Nvidia RTX Spark Explained: What Windows on Arm Laptops Mean for Buyers in 2026

For about forty years, buying a Windows PC has meant choosing between two companies for the processor: Intel or AMD. Every Windows laptop and desktop most of us have ever owned ran on the same basic family of chips, known as x86. There were a few attempts to change that over the years, and most of them didn’t go very well. So when Nvidia, the company best known for graphics cards, walked on stage in Taipei at the end of May 2026 and said it had built a chip that would “reinvent the PC,” a lot of people were sceptical. We were too.

That chip is called RTX Spark. It’s an all-in-one processor that combines an Arm-based CPU, a Blackwell graphics chip from the same family as the RTX 50 series, and up to 128GB of shared memory, all in one package that goes into thin laptops and small desktops running Windows 11. The first laptops go on sale on October 16, 2026. Microsoft’s Surface Laptop Ultra starts at $2,599 in the US, and Dell’s XPS 16 Creator Edition starts at $3,799. ASUS, HP, Lenovo and MSI have machines coming too, with Acer and Gigabyte following later.

This is one of the biggest shake-ups the Windows world has seen in a long time. It’s also one of the most confusing, because it mixes several things that people tend to misunderstand at the best of times: what Arm actually is, why “unified memory” matters, what Windows on Arm can and can’t run, and what all the “AI PC” marketing really means in practice.

So in this guide we’re going to go through all of it, slowly and honestly. What RTX Spark is, how it’s built, where it came from, what software and games run on it, which laptops use it and what they cost, what early benchmarks suggest (and why you should be careful with them), and, most importantly for anyone reading this from Pakistan, who should actually consider one and who shouldn’t.

A quick disclaimer before we start. As of the time of writing, retail RTX Spark laptops haven’t reached buyers yet. No major independent review site has published full tested results on a shipping unit. So a lot of what’s out there is Nvidia’s own claims, leaked benchmarks, and early hands-on impressions. We’ll be clear about which is which. Where something is a manufacturer claim, we say so. Where it’s a leak, we say so. And when proper reviews arrive, some of the details here may change.

The short version, if you’re in a hurry

What it is: RTX Spark is Nvidia’s first processor for Windows laptops and small desktops. It puts a 20-core Arm CPU (designed with MediaTek), a Blackwell RTX GPU with up to 6,144 CUDA cores, and up to 128GB of unified LPDDR5X memory together in one package.
Why it matters: It’s the first time a company with serious graphics muscle has gone after the Windows laptop CPU market. Until now, Windows on Arm laptops (mostly Qualcomm Snapdragon) were good at battery life but weak at graphics and heavy creative work.
The AI angle: Nvidia rates it at up to one petaflop of AI compute (at FP4 precision), and says it can run AI language models with up to 120 billion parameters locally on the laptop, thanks to the big shared memory pool.
Games: Nvidia says hundreds of games will be playable at launch, including Cyberpunk 2077, Fortnite and Hogwarts Legacy, with anti-cheat support for titles like Counter-Strike 2 and Marvel Rivals. Some big ones, like Call of Duty, won’t arrive until 2027.
Price: Expensive. The first laptops start around $2,600 to $3,800 in the US, and high-memory configurations go much higher.
The catch: It runs Windows on Arm, which means some older software, drivers and games don’t run natively and rely on emulation, and a few things don’t run at all yet. Memory is soldered, so whatever you buy is what you’ll have forever.
Our view: Exciting technology, and very interesting for AI developers and creators who travel. For most buyers, and certainly for most desktop builders, it’s worth waiting for proper reviews and the second wave of cheaper models before jumping in.

What’s in this guide

What exactly is RTX Spark?
How we got here: Nvidia, Arm and Windows
Unified memory explained, and why AI people care so much
Windows on Arm in 2026: what runs and what doesn’t
Games on RTX Spark
The laptops and desktops: Surface Laptop Ultra, Dell XPS 16 and more
Performance: what we know, what we don’t
How it compares with Snapdragon, Apple and AMD
The “agentic” Windows pitch
Who should buy one, and who shouldn’t
What it means for desktop PC builders
Buying from Pakistan: things to think about
Myths and misunderstandings
Frequently asked questions

Part 1: What exactly is RTX Spark?

Let’s start by taking the chip apart, at least on paper. RTX Spark is what the industry calls a system-on-chip, or SoC, and Nvidia calls it a “superchip.” In a normal Windows laptop, you have a CPU from Intel or AMD, a separate graphics chip (often an Nvidia GeForce laptop GPU), memory for the CPU, and separate memory (VRAM) for the GPU. RTX Spark folds all of that into one package.

The CPU: 20 Arm cores, designed with MediaTek

The processor part is what Nvidia calls a Grace CPU, with up to 20 cores. It’s based on Arm’s designs rather than x86, and Nvidia developed it together with MediaTek, the Taiwanese company best known for smartphone chips. Reporting on the closely related GB10 chip (more on that in a moment) describes the 20 cores as 10 high-performance Cortex-X925 cores and 10 efficiency-focused Cortex-A725 cores.

If your used to Intel’s mix of performance cores and efficiency cores, the idea is similar. The big cores handle demanding, responsive work, and the small cores take care of background tasks while sipping power. Microsoft said it added workload scheduling tweaks to Windows specifically so the scheduler can spread work sensibly across all 20 cores, and enabled its power and thermal framework on the platform.

The GPU: Blackwell, the same generation as RTX 50

The graphics part is a Blackwell RTX GPU, the same architecture used in the GeForce RTX 50 series. In its full configuration it has 6,144 CUDA cores and fifth-generation Tensor Cores, which are the parts of the GPU that accelerate AI work. Those Tensor Cores support FP4, a very low-precision number format used to run AI models faster and in less memory. That’s where the headline “up to one petaflop of AI compute” figure comes from: it’s measured at FP4, which is worth remembering when people compare it with other chips quoted at different precisions.

For a sense of scale, 6,144 CUDA cores is a lot for something that lives inside a thin laptop alongside the CPU. Nvidia has said it supports the full stack of its graphics features: DLSS (including the new DLSS 5), Reflex, G-SYNC, ray tracing and RTX Remix.

NVLink-C2C: the glue in the middle

The CPU and GPU are connected by NVLink-C2C, a high-speed chip-to-chip link Nvidia already uses in its huge data centre “superchips.” It lets the CPU and GPU share data very quickly, which is a big part of what makes the unified memory design work well.

Memory: up to 128GB, shared by everything

RTX Spark uses LPDDR5X memory, the fast, power-efficient type used in high-end phones and thin laptops, in capacities up to 128GB. Crucially, it’s unified: the CPU and GPU draw from one shared pool instead of each having their own separate memory. We’ll explain why that matters so much in Part 3, but the short version is that it lets the GPU use far more memory than any laptop graphics card normally gets, which is a big deal for AI models and large 3D scenes.

Manufacturing

Nvidia’s chip is built on TSMC’s 3nm manufacturing process, according to coverage of the announcement, the same class of process used in Apple’s latest chips. A smaller, more advanced process generally means more transistors and better efficiency.

There’s more than one version

Here’s where it gets slightly messy. Nvidia’s marketing talks about “up to” 20 cores, “up to” 6,144 CUDA cores and “up to” 128GB, and those “up to”s are doing real work. The laptops now on sale make it clear there are at least two configurations of the main chip:

A full version with a 20-core CPU and 6,144 GPU cores.
A cut-down version with an 18-core CPU and 5,120 GPU cores.

Retail listings show the base Surface Laptop Ultra and the base Dell XPS 16 Creator Edition both using the 18-core, 5,120-core version, with the full chip in more expensive configurations. Leaked documents reported by VideoCardz also describe smaller “N1” variants with fewer cores and lower memory limits, aimed at cheaper, lower-power laptops, but Nvidia hasn’t officially announced those, so treat them as unconfirmed for now.

The takeaway: when you see a laptop advertised as “RTX Spark,” check exactly which configuration it has. The difference between the 18-core and 20-core versions, and between 24GB or 32GB and 128GB of memory, is significant, and so is the price gap.

Part 2: How we got here, Nvidia, Arm and Windows

RTX Spark didn’t come out of nowhere. To understand why it’s a big deal, and why some people are cautious, it helps to know a bit of the history.

x86 vs Arm in plain language

Every processor speaks a “language” called an instruction set. Intel and AMD chips speak x86. Almost every smartphone and tablet in the world, plus Apple’s Macs since 2020, speak Arm. Software compiled for one doesn’t run directly on the other, it has to be either rebuilt (recompiled) for the new language or translated on the fly by an emulator.

Arm chips earned a reputation for efficiency in phones, where battery life is everything. For a long time, though, they weren’t seen as powerful enough for serious desktop work. Apple changed that perception dramatically when it moved its Macs to its own Arm-based M-series chips in 2020, with excellent performance and battery life. Since then, the question in the Windows world has been: can someone do the same for Windows?

Windows on Arm’s rocky history

Microsoft has been trying to make Windows work on Arm for over a decade, and the early attempts were rough. Back in 2012, the original Surface RT ran a cut-down version of Windows on an Arm chip, which, interestingly, was an Nvidia Tegra. It couldn’t run normal Windows desktop apps at all and was widely considered a flop.

Later generations of Windows on Arm could run regular x86 apps through emulation, but performance was poor and compatibility was patchy. For years the main supplier of Arm chips for Windows laptops was Qualcomm. A 2023 Reuters report said Microsoft’s exclusivity arrangement with Qualcomm for Windows on Arm chips was set to expire in 2024, and that Nvidia was developing its own Arm CPUs for Windows. That was the first public hint of what eventually became RTX Spark.

Things genuinely improved with Qualcomm’s Snapdragon X Elite chips in 2024, which made Windows on Arm laptops properly competitive for everyday use, with excellent battery life. Microsoft also launched a much better emulator called Prism. But even then, Snapdragon laptops had a clear weak spot: graphics. They were fine for office work, browsing and video, and okay for light creative work, but not something you’d buy for 3D work or serious games.

Nvidia’s Arm experience

Nvidia isn’t new to Arm at all. It made the Tegra chips found in early Android tablets, the Shield TV and, most famously, the original Nintendo Switch. In the data centre, its Grace CPU, also Arm-based, is paired with its AI GPUs in giant “superchips” like Grace Hopper and Grace Blackwell.

At CES in January 2025, Nvidia announced Project DIGITS, a small desktop “AI supercomputer” built around a chip called GB10 Grace Blackwell, which combined an Arm CPU co-designed with MediaTek and a Blackwell GPU with 128GB of unified memory. Project DIGITS was later renamed DGX Spark and went on sale in late 2025, running Nvidia’s own Linux-based DGX OS. It was aimed squarely at AI developers who wanted to run large models on their desk.

Around the same time, reports started appearing about chips codenamed N1 and N1X for Windows laptops. In September 2025, Tom’s Hardware reported that Jensen Huang said the N1 was the Arm product used in DGX Spark and related products.

RTX Spark is essentially DGX Spark for Windows

When RTX Spark was announced, it became clear that it’s very closely related to the GB10 chip in DGX Spark. The Register described the N1X and GB10 as essentially the same chip, and ServeTheHome questioned whether N1X is really a distinct chip at all or simply GB10 rebranded for consumers. The main difference is the software: DGX Spark ships with Linux for AI developers, and RTX Spark ships with Windows 11 for everyone else.

That’s actually useful for buyers, because it means DGX Spark gives us a preview of how the silicon behaves in real AI workloads, which we’ll use later when talking about performance.

The announcement

Nvidia and Microsoft announced RTX Spark on May 31, 2026, at Nvidia’s GTC Taipei event during Computex. Jensen Huang’s pitch was bold, as usual: “The PC is being reinvented,” he said, and “This is the new PC. The personal AI computer.” Microsoft’s Satya Nadella said the goal was “to deliver unmetered intelligence to every home and every desk with Windows.” According to Windows Latest’s coverage, Huang also claimed the two companies had worked on the platform for three years, and that the machine “literally runs everything the world has ever created.” That last line is the kind of claim worth taking with a healthy dose of salt, as we’ll see in the compatibility section.

Microsoft followed up with more details at its Build developer conference in early June, and then on October 7, 2026, at an event in San Francisco, it revealed prices for the Surface Laptop Ultra and a developer desktop called the Surface RTX Spark Dev Box.

Part 3: Unified memory explained, and why AI people care so much

If you only remember one technical thing about RTX Spark, make it this one. The unified memory design is what makes it genuinely different from a normal performance laptop with an Nvidia graphics card, and it’s the main reason AI developers are excited.

How a normal laptop handles memory

In a typical Windows laptop with a GeForce GPU, there are two separate pools of memory:

System RAM (say 16GB or 32GB of DDR5), used by the CPU and Windows.
VRAM on the graphics chip (often 8GB, 12GB, or at the very top end 24GB of GDDR7), used by the GPU.

The GPU is very fast at working with its own VRAM, but it can only comfortably use what’s there. If a task needs more memory than the VRAM holds, data has to be shuffled back and forth over a much slower connection, and performance falls off a cliff. For games that’s usually fine, because games are designed around typical VRAM sizes. For AI models, it’s a hard wall.

Why AI models are memory-hungry

A large language model, the kind of AI that powers chatbots, is essentially a giant collection of numbers called parameters. A model’s memory needs depend on how many parameters it has and how precisely each number is stored. Roughly speaking, a model with 70 billion parameters stored at 4-bit precision needs somewhere around 35 to 40GB of memory just for the model itself, plus extra for the conversation context. A 120-billion-parameter model needs more still.

No laptop graphics card comes anywhere close to that much VRAM. Even the biggest desktop graphics cards, which are huge and expensive, top out at 32GB on the consumer side. So until now, running really large AI models locally meant expensive multi-GPU workstations, servers, or a Mac with lots of unified memory.

What unified memory changes

With unified memory, the GPU can use most of the shared pool. On a 128GB RTX Spark machine, a very large chunk of that is available to the GPU, which means models that would never fit on a laptop graphics card can be loaded and run locally. That’s why Nvidia says RTX Spark can run 120-billion-parameter models with up to a million tokens of context, render 3D scenes of 90GB or more, and edit 12K video.

Apple has been using this same approach in its M-series chips for years, and it’s one of the main reasons MacBook Pros with lots of memory became popular with AI developers. RTX Spark is the first time Windows laptops get a comparable design with a powerful GPU attached, plus Nvidia’s CUDA software, which is the standard for most AI development.

The trade-off: bandwidth

Unified memory isn’t magic though. LPDDR5X is much slower than the GDDR7 on a high-end desktop graphics card. Nvidia’s DGX Spark, which uses the closely related GB10 chip, is rated at 273GB/s of memory bandwidth. Some coverage puts RTX Spark at around 300GB/s. For comparison, a desktop RTX 5090 has well over 1,500GB/s.

Why does that matter? When an AI model generates text, it has to read through its parameters for every word (technically every token) it produces. So the speed of generating text is heavily limited by memory bandwidth. Reviews of DGX Spark found exactly that: it can hold very big models that desktop cards can’t, but it generates text from them more slowly than a big desktop GPU would if the model fit. LMSYS, which tested DGX Spark, concluded that the shared LPDDR5X bandwidth is “the key bottleneck in AI inference performance.” In one of their tests, a 70-billion-parameter model at FP8 decoded at around 2.7 tokens per second, which is usable for experiments but slow for conversation. Smaller models run much faster.

So the honest summary is: RTX Spark is about capacity, not raw speed. It lets you run things that otherwise wouldn’t run at all on a laptop. If your models fit comfortably in a big desktop graphics card, that card will usually be faster.

The other trade-off: you can’t upgrade it

Unified LPDDR5X memory is soldered to the package. You choose your memory when you buy the laptop, and that’s it for the life of the machine. Given what we covered in our previous article about the 2026 memory crisis, that’s a significant decision: high-memory configurations cost a lot more, and you can’t add more later. If you think you’ll want to run larger AI models, buy more memory upfront, because you’ll never get the chance again.

The good news, at least on the Surface Laptop Ultra, is that storage is not soldered. Microsoft has confirmed the SSD is user upgradable, with options from 512GB to 2TB from the factory.

Part 4: Windows on Arm in 2026, what runs and what doesn’t

This is the section that matters most if you’re thinking of actually buying one. Hardware is only as good as the software it runs, and the history of Windows on Arm is full of people who bought a laptop and then found out their favourite app, printer or game didn’t work.

The good news is that things are much better in 2026 than they were even two or three years ago. The less good news is that “much better” isn’t the same as “everything works.”

Three ways software runs on Windows on Arm

When you run a program on an RTX Spark laptop, one of three things happens:

It runs natively. The developer has built an Arm64 version of the program. This is the best case, it runs at full speed with full efficiency.
It runs through emulation. The program only exists in an x86 version, and Windows’ built-in emulator, called Prism, translates it to Arm instructions on the fly. Most of the time this works, with some performance cost.
It doesn’t run. Usually because it depends on something emulation can’t handle, most often a low-level driver.

Native apps: the big names are there

A lot of the most important software has been rebuilt for Arm. Browsers like Chrome and Edge run natively, as do Microsoft Office and most of Microsoft’s own apps. For creative work, Nvidia and Microsoft list native Arm versions of Adobe Photoshop and Premiere, Blender, DaVinci Resolve, Maxon Cinema 4D and Redshift, Topaz Photo, CapCut, Cubase, Bitwig Studio and Affinity. Nvidia says it worked with over 100 Windows software providers ahead of launch.

Windows Central made a fair point in its coverage of Arm laptops this year: it’s getting tougher to find everyday apps that need emulation at all, unless you get into specialised stuff.

Prism emulation: better than it used to be

For everything else there’s Prism. Microsoft introduced it with the Snapdragon X laptops and has improved it steadily. One important upgrade came in 2025, when Prism gained support for x86 extensions including AVX and AVX2, which many creative apps and games need. Windows Latest gave a nice example: the music software Ableton Live doesn’t have a native Arm version and used to fail under emulation because it depends on those extensions, but it works after the update. Microsoft says it worked with Nvidia to optimise Prism specifically for RTX Spark as well.

MATLAB is a good example of a professional app that, according to Nvidia’s own list, runs on RTX Spark through Prism rather than natively.

Emulation does cost performance, and how much depends heavily on the app. Light apps feel basically native. Heavy, CPU-bound apps can be noticeably slower than they’d be natively. We’ll know much more about RTX Spark’s emulation performance once reviewers get retail units.

What still causes problems

Here’s where you need to be careful. Things that commonly don’t work, or need checking, on Windows on Arm include:

Drivers for older hardware. Drivers can’t be emulated. If you have an older printer, scanner, audio interface, specialised USB device or graphics tablet, it needs an Arm64 driver. Most current mainstream devices are fine, but niche and older equipment is the biggest risk.
Some VPNs, security tools and system utilities. Anything that installs a low-level driver, including certain VPN clients, antivirus products, and hardware monitoring tools, needs an Arm version.
Android development. Windows Latest pointed out that Android Studio remains a problem area because it needs to emulate Android.
Specialised professional software. Some engineering, CAD and industry-specific tools haven’t been ported. AutoCAD and Premiere Pro were mentioned by Windows Central as the kind of “specialised stuff” where you may still hit emulation.
Anti-cheat in online games. More on this below.

LTT Labs summed it up well in July 2026, noting that a fair share of the compatibility issues the community has noticed over the years still persist today. Our advice is simple: before buying any Windows on Arm laptop, make a list of every app, game and device you rely on, and check each one. Don’t assume.

Part 5: Games on RTX Spark

Games have always been Windows on Arm’s weakest point. Nvidia clearly knows that because its put a lot of effort into making games part of the RTX Spark launch story.

What Nvidia says will work at launch

According to Nvidia, RTX Spark launches with “hundreds” of top games playable on day one, October 16, 2026. Microsoft also says hundreds of PC games, including big releases and indie titles, will be supported. Titles Nvidia has named include Cyberpunk 2077, Fortnite, Gears of War: E-Day, Hogwarts Legacy, Marvel Rivals and NBA 2K27.

Nvidia’s headline performance claim is that RTX Spark can run big games at 1440p at over 100 frames per second with ray tracing, DLSS and Reflex. That’s a manufacturer claim made with DLSS upscaling and frame generation switched on, so it’s not the same as “100fps at native 1440p.” It’s also not independently tested yet. Other reports suggest graphics performance close to a GeForce RTX 5070 Laptop GPU, depending on the game. Again, wait for reviews before you treat any of that as fact.

DLSS 5 on day one

RTX Spark supports Nvidia’s DLSS suite, including DLSS 5, the new neural rendering technology Nvidia announced in March 2026 (we’ll cover it in detail in a future article). Nvidia says DLSS 5 arrives on launch day for NBA 2K27, with Hogwarts Legacy getting its DLSS 5 update on November 5 and Where Winds Meet adding it in late November.

Anti-cheat: the big hurdle, partly solved

Online competitive games use anti-cheat systems that run at a very deep level of Windows, as kernel drivers. Like other drivers, they can’t simply be emulated, they need native Arm versions. For years this was the single biggest reason popular online games didn’t work on Arm laptops.

Nvidia says it worked with anti-cheat developers so that major systems work on RTX Spark. It names BattlEye, Easy Anti-Cheat, Vanguard (Riot’s system), Tencent Anti-Cheat Expert, EA Javelin and Activision’s Ricochet as systems with support either now or coming. Games Nvidia lists as playable online with anti-cheat on October 16 include ARC Raiders, Counter-Strike 2, Fortnite, Marvel Rivals, Rocket League, THE FINALS, War Thunder, and World of Tanks and World of Warships.

But, and this is a big but, several major titles aren’t ready yet:

Call of Duty is slated to come to Windows on RTX Spark in 2027, with Ricochet anti-cheat support being developed for Arm64.
EA Javelin anti-cheat (used in EA’s games) is listed for early 2027.
PUBG: Battlegrounds is targeted for the first half of 2027.
Arena Breakout: Infinite is coming in 2027.
Delta Force is only getting “limited testing” in a future season, with no date.

Some other titles are getting updates soon after launch: Marvel Rivals gets further optimisations with Season 11 on November 13, and World of Warcraft supports RTX Spark from its November 4 update according to Nvidia.

If online multiplayer is a big part of why you’d buy a laptop, check your specific games against these lists before buying. “Hundreds of games” is a good start, but if the one you play every night isn’t on the list, it doesn’t help you much.

Older games and mods

Single-player games without anti-cheat will generally run through Prism if they don’t have native Arm builds. Most PC games today are x86 only, so emulation is how the majority of your library will run. Performance in those depends on how heavy the game is on the CPU, since emulation adds CPU overhead. GPU-heavy games may be less affected than CPU-heavy ones like big strategy titles or simulators. Mods that inject code into games, overlays and some launchers can also be hit or miss.

The Xbox PC app has also been updated over time to offer Arm64-compatible versions of games where they exist, and Nvidia lists Xbox among its RTX Spark partners.

Our honest take on RTX Spark for games

RTX Spark is by far the most capable Windows on Arm machine for games so far, and the launch support is much stronger than anything Arm laptops have had before. But if your main goal is playing games, especially online competitive ones, a traditional x86 laptop with a GeForce RTX 50 series laptop GPU is still the safer, cheaper and more compatible choice in 2026. And a desktop is safer still. RTX Spark makes sense when games are one part of what you do on a thin, portable machine, not the main reason you’re buying it.

Part 6: The laptops and desktops

Nvidia doesn’t sell RTX Spark laptops itself, it sells the chip to PC makers. The launch partners are ASUS, Dell, HP, Lenovo, Microsoft Surface and MSI, with Acer and Gigabyte models to follow. Nvidia says designs will range from 14 to 16 inches, as thin as 14mm and as light as three pounds (about 1.4kg), many with tandem OLED displays and G-SYNC. Here are the machines with confirmed details so far.

Microsoft Surface Laptop Ultra

Microsoft’s flagship launch device is a 15-inch laptop, and it’s clearly aimed at the MacBook Pro crowd. Microsoft is even offering up to $1,000 off for people who trade in a MacBook Pro.

Display: a 15-inch “PixelSense Ultra” Mini-LED touchscreen, 120Hz, with up to 2,000 nits HDR peak brightness and 1,000 nits SDR. Resolution is listed at around 3270 x 2180.
Chip: 18-core RTX Spark (5,120 GPU cores) in the base model, with the full 20-core, 6,144-core chip in higher configurations.
Memory: the base model reportedly starts at 24GB of unified memory, with options up to 128GB.
Storage: 512GB to 2TB, and importantly, not soldered. Microsoft has confirmed it’s user upgradable.
Ports: two USB-C/USB4 ports, one USB-C/USB4 with a magnetic connector, HDMI 2.1b, USB-A, SD card reader, and a headphone jack. That’s a refreshingly generous selection for a premium thin laptop.
Size and weight: under 18mm thick and 2.0kg (4.41 lbs) according to Microsoft.
Battery: ServeTheHome lists a 92Wh battery, the maximum size generally allowed on flights is 100Wh, so that’s near the top.
Price: starts at $2,599 in the US. A second base model with the more powerful chip starts at $3,700, and maxed-out configurations go up to $5,900. TechCrunch reported Microsoft said the highest-end model was already out of stock at announcement.
Availability: sales begin October 16, 2026.

Note that at 2.0kg and nearly 18mm the Surface Laptop Ultra is noticeably thicker and heavier than Nvidia’s “as thin as 14mm and as light as three pounds” headline. That figure describes the thinnest designs the platform allows, not every laptop. A 15-inch machine with a big battery and cooling for an 80W-class chip is going to weigh more.

Dell XPS 16 Creator Edition

Dell’s entry is aimed, as the name says, at creators.

Chip: 18-core RTX Spark with 5,120 GPU cores in the base model.
Memory: 32GB unified memory in the base model, with 64GB and 128GB options (pricing for those wasn’t published at the time of writing).
Storage: 512GB SSD in the base model.
Display: 16-inch 3.2K OLED touchscreen at 120Hz.
Battery: 99Wh.
Size and weight: 17.8mm thick and about 2.02kg.
Price: $3,799.99 for the base configuration at Best Buy, with preorders open and an October 16 release listed.

It’s worth pointing out that Dell’s base model costs about $1,200 more than Microsoft’s base model, but comes with 32GB instead of 24GB of memory and an OLED panel. As always with premium laptops, compare like for like.

Surface RTX Spark Dev Box

Microsoft also announced a compact desktop for developers, the Surface RTX Spark Dev Box, starting at $6,000. It comes with 128GB of unified memory, runs the chip in a 100-watt thermal envelope (more headroom than a laptop), and ships with developer tools including VS Code, GitHub Copilot CLI, WSL and PowerShell 7 preinstalled. This is not a home PC, it’s a tool for people building AI software.

Others coming

HP has an upcoming OmniBook model, and ASUS, Lenovo and MSI have designs on the way. Nvidia also mentioned a DGX Station for Windows, a bigger deskside system for businesses running AI agents. Expect a range of configurations, and hopefully some cheaper ones, to arrive over the coming months.

Part 7: Performance, what we know and what we don’t

We want to be really careful here, because this is where a lot of online coverage gets ahead of itself. As of writing, no established review outlet has published controlled, repeatable tests on retail RTX Spark laptops. What exists falls into three buckets: Nvidia’s claims, leaked benchmarks, and one unusual unofficial review.

Nvidia’s claims

Up to 1 petaflop of FP4 AI compute.
Run AI models with up to 120 billion parameters locally, with up to 1 million tokens of context.
Play big games at 1440p at over 100fps, with ray tracing, DLSS and Reflex enabled.
Edit 12K 4:2:2 video and render 3D scenes of 90GB or more.
Up to 2x faster AI and graphics performance in Adobe apps (an Adobe partnership claim).

These are useful as a description of what the chip is designed for. They’re not independent measurements.

Leaked CPU benchmarks

Several RTX Spark results have appeared in benchmark databases before launch:

Geekbench: a leaked 20-core result scored roughly 2,570 single-core and 23,126 multi-core. According to coverage of that leak, that beats AMD’s Ryzen AI Max+ 395 in multi-core, but trails Qualcomm’s Snapdragon X2 Elite Extreme and Apple’s M5 Max by a wide margin. Tom’s Hardware also reported two variants appearing on Geekbench, a full 20-core and a cut-down 18-core, with the full chip beating most x86 mobile chips. The listings were later removed.
Cinebench: results from a pre-release unit showed 1,386 points multi-core and 123 single-core in Cinebench 2024, and 5,771 multi-core and 540 single-core in Cinebench 2026, at an 80W sustained power limit.

Leaked benchmarks on pre-release hardware often don’t reflect final performance, because drivers, firmware and power tuning change before launch. Treat them as a rough idea only.

The “found on a roadside” review

In one of the stranger tech stories of the year, TechPowerUp and TweakTown reported on a tech enthusiast who said they found a pre-release Surface Laptop Ultra, tested it for about a month, and posted results. In AI tests, performance landed roughly around a desktop GeForce RTX 4070. In 3DMark graphics tests, it was closer to a desktop RTX 3060 Ti. The tester also reported a predictable lock-up after about an hour of use, which could easily be down to pre-release firmware.

We mention this because it’s been widely shared, but please don’t treat it as a proper review. It’s a single pre-production unit, tested by an unknown person, under unknown conditions. It’s interesting, not conclusive.

What DGX Spark tells us about AI performance

Since DGX Spark uses essentially the same silicon, its reviews give a better grounded picture of AI performance. They consistently found that the system’s strength is memory capacity, not speed. It can load models no single consumer graphics card can hold, but text generation speed is limited by memory bandwidth. LMSYS measured around 49.7 tokens per second generating text with the GPT-OSS 20B model, compared with about 215 on a far more expensive RTX Pro 6000 Blackwell desktop card. StorageReview measured about 34 tokens per second with Llama 3.1 8B at FP4. Very large models run much slower, but they run, which is the point.

RTX Spark laptops will likely land in a similar range for AI, possibly a little lower in thin laptops with tighter power limits. That’s very usable for coding assistants, document analysis, private chatbots and experimenting with bigger models. It’s not going to replace a powerful desktop GPU for heavy AI work where the model fits in VRAM.

What we still don’t know

Real battery life under normal use and under load.
Sustained performance, fan noise and heat in thin chassis.
Actual frame rates in tested games, at native resolution and with DLSS.
Emulation performance for heavy x86 apps.
Long-term driver stability.

These are exactly the things that decide whether a laptop is pleasant to live with, and we won’t know them until proper reviews land after October 16.

Part 8: How it compares with Snapdragon, Apple and AMD

RTX Spark isn’t entering an empty market. Here’s how it stacks up, at least on paper, against the main alternatives. Remember that full reviews are still to come, so some of this will shift.

Qualcomm Snapdragon X2 Elite

Qualcomm is the incumbent in Windows on Arm, and the company that did a lot of the hard work making the platform viable: pushing developers to port apps, working on anti-cheat support and helping Microsoft improve Prism. Its newer Snapdragon X2 Elite chips use custom Oryon CPU cores, and early leaks suggest they’re stronger than RTX Spark in CPU benchmarks, at least the X2 Elite Extreme.

Where RTX Spark should clearly win is graphics and AI on the GPU. Preview estimates put RTX Spark’s graphics well ahead of Snapdragon X2 Elite. Snapdragon laptops are also generally thinner, lighter, cheaper, and likely to have better battery life, because they’re not carrying a large GPU. Think of it this way: Snapdragon X2 is the efficient everyday Arm laptop, RTX Spark is the powerful creative and AI-focused Arm laptop.

Apple MacBook Pro with M5 Pro and M5 Max

Apple is the obvious comparison, because Apple proved the unified memory approach works brilliantly and has been refining it for years. Leaked CPU numbers suggest Apple’s M5 Max is well ahead of RTX Spark on the CPU side. Notebookcheck’s estimates place RTX Spark’s graphics somewhere between the M5 Pro and M5 Max in synthetic tests and creative applications.

RTX Spark’s advantages over a Mac are mainly about software: it runs Windows, it supports Nvidia’s CUDA (which a huge amount of AI software is built on), and it has much broader support for PC games, even with the Windows on Arm limitations. If your work depends on CUDA-specific tools, that alone can decide it. The Mac’s advantages are maturity, a native software ecosystem built entirely for Arm, and Apple’s long track record with battery life and build quality.

AMD Ryzen AI Max (Strix Halo)

AMD’s Ryzen AI Max series, codenamed Strix Halo, is probably RTX Spark’s most direct Windows rival. It’s also a big chip with a powerful integrated GPU and a large unified memory pool, up to 128GB, aimed at exactly the same “run big AI models on a laptop or mini PC” crowd.

The key difference: Strix Halo is x86. Every Windows app, game and driver runs natively, with no emulation and no anti-cheat question marks. That’s a huge practical advantage. RTX Spark’s advantages are Nvidia’s GPU architecture and CUDA software, which matter a lot for AI developers. Preview estimates suggest RTX Spark’s graphics are well ahead of Strix Halo, and the leaked Geekbench result put RTX Spark ahead of the Ryzen AI Max+ 395 in multi-core CPU, though that’s one leak.

A normal x86 laptop with a GeForce RTX 50 GPU

Don’t forget the boring option. A good Intel or AMD laptop with a GeForce RTX 5070 or 5080 laptop GPU will run every Windows game and app natively, often for less money than an RTX Spark machine. For pure frame rates in games, especially competitive ones, it’s still the safer choice. The downsides are battery life, weight and VRAM: even the biggest laptop GPUs have nowhere near enough VRAM for large AI models, which is exactly the gap RTX Spark targets.

A quick comparison table

Option	Strengths	Weaknesses
RTX Spark	Strong GPU for its size, CUDA, up to 128GB unified memory, big game support push	Windows on Arm compatibility gaps, expensive, untested at retail, memory not upgradable
Snapdragon X2 Elite	Efficiency, battery life, strong CPU, cheaper	Much weaker graphics, same Arm compatibility limits
Apple M5 Pro / Max	Mature Arm ecosystem, strong CPU, excellent efficiency	macOS, no CUDA, limited PC game library
AMD Ryzen AI Max	x86 compatibility, big unified memory, strong iGPU	No CUDA, GPU likely behind RTX Spark
x86 + RTX 50 laptop GPU	Full compatibility, best for games, wide price range	Limited VRAM for AI, heavier, worse battery life

Part 9: The “agentic” Windows pitch

You’ll hear the word “agentic” a lot in RTX Spark marketing. Nvidia’s announcement headline literally talked about turning Windows into an “agentic AI OS.” What does that actually mean?

What an AI agent is

A chatbot answers questions. An AI agent goes further and actually does things: opens apps, edits files, fills in forms, searches your documents, chains several steps together to finish a task. For example, rather than asking an AI how to rename a hundred photos by date, you’d ask an agent to do it, and it would.

That’s powerful, and also a bit scary, because an agent that can act on your computer can also make mistakes on your computer. Microsoft’s own developer documentation openly warns that agents can hallucinate, can be manipulated through what’s called cross-prompt injection (where malicious instructions are hidden in content the agent reads), and should run with separate accounts and limited access.

What Microsoft and Nvidia are building

Several pieces were announced across the spring and autumn:

Execution Containers. At its October 7 event, Microsoft said a revamped Windows 11 on these PCs includes “Execution Containers,” which make it easier to sandbox AI agents, keeping them in a controlled box rather than giving them free rein over your system. Satya Nadella said the feature will come to all Windows 11 users, not just RTX Spark machines.
Agent identity and controls. Microsoft’s Build 2026 documentation talked about giving agents distinct identities, and enterprise controls through tools like Intune, so businesses can manage what agents are allowed to do.
Nvidia OpenShell. Nvidia described a runtime that lets users define agent permissions, route requests to local or cloud AI models based on privacy rules, and mask personal information before anything is sent to a cloud model.
Taskbar access. Windows agent features will be reachable from the taskbar.
Third-party agents. Nvidia named agent developers including Nous Research (Hermes Agent) and the OpenClaw Foundation as partners. Nadella said Microsoft is enabling this “not just for our apps but apps for anyone’s agent.”

Why local hardware matters for agents

Here’s the connection to RTX Spark. Agents can run on cloud AI models, but that means sending your files, screen contents and requests to someone else’s servers, and often paying per use. Running the AI model locally keeps your data on your machine and costs nothing per request. Microsoft pitched its RTX Spark devices as being able to run AI models locally “at no cost per use.” With 128GB of unified memory, a laptop can run fairly capable models entirely offline.

RTX Spark devices also join Microsoft’s Copilot+ PC category, and include an NPU (neural processing unit) for low-power AI tasks alongside the big GPU.

Should you care today?

Honestly, for most people, not yet. Agentic AI on Windows is at a very early stage, and the most useful applications are still being built. If you’re a developer working on agents or AI apps, RTX Spark is genuinely interesting as a platform to build on. If you’re an everyday user, treat agent features as a bonus that may become useful over the next couple of years, not as a reason to spend $2,600 today. We’ll be covering AI agents in much more depth in a later article in this series.

Part 10: Who should buy one, and who shouldn’t

Let’s get practical. Here’s our honest view of who RTX Spark makes sense for, based on what we know right now.

Good candidates

AI developers and researchers who need portability. This is the clearest fit. If you build or test AI applications and want to run large models locally, with CUDA, on a laptop you can carry to a client meeting or a university lab, RTX Spark is the most interesting option on the Windows side. A high-memory configuration (64GB or ideally 128GB) gives you capacity no other Windows laptop with an Nvidia GPU can match.

Creators who travel. Video editors, 3D artists and photographers who use apps with native Arm versions (Premiere, Resolve, Blender, Photoshop and friends) and want a powerful machine that isn’t a heavy, loud performance laptop. The big unified memory pool helps with large 3D scenes and high-resolution video in a way limited VRAM can’t.

People who want one premium laptop for everything, and can afford early adoption. If you want a thin, high-end Windows machine that can handle creative work, some games and serious AI experiments, and you’re comfortable checking compatibility and living with the occasional rough edge of a brand-new platform, RTX Spark could be a great all-rounder.

Probably not for you, at least yet

If games are your main reason for buying. As covered above, a normal laptop with a GeForce RTX 50 GPU, or better still a desktop, offers better compatibility and better value for games in 2026. Online titles like Call of Duty won’t even run on RTX Spark until 2027.

If you rely on specialised software or hardware. Engineering tools, industry-specific apps, older printers and scanners, audio interfaces, certain VPN or security software: check every single one. If anything critical lacks Arm support, this isn’t your laptop.

Students and budget buyers. At $2,600 and up in the US before any import costs, RTX Spark is premium hardware. For study and everyday use, a good mid-range x86 laptop or a Snapdragon X laptop will do the job for a fraction of the price.

Anyone who wants to upgrade later. Memory is soldered, and so is the GPU, because it’s all one chip. If you like the idea of upgrading parts over time, that’s what desktop PCs are for.

Anyone who hates being a beta tester. First-generation platforms usually have early bugs: driver issues, firmware updates, apps that misbehave. They get fixed, but if you want something that just works from day one, waiting six months for reviews, updates and maybe cheaper second-wave models is very reasonable.

A buying checklist

If you are seriously considering one, run through this list first:

List your apps. Every program you use regularly. Check whether each has a native Arm64 version, runs well under Prism, or doesn’t run.
List your devices. Printers, scanners, drawing tablets, audio gear, docks, anything that needs a driver.
List your games. Especially online ones with anti-cheat. Compare against Nvidia’s launch list and the “coming in 2027” list.
Pick the right chip. 18-core/5,120 or 20-core/6,144. The difference matters for heavy work.
Pick memory for the next five years. You can’t add more later. For AI work, 64GB is a sensible floor and 128GB is ideal. For creative work, 32GB minimum. Don’t buy the 24GB base model if you plan to run large models.
Check storage upgradability. Surface Laptop Ultra’s SSD is user-replaceable, which helps. Check other models individually.
Wait for reviews of the exact model you’re considering, especially battery life, heat and noise.
Check warranty support where you live (more on that for Pakistan below).

Part 11: What it means for desktop PC builders

Most of our customers build or buy desktop PCs, so it’s fair to ask: does any of this matter if you’re never going to buy a $2,600 laptop?

In the short term: not much

RTX Spark is a laptop and small-form-factor chip. It doesn’t affect the parts you put in a desktop tower today. Your next desktop is still going to use an Intel or AMD processor, a socketed motherboard and, if you want serious graphics, a discrete graphics card. Desktop GPUs remain far more powerful than anything in a thin laptop, and for games and most creative work, a well-built desktop still gives you much more performance per rupee, plus the ability to upgrade.

There’s also the compact desktop angle. Microsoft’s Surface RTX Spark Dev Box and similar small desktops could appeal to developers who want a quiet, tiny AI box on their desk, the same niche DGX Spark targeted. For most home users though, a normal desktop with a good graphics card is the better buy.

In the longer term: it could change a lot

The bigger story is that Windows is no longer tied to x86. With Qualcomm and now Nvidia making serious Arm chips for Windows, software developers have much stronger reasons to build native Arm versions of their apps and games, and anti-cheat vendors have to support Arm. That benefits the whole Windows on Arm ecosystem over time.

There have already been interesting side effects. One report in 2026 described how someone managed to get a regular, standalone Nvidia graphics card working on a Windows on Arm system using the RTX Spark drivers, although game performance didn’t look ideal yet. It’s an early experiment, but it hints at a future where Arm-based desktops with discrete graphics cards might become possible.

Intel and AMD aren’t standing still either. Both have next-generation desktop processors on the way, Intel’s Core Ultra 400 “Nova Lake” and AMD’s Zen 6 Ryzen, which leaks suggest are likely to arrive around 2027. Competition from Arm pushes everyone to improve efficiency and integrated graphics, which is good news for buyers regardless of which camp you’re in.

For desktop builders, our advice hasn’t changed: build on a platform with a future (AMD’s AM5 has support promised through 2029), choose a graphics card with enough VRAM for what you do, and if local AI is a big interest, VRAM and system RAM matter more than ever. We’ll cover that in detail in our upcoming article on running AI on your own PC.

Part 12: Buying from Pakistan, things to think about

RTX Spark laptops launch first in markets like the US, and as with most premium laptops, they’ll reach Pakistan either through official channels later or through importers. A few things are worth thinking about before buying one here.

Price

The US starting prices of $2,599 for the Surface Laptop Ultra and $3,799 for the Dell XPS 16 Creator Edition are before any import duties, taxes, shipping and seller margins, and before local sales tax. Imported premium laptops in Pakistan typically cost noticeably more than their US sticker price once all of that is added. Higher memory configurations, which are the ones that matter most for AI work, cost much more again. Budget accordingly and compare final landed prices rather than converting the US price in your head.

Warranty and support

This is the big one. A laptop imported without an official local warranty means any repair may involve shipping it abroad, which is slow, expensive and risky for a machine that costs this much. With a first-generation platform, the chance of needing support in the first year is a bit higher than usual. Ask clearly who provides warranty service, for how long, and what happens if the motherboard fails. Since the memory and processor are part of one board, a hardware fault is almost always a mainboard replacement.

Keyboard layout and power

Small things, but worth checking: imported units may have a US keyboard layout (usually fine here) and a US-style charger plug, which will need an adapter or a replacement cable. USB-C charging makes this easier than it used to be.

Software and local services

Check that any local software you need works on Windows on Arm. Some banking security apps, government software, and older business tools used in Pakistan are x86-only and may rely on drivers or plugins that don’t work under emulation. If your work depends on a specific local application, test it or ask the developer before buying.

Our honest suggestion

For most buyers here, RTX Spark is a “watch and wait” product in 2026. Let the first wave of reviews come out, let prices settle, and see which models get proper local distribution and warranty support. If you’re an AI developer or creator who needs one now, buy from a seller who can clearly explain the warranty, and choose the memory configuration carefully, because you’re stuck with it.

Part 13: Myths and misunderstandings

“RTX Spark is a graphics card.” No. It’s a complete processor package, CPU, GPU and memory together, that goes into laptops and small desktops. You can’t buy it as a card to put in your existing PC.

“It’s just a Snapdragon laptop with Nvidia branding.” Not really. It uses a different CPU design (developed with MediaTek, not Qualcomm), and the GPU is far more powerful than the graphics in Snapdragon chips. Both run Windows on Arm, which is where the similarity ends.

“It runs everything the world has ever created.” That’s Jensen Huang’s line, and it’s marketing. Most mainstream software runs, either natively or through emulation. Some drivers, specialised tools and anti-cheat games don’t yet. Check your own list.

“One petaflop means it’s faster than a desktop RTX 5090 for AI.” The petaflop figure is measured at FP4, a very low precision format, and peak compute isn’t the whole story. For generating text from AI models, memory bandwidth matters enormously, and a desktop RTX 5090’s GDDR7 is several times faster than RTX Spark’s LPDDR5X. RTX Spark’s advantage is that it can hold much bigger models, not that it runs them faster.

“Arm means bad performance.” That used to be a fair assumption on Windows, but it’s outdated. Apple’s Macs proved Arm can be very fast, and leaked RTX Spark CPU results put it ahead of most x86 laptop chips in multi-core tests. Emulated apps lose some performance, native apps don’t.

“Every RTX Spark laptop is 14mm thin and weighs 1.4kg.” That’s the thinnest the platform allows. The first big launch models, the Surface Laptop Ultra and Dell XPS 16 Creator Edition, are both around 18mm and 2kg.

“You’ll be able to add more RAM later.” No. The memory is part of the package. What you buy is what you’ll always have.

“This is the end of Intel and AMD.” Very unlikely any time soon. The vast majority of Windows PCs, and nearly all desktops, will keep using x86 for years. What RTX Spark does is add real competition, which is good for everyone.

Part 14: Frequently asked questions

What is Nvidia RTX Spark?

RTX Spark is Nvidia’s Arm-based “superchip” for Windows laptops and compact desktops. It combines a Grace CPU with up to 20 cores, a Blackwell RTX GPU with up to 6,144 CUDA cores, and up to 128GB of unified LPDDR5X memory, connected by NVLink-C2C. It was announced on May 31, 2026, and the first laptops go on sale on October 16, 2026.

Is RTX Spark the same as N1X?

N1X was the codename used in leaks and early reports. RTX Spark is the official product name. It’s closely related to the GB10 chip used in Nvidia’s DGX Spark AI desktop.

Which laptops use RTX Spark?

At launch, the Microsoft Surface Laptop Ultra (from $2,599) and Dell XPS 16 Creator Edition (from $3,799) are the most detailed. ASUS, HP, Lenovo and MSI have models coming, and Acer and Gigabyte will follow. Microsoft also has the Surface RTX Spark Dev Box desktop from $6,000.

Can RTX Spark run normal Windows apps?

Yes, most of them. Many popular apps run natively on Arm, and most other x86 apps run through Microsoft’s Prism emulator. Apps that depend on low-level drivers, some specialised professional software and certain older hardware may not work.

Can I play games on RTX Spark?

Yes. Nvidia says hundreds of games are playable at launch, including Cyberpunk 2077, Fortnite and Hogwarts Legacy, and anti-cheat games like Counter-Strike 2 and Marvel Rivals. Call of Duty and PUBG are expected in 2027. For dedicated play, a traditional x86 laptop or desktop with a GeForce GPU is still more compatible.

How fast is RTX Spark for games?

Nvidia claims over 100fps at 1440p in big games with ray tracing, DLSS and Reflex enabled. Some reports suggest graphics performance in the range of an RTX 5070 laptop GPU depending on the game. Independent reviews of retail units hadn’t been published at the time of writing.

Is RTX Spark good for AI?

For running large AI models locally on a laptop, it’s one of the most capable options available, mainly because of its large unified memory and Nvidia’s CUDA software support. Text generation speed is limited by memory bandwidth, so it’s better at holding big models than running them very fast.

Does RTX Spark support DLSS 5?

Yes. Nvidia says DLSS 5 is available on launch day in NBA 2K27, with Hogwarts Legacy getting it on November 5 and Where Winds Meet later in November.

Can I upgrade the RAM on an RTX Spark laptop?

No. The unified memory is part of the chip package and can’t be upgraded. The SSD can be upgraded on some models, including the Surface Laptop Ultra.

Should I buy RTX Spark or a MacBook Pro?

If you need Windows, CUDA-based AI tools or PC games, RTX Spark makes more sense. If you’re happy on macOS and want the most mature Arm laptop experience, the MacBook Pro is the safer choice for now.

Will RTX Spark laptops be available in Pakistan?

They’ll likely appear through importers and possibly official channels over time. Expect prices well above US list prices once duties and taxes are added, and pay close attention to warranty support.

Should I wait for the second generation?

If you don’t need one urgently, waiting for full reviews, software updates and possibly cheaper models is sensible. First-generation platforms almost always improve after launch.

The bottom line

RTX Spark is the most serious attempt yet to change what a Windows laptop is. It takes Nvidia’s graphics and AI strengths, pairs them with a capable Arm CPU and a huge pool of unified memory, and puts it all in machines thin enough to carry every day. For AI developers and creators, that’s genuinely exciting, there’s nothing else quite like it on Windows.

But it’s also a first-generation product on a platform that still has gaps. Windows on Arm has come a very long way, yet some apps, drivers and games still don’t run, the most popular online shooter won’t arrive until 2027, and the laptops are expensive, with memory you can never upgrade. And as of launch day, nobody outside Nvidia, Microsoft and their partners has properly tested a retail unit.

Our advice: watch this space closely, read the reviews when they land, and if you’re not an AI developer or a creator who really needs this kind of portable power, give it a generation to mature. For desktop builders, nothing changes today, a good desktop with a proper graphics card is still the best value for performance. But the long-term direction of Windows just got a lot more interesting.

If you’re trying to decide between a laptop and a desktop for your work, or want help speccing a desktop that’ll last, get in touch. We’re always happy to talk it through, and sometimes the honest answer is that what you already have is good enough for now.

Sources

Planning a build or an upgrade?

Tell us what you already have and we'll help you spend only where it matters.

Browse the shop Ask on WhatsApp