One Week with RTX Spark: Insights from Nvidia’s Latest Laptops

Nvidia’s RTX Spark N1X processor has officially launched, and we’ve been putting a range of high-end laptops featuring this new “superchip” through their paces. The RTX Spark is crafted to deliver top-tier performance as an Arm-based SoC for laptops and compact desktops, competing with the likes of Apple, Intel, AMD, and Qualcomm. With a multitude of cores, Nvidia RTX Blackwell graphics, and ample power along with unified memory, it’s capable of handling substantial local AI models.

The standout device is Microsoft’s Surface Laptop Ultra, boasting an impressive 128GB of unified memory, which has been thoroughly reviewed. Additionally, there’s a 64GB version of the Asus ProArt P16 that has also been tested, alongside a top-spec 128GB configuration. I’m also working with a 64GB Lenovo Yoga 9n 2-in-1 and a 32GB HP OmniBook Ultra.

All these laptops are equipped with the same N1X chip featuring 20 CPU cores and 6,144 GPU CUDA cores, and they come with hefty price tags — from the HP OmniBook at $3,199.99 to the Asus ProArt’s staggering $6,999.99. There’s a more budget-friendly 18-core / 5,120 CUDA N1X chip available, starting at $2,599.99 for a Surface Laptop Ultra or a smaller Asus ProArt P14 with 24GB of RAM.

These laptops are designed for developers, coders, and content creators who also want the capability to utilize local AI and enjoy gaming with the latest Nvidia technologies like DLSS 5. There’s a lot to explore, and while we haven’t had extensive time to test everything, below are our impressions from a week of use across five laptops.

Is Nvidia’s “superchip” truly impressive? Are competitors like Apple, Intel, AMD, and Qualcomm on alert? And have we managed to catch any sleep?

**Does it just work?**

Antonio Di Benedetto: The experience of using Windows on Arm has significantly improved compared to a few years ago. Thanks to advancements with Qualcomm’s Snapdragon X chips, there’s been a surge in native Arm-based applications for Windows. Many x86 applications also run smoothly through Microsoft’s Prism emulation layer.

Adobe has offered native Arm versions of Photoshop and Premiere for a while, and even though Lightroom Classic remains x86-based, I haven’t encountered any issues using it on Arm. Unless you’re relying on specialized software that’s strictly x86, Windows on Arm is quite functional — especially for gaming without compatibility concerns.

Sean Hollister: Occasionally, I’ve noticed apps taking longer to load than expected, leading me to wonder if they would even open. However, they usually do, even if it sometimes takes a couple of attempts. I experienced a few hiccups with Evernote disappearing when I alt-tabbed, but a quick refresh resolved that. Some apps also don’t seem to respect my dark mode preference.

Overall, I share Antonio’s sentiment. I’ve been using an RTX Spark laptop as my daily driver with all my usual apps — Signal, Discord, Audacity, CapCut, Evernote, Irfanview, Slack, ScreenToGif, and more — and everything has functioned without a hitch. I never had to think about whether I was using an Arm or x86 app. The real challenges arose with gaming and AI, which you’ll read about next.

Tom Warren: After using a Windows on Arm laptop with a Qualcomm chip for the past two years, I can confidently say that the experience has improved to the point where I can’t distinguish between native and non-native apps anymore. The same holds true for the RTX Spark laptops, and I believe Nvidia’s entry into this space will enhance the experience for creative applications and gaming.

**Can it game?**

TW: Windows on Arm has never been seen as a serious gaming platform until now. While native Arm games are still scarce, Nvidia’s GPU helps push through the emulation layer. Your gaming experience will vary by title, but generally, I’ve found that medium settings at 1440p with DLSS performance can achieve a stable 60fps in most games.

I typically game on my desktop with an RTX 5090 on a 32-inch 4K display at max settings, but I didn’t mind the medium settings on a 15-inch screen. I managed to run Black Myth: Wukong at an average of 70fps at native resolution (3270 x 2180) with medium settings and DLSS performance mode enabled. Cyberpunk 2077 also performed well, averaging 66fps, while Forza Horizon 6 hit around 57fps.

Some titles may require lowering the resolution or settings, though. This setup isn’t as powerful as a dedicated gaming laptop, but for something comparable to a MacBook Pro, it’s more than adequate for gaming on the go.

SH: As Tom mentioned, the GPU power is impressive, but the CPU might limit some gaming experiences. I achieved smooth 60fps in Returnal (at 1440p with high settings), Expedition 33 (DLSS “Performance” upscaled 1200p with high settings), Jedi Survivor (DLSS “Performance” upscaled 1200p with medium settings), and Shadow of the Tomb Raider (native 1800p at highest settings).

However, when I tried Helldivers 2, even dropping to a DLSS-upscaled 1080p didn’t guarantee 60fps during the most challenging missions due to the number of enemies on screen. Star Wars Zero Company was even more CPU-intensive, with framerates capped between 37fps and 39fps regardless of graphics settings.

Even older, less demanding games can have quirks. I tested the 2003 title Max Payne 2 and noticed framerate dips to 50fps. I also encountered some stutters in the excellent UFO 50 collection, although other 2D games ran smoothly.

I’m uncertain if the RTX Spark is suitable for VR yet. The Titanfall 2 VR mod worked initially but lost stability after a few minutes. This could be due to my Wi-Fi, as the driver for Valve’s Steam Frame wireless adapter wouldn’t install on the Spark, nor did Steam’s streaming audio solution, resulting in audio output from the laptop instead of the headset.

Just to clarify, all of this assumes you’re okay with a laptop that runs hot and loud. That’s typical for thin gaming notebooks, but it’s noticeably louder and hotter than a good desktop or handheld.

AD: I’ve longed for a MacBook Pro-like laptop that can handle full PC gaming, so I was excited to see if these laptops could deliver. Setting up the Asus ProArt P16 with its massive 128GB of memory, I booted up Big Walk for an online session with friends. My initial thought was, “Well, that looks okay, I guess.” While Big Walk isn’t a graphical powerhouse, I expected the N1X to handle it better.

I had to lower the P16’s 4K display to 2560 x 1600 and adjust the graphics settings from Highest or Very High to High to maintain a solid 60fps. Does it matter for a co-op puzzle game? Not really, but it was a reminder that this “superchip” isn’t a graphical miracle worker.

As Tom and Sean noted, RTX Spark can handle decent gaming, but manage your expectations for medium or low settings in some cases, and be prepared to lean on DLSS and frame generation. There’s no denying that more affordable desktops and laptops with discrete Nvidia graphics cards can outperform an N1X. Nvidia claims it’s akin to having an RTX 5070 Laptop GPU in an SoC, but in my experience, it’s more comparable to an RTX 5060 Laptop GPU — still capable, but if you prioritize gaming, traditional Nvidia graphics cards offer better value.

**How’s the battery life?**

SH: Nvidia’s partners have equipped their first wave of RTX Spark laptops with some of the largest batteries permissible for air travel — but even with a 99.9 watt-hour pack like the one in my Asus ProArt P16, these laptops won’t win any longevity awards. While Tom and Antonio tested various use cases, I made the P16 my daily driver for regular work while closely monitoring the battery. This mostly involved writing stories in a web browser, and it didn’t quite last the entire day.

My best run was 5 hours and 40 minutes of real-world work at 50 percent brightness, including an hour-long video call with Tom and Antonio to strategize our RTX Spark coverage. The next day, I only got about 5 hours from a full charge because I spent two of those hours editing a video in CapCut. The third time I drained the battery, it also took 5 hours.

With web work, I typically saw a battery drain of 13-17 watts on default settings, which increased to 20-30 watts with light video work, up to 50 watts during brief video exports, and spiked to 120 watts while playing Indiana Jones and the Great Circle. The battery couldn’t sustain that rate, and at 30 percent charge, I experienced a significant performance drop to 60 watts.

The good news is that the Asus charges quickly from a 140W USB-C PD power source. It took 20 minutes to go from dead to 50 percent, another 15 minutes to reach 70 percent, and a total of 1 hour and 15 minutes to hit 95 percent.

AD: I’ve been running numerous benchmarks while plugged in, so I haven’t had as much time to test battery life as I’d like. Anecdotally, it seems decent. As Sean noted, you’re unlikely to get through a full eight-hour workday unless you’re being very efficient. If you run anything demanding, like a local AI model, you won’t even make it to lunch. I plan to conduct more thorough battery tests as I continue using these laptops, including our usual rundown tests for direct comparisons.

**Does local AI seem useful?**

TW: Not yet. It feels like a beta aimed at developers right now, from the installation process of AI agents flashing command prompts and Visual Studio to the fans ramping up for even basic queries.

I struggled to find compelling use cases for the power of these chips in local AI work. Many tasks could be completed much faster and without the laptop fans kicking in by using a free cloud model. The real advantage lies in running an agent at full throttle to build something overnight, handling tasks that developers would appreciate.

It’s also nice to have everything on a local device, avoiding the need to share emails and files with a cloud service and the associated privacy concerns. Still, I often found myself closing OpenClaw or Hermes and not letting them run in the background, especially when they had access to local files. I’m just not ready to fully trust AI, whether local or cloud-based.

Both Microsoft and Nvidia envision a future where AI agents automate tasks on your PC, but I’m not entirely convinced. While there are mundane tasks I could automate, I enjoy using my PC for photo and video editing, writing, and gaming. I don’t want to watch a robot do all that for me; that’s part of the joy of computing.

SH: After Nvidia’s Editor’s Day, I thought I might finally have a reason to try local AI. I asked Hermes, running Qwen3.8-27B, to create a daily news briefing similar to the one shown in Nvidia’s demo. It took 10 minutes and 10 percent of my entire 99.9 watt-hour battery to produce a list of stories that the cloud-based Claude (Opus 5.5) generated in half the time — while monopolizing my GPU and heating up the Spark laptop. The laptop remained hot and sluggish as long as Hermes was open.

Claude also delivered better results, with more detail and better sourcing. While I got the local AI tokens for “free,” I wouldn’t trust Claude with my email, and I probably could have used a smaller, lighter model. I went with the program’s recommendation, and that was the outcome. I even tried a clean reinstall of Hermes to rule out misconfiguration, but this time, the prompt took even longer and consumed 60 percent of my battery. Ouch.

AD: I’m still in the early stages of learning to use local AI, and these RTX Spark laptops are certainly a viable — albeit pricey — option. However, even with 128GB of memory for larger models, the experience isn’t user-friendly for beginners. Microsoft’s Get Started intro in Windows 11 offers easy one-click installs for Hermes Agent, OpenClaw, and Perplexity (which all have ESRB ratings for some reason), but after that, you’re largely on your own. I stuck with Hermes since I was already using it on the Mac Studio, but if you’re on a system with 32GB of RAM or less, it simply tells you there isn’t enough memory for its recommended local Qwen models — either directing you to the cloud or leaving you to find a smaller model yourself.

Using local AI has significant privacy benefits over cloud models, but it’s less convenient since local LLMs are slower to process and more demanding on your system. Running a local model through your laptop’s memory drains the battery quickly, and the results from LLMs can be hit or miss. I drained 30 percent of the Lenovo Yoga 9n’s battery in 36 minutes while getting a Qwen 3.8 27B model to resize and add a white background to five product images in PNG format — and it cropped the images incorrectly, cutting off the products. After a couple of follow-up prompts, I managed to get it right, but I had to grab the power cable to prevent Hermes from draining the laptop in the process.

**Is the hardware truly great?**

AD: These laptops come with a hefty price tag, but that also means the manufacturers are bringing their best. Across the Asus, HP, Lenovo, and based on Tom’s feedback on the Surface, the RTX Spark line features excellent haptic trackpads, keyboards, displays, and speakers. The metal builds of the three I’m testing feel premium. The OLED screens look stunning, and each of these laptops is relatively thin for workstation-class devices.

If I had to choose one based solely on hardware, I’d likely go for the Asus ProArt. It checks all the boxes for me with three USB-C ports, a bright display, a superior inverted-T arrow cluster, and a full SD card slot. The HP feels the most solid, with a robust chassis, but at nearly five pounds, it’s the heaviest, and its microSD card slot is less useful.

The Lenovo has become my sleeper favorite. As the only 2-in-1 in this group, it offers more flexibility and fun. You can use its large stylus on the screen or the spacious trackpad, making it great for graphics and design work. Tent mode with a large 16-inch 120Hz OLED is a joy for occasional gaming or watching shows and movies, though it’s not as bright as the Asus.

SH: I’ve primarily tested the Asus P16, and I was impressed with the build quality of the entire range of RTX Spark laptops at Nvidia’s Editor’s Day. The Asus P16 might not be my personal choice due to its glossy screen and overly sensitive touchpad. I prefer using laptops in bright environments, and I find myself needing to crank up the brightness on this otherwise excellent OLED display. Additionally, my palms often brush the touchpad while typing, causing unintended cursor movements.

The speakers and webcam could use improvement, but the Windows Hello face login and proximity sensing are fantastic. The screen dims when I walk away, and I’m logged back in just by looking at my laptop. Despite the heat, the Asus fans are loud but not shrill — they produce a whoosh rather than a whine during gaming.

TW: While I encountered some odd bugs with the Surface Laptop Ultra during testing, I can’t fault the hardware. It has a similar look and feel to a MacBook Pro, and the magnetic USB-C port is a feature Apple should have implemented — it’s that good. The large haptic trackpad continues the tradition of great Surface trackpads, with the added benefit of subtle haptic feedback when moving windows or aligning objects in apps like PowerPoint.

I would have preferred an OLED display on the Laptop Ultra, but switching from the default “Vivid” color option to “Enhanced” improved the contrast on the mini-LED panel. I also appreciate the port selection on the Laptop Ultra, especially the full SD card reader and USB-A port.

**Is it fast? Does it compete with Apple M-series chips?**

AD: Yes, the RTX Spark N1X is a powerful chip. The high-end laptop configurations we’ve been testing, with their substantial unified memory, are designed to compete with the Pro and Max chips in Apple’s MacBook Pros. While they can match some GPU performance tests, like tying the top-spec M5 Max in a Blender render, Apple’s CPUs are generally more powerful.

Apple has invested heavily in its in-house processors, particularly the fast “super cores” of its M6, M5 Pro, and M5 Max chips. In terms of single-core performance, nothing competes with those chips — not even Nvidia’s “superchip.” The most disappointing aspect of the RTX Spark’s 20-core CPU is that even a basic 12-core M6 Mac Mini competes with it in multicore performance, and a Mac Mini starts at just $900.

It’s impressive to see this level of graphics performance without needing a discrete GPU, allowing for thinner designs with quieter fans. I’m all for it, just as I was with AMD’s Strix Halo APU — but like Strix Halo, Nvidia’s RTX Spark is a costly component. There’s no denying that Apple’s offerings are more well-rounded performers, especially if gaming isn’t your primary focus.

TW: It’s definitely competitive, but I think Qualcomm comes closer to Apple’s M-series chips on the CPU front. Qualcomm has had years of experience with Windows on Arm before acquiring Nuvia and developing a competitive chip with the Snapdragon X series.

Nvidia will need to invest similar effort to catch up on the CPU side for future RTX Spark chips. The development of the RTX Spark took at least three years, and I’ve heard from sources that the chips were initially slated for release last year. So, while progress is being made, the CPU performance still leaves something to be desired.

Nvidia chose MediaTek for its Arm-based CPU, a company better known for Wi-Fi hardware in laptops than CPUs. I hope the next generation offers more power. Throughout my time testing the Surface Laptop Ultra, I couldn’t help but dream of a combination of a Qualcomm CPU and Nvidia GPU. While that may never happen, it would certainly be intriguing.

**So what’s our take on RTX Spark so far?**

TW: I believe it will spark interest in Windows on Arm for creative and gaming tasks, as we haven’t had capable Arm-based Windows laptops for gaming until now. Nvidia is targeting creators and developers with its initial RTX Spark laptops, but there’s definitely a need for a more affordable version aimed at gamers. Perhaps that’s where the rumored N1 chips come into play.

SH: They’re impressive, but too expensive. They run hot and loud during tasks that would justify that price. I care more about what they represent and the direction they’re headed than about owning one.

AD: I appreciate the increased competition in the laptop market, but the exorbitant prices make RTX Spark laptops largely unattainable. I hope to see more affordable options in the future, along with a second generation that boosts CPU performance to better compete with Apple. I also hope this drives development for Arm-based and Arm-compatible games, potentially leading to easier game porting to macOS. That would be a fun outcome.

While it’s exciting to continue testing these high-end RTX Spark models, I’m particularly eager to try a base Surface Laptop Ultra or Asus ProArt P14 with 24GB of RAM to see how they compare to the entry-level 14-inch M5 MacBook Pro. That’s my personal sweet spot in Apple’s lineup — the people’s MacBook Pro. (Well, it used to be, especially now that it starts at two grand.)

It’s an exhilarating time in computing regarding chip options, but it’s also a frustrating time due to skyrocketing RAM prices. I often wonder how much laptops like these RTX Sparks or even Googlebooks would cost if AI demands hadn’t inflated RAM and storage prices. I’m not sure who will spend up to $7,000 on these laptops when overall PC sales are declining, but it’s cool that they exist.