local AI, RTX Spark or Ryzen AI Max+ PRO 495?
It depends on the model size and your software. RTX Spark brings CUDA and RTX features with up to 128 GB, while the AMD platform offers a larger 192 GB pool and x86 compatibility. Independent speed tests are still thin.
