Microsoft Surface Laptop Ultra review: A monster AI laptop with a monster price | usagoldmines.com

Price When Reviewed

Base price: $2,599.99 Reviewed price: $3,699.99

Best Prices Today: Microsoft Surface Laptop Ultra

Retailer

Price

$3,699.99

View Deal

$3699.99

View Deal

$3699.99

View Deal

Price comparison from over 24,000 stores worldwide

Microsoft’s Surface Laptop Ultra and its newfangled Nvidia RTX Spark chip characterize what many would call a new axis of productivity: AI. With it comes an old, familiar feeling of FOMO, or the fear of missing out.

For decades, the equation has been simple: A faster CPU means applications run faster. A faster GPU equates to higher frame rates in PC games. Now there’s AI, which means memory matters. Memory is shared between the Arm CPU and powerful RTX GPU inside the Nvidia Spark N1X chip powering the Surface Laptop Ultra, and more memory means the ability to load more powerful AI models.

As regards the Surface Laptop Ultra model that Microsoft shipped us, my first impression was one of disappointment, as it came with just a “measly” 32GB of memory (along with its whopping $3,699 price tag). By contrast, RTX Spark machines with 48GB to 128GB of RAM encourage users to load AI agents as part of the setup process. Ours and others like it do not, implying that there’s both first- and second-class experiences within the Surface Laptop Ultra and other RTX Spark devices. A new AI model is like a new video game: We all want to play it. Some can’t.

So here’s how I approached it: I began first by testing the Surface Laptop Ultra as an “ordinary” productivity device, like every other laptop we touch. I then pivoted into examining it as a content-creation machine, with tests of image and video editing. After I touched on gaming, I then took a harder look at the Ultra in its natural setting: as a portable AI workstation, owned by those with money to burn.

For me, the latter half of this review was a chance to ask the question, why do we need the Surface Laptop Ultra when gaming laptops exist? I think it’s one worth exploring, with valid reasons both for and against this new class of devices.

Mark Hachman / Foundry

Microsoft Surface Laptop Ultra: Configuration options

Microsoft’s Surface Laptop Ultra starts at a base price of $2,599, which is hundreds of dollars more than the existing 15-inch, $1,615 Surface Laptop 8. That laptop uses Qualcomm’s Snapdragon X2 Elite Arm CPU, while the Surface Laptop Ultra ushers in a more powerful Arm alternative, the Nvidia RTX Spark N1X, a collaboration between Mediatek and Nvidia.

While you don’t have any choice in different display sizes, you can choose between two basic configurations, a Surface Laptop Ultra with an 18-core N1X CPU/5,120-core GPU RTX Spark N1X ($2,599 and up) or a version with a 20-core RTX Spark N1X with a 6,144-core GPU ($3,699 and up). More importantly, the two configurations include different memory options: 24GB and 32GB for the 18-core version, and 32GB, 48GB, 64GB, and 128GB for the 20-core version. (At press time, the 128GB option was listed as out of stock.)

There are three storage options, a PCI Express Gen 4 512GB option on the base model, plus the choice of between 1TB and 2TB of PCIe Gen 5 SSD storage. The latter is reserved for the 64GB RAM configuration, not the 128GB RAM option.

You have two choices of colors: Nightfall (black) and Platinum. Not all configurations are available in all colors.

Oh, and Microsoft’s getting snarky: Trade in your MacBook Pro and get up to $1,000 back. That’s good, as the top price for a Microsoft Surface Laptop Ultra is $5,899.

Surface Laptop Ultra features:

• CPU: Nvidia RTX Spark N1X 18/20 core CPU, 5,120/6,144-core GPU (20-core CPU as tested)
• Memory: 24GB/32GB/48GB/64GB/128GB LPDDR5x (32GB as tested)
• Graphics/GPU: RTX Blackwell 5,120-core/6,144 core GPU (6,144-core GPU as tested)
• NPU: Yes? But maybe not? To be continued below.
• Display: 15-inch (3720 x 2180) PixelSense Ultra mini-LED display
• Storage: 512GB removable PCIe4 SSD/1-2TB removable PCIe5 SSD (1TB as tested)
• Webcam: 1080p
• Connectivity: 2 USB4 USB-C, 1 USB4 USB-C (magnetic connect), HDMI 2.1b, 10Gbps USB-A, SD, 3.5mm headphone jack
• Networking: Wi-Fi 7, Bluetooth
• Biometrics: Windows Hello (webcam)
• Battery capacity: 89Wh (95Wh as tested)
• Dimensions: 12.95 x 9.4 x 0.71 inches
• Weight: 4.41 pounds
• MSRP: $2,599 to $5,899 ($3,699 as tested)

Surface Laptop Ultra: Out of the box and construction

The word to describe the Surface Laptop Ultra is dense.

Many other laptops feel like a construction of parts. The Surface Laptop Ultra feels like somehow the electrical traces and fans and compute were all routed through a single piece of metal. Technically, it’s made from “precision‑molded polymer composites” for the main body, combined with “aluminum and magnesium alloy frames” for structural rigidity and heat dissipation. The keyboard tray feels slightly plasticy, but that’s all.

Otherwise, it’s about the most solid laptop I’ve ever used. There’s no give to the keyboard, and I didn’t notice any twisting to the display, which reclines to about 40 degrees or so off of the horizontal.

Mark Hachman / Foundry

Physically, the most noteworthy addition to the Laptop Ultra is the return of the magnetic I/O connector. For over a decade, Microsoft built its Surface devices with a magnetic Surface Connector, which delivered power and (if it was connected to a Surface Dock) data as well. With the Surface Laptop Ultra, Microsoft designed in something novel: a magnetic USB-C connector, which snaps in and out securely to avoid yanking the laptop off a table if ever your foot accidentally catches the cord. I’ve never had that happen, but it’s a comfort nevertheless.

The thing I don’t like about the charger is that it’s essentially a larger version of the existing Surface charger, with the plug built right into it. That means that the 140-watt USB-C charger (its most powerful to date) monopolizes at least one end of a power strip, unless you buy one with plugs perpendicular to the strip itself. You can buy a cord extension to the charger which eliminates the problem, but that costs extra. Unfortunately, it’s also proprietary, too, as I was unable to plug the Surface’s magnetic USB-C adapter into other laptops.

Otherwise, if you’re a fan of ports, the Surface Laptop Ultra has them in spades: a whopping three USB4 (the generic equivalent of Intel’s Thunderbolt 4) ports, including one with the magnetic connector; an HDMI 2.1b port; a 10Gbps USB-A port; an SD card slot, and a 3.5mm headphone jack.

Normally, most laptops vent air by pulling it from a grille on the underside of the laptop and routing it over the components, then outside, typically via the hinge. Instead, this laptop appears to pull air through grillwork mounted on the outer edges, much like the older Surface Laptop Studio series. The fans are loud, though not especially noisy. I thought a heard a note of coil whine at some point, but it didn’t last.

Mark Hachman / Foundry

Surface Laptop Ultra: Display

Microsoft has flirted with OLED screens in the past, which produced some speckling. In this case, the company chose a Mini-LED-based PixelSense Ultra screen instead, the first Surface to use the technology.

Mini-LED displays use a collection of small LED lights, rather than a giant LED backlight, providing a middle ground between an IPS display and an OLED. An OLED’s pixels can be turned off and on individually, as can the Mini-LEDs, and can suffer a “bloom” or “halo” effect while they’re on. I didn’t notice anything like that, which makes me think there are quite a few Mini-LEDs behind the display. (I asked, but never heard back from Microsoft.)

Microsoft provides three modes: The default Vivid mode, a mode for sRGB video and photo editing, and an “enhanced” mode for watching videos, with more visual pop that tends to dial up the contrast. I’ve included the color gamut for the default mode below.

Mark Hachman / Foundry

Microsoft claims that the display puts out 1,000 nits while in SDR mode, with HDR brightness at about 2,000 nits. I found the display put out far less than that, with 792 nits maximum in an SDR mode. It’s still a terrific display, but my review unit didn’t live up to its promised ratings.

Surface Laptop Ultra: Mics and audio

Microsoft has always manufactured Surfaces with surprisingly solid audio, even including its earlier tablet offerings. The Surface Laptop Ultra lives up to its reputation.

When I played back my test tracks, the music sounded excellent up and down the spectrum, maintaining cohesion even at maximum volume — which is more than sufficient for individual playback or for watching a movie with a companion or two. I’m not going to say that it sounds like the best laptop ever, but it’s certainly top tier.

As for the mics, voice recording is superb, and my voice sounded warm and natural, honestly about as good as most mics I’ve used.

Noise cancellation was surprisingly a mixed bag, however. I test by holding a phone playing back music (a rock track as well as white noise) behind my head and then move it off to the side, while recording my voice using the Windows Sound Recorder app and the built-in mics. I expected more here. Other laptops (Asus generally nails this) perform audio filtering quite well.

The laptop stumbled rather badly trying to distinguish my voice from the background music, but did far better when cancelling out the white noise of a recorded rainfall. In that case, it preserved my voice’s tones while wiping out the sounds of the rain almost completely.

For all of its “AI” positioning, I didn’t see an NPU listed in the spec sheet, nor did I see it active while recording audio. Windows reported one, though. Supposedly this is a Copilot+ PC, yet I see no rated NPU TOPS.

Surface Laptop Ultra: Keyboard, trackpad

The best laptop keyboard I can recall using was the sadly discontinued Surface Book series, whose key travel rivaled that of a mechanical keyboard. The Surface Laptop Ultra feels on par with recent Surface Laptops, and perhaps a bit better. The key action feels a bit stiffer than other laptops I compared, and might loosen up a bit over time. Overall, it’s a good keyboard, though not outstanding.

Microsoft continues to play with its function keys, however. On the top row of the keyboard there’s now a dedicated microphone on/off toggle, plus controls for adjusting the screen brightness and volume. There’s also a screen capture key, the first I can recall seeing on a Surface keyboard.

Microsoft says its haptic keyboard on the Surface Laptop Ultra is its largest ever. That might be; haptic keyboards are clickable throughout the breadth of its surface area, and this follows suit. To me, a haptic keyboard adds a bit of luxe to the whole experience, and for this price tag you’d better believe I expected one. Gestures worked as advertised, with no issues.

Surface Laptop Ultra: Webcam

I was not impressed with the Surface Laptop Ultra’s webcam, and I’m shocked by that. The webcam is just an ordinary 1080p webcam, and not the superior webcams that helped jumpstart the webcam renaissance during the pandemic. While the color balance was fine, the fuzzy 1080p (am I too used to testing top-notch 4K webcams for our recommendations?) left me feeling like Microsoft cut a corner here — and for the price, I don’t expect many corners to be cut.

Mark Hachman / Foundry

No NPU also means no Windows Studio Effects, the filtering and zooming and panning effects that are built into Copilot+ PCs. It’s just an ordinary webcam, and that’s not something Surface fans expect.

Surface Laptop Ultra: Performance

My experience with the Surface Laptop Ultra began poorly. Apps simply crashed or refused to load, or loaded so slowly that I could literally see the display painting the screen like AOL’s dialup service during the 1990s, as was the case with 3DMark. My initial battery rundown test uses Windows Media Player, with a video file looped over and over until the battery expires. Media Player, however, froze when loading my test, and froze again during the run, on two occasions.

I notified Microsoft, and they believed that I had been accidentally sent a defective unit. That might be the case. But Microsoft snuck in a security update after I began testing it. By the time I started intensively testing the laptop, those issues had gone away. I can’t say for certain what happened. Because of my limited testing time, I chose not to run our standard battery rundown test, but I replaced it with what I feel are more accurate tests, anyway.

(Microsoft sent a replacement Surface Laptop Ultra — with 48GB of RAM — which I didn’t have a chance to test before this review published. The initial issue I had was with stability, not performance, so I think my results are valid.)

Nevertheless, there are several caveats worth noting.

First, the Nvidia N1X is an Arm CPU, so certain applications won’t run natively. (Those apps use Windows’ own Prism translator, sometimes robbing them of a bit of performance.) But the CPU wasn’t the issue, the GPU was. Some applications (Adobe Premiere, for example) refused to run because of an unknown GPU. However, a beta version of Premiere accommodated the new CPU architecture.

Microsoft also released the Surface Laptop Ultra as it normally does, by dialing down the Windows power slider into its most efficient mode. On the other hand, Microsoft and Nvidia also claim that, like Qualcomm’s Snapdragon, Nvidia’s N1X should run as close to the performance of wall power when on battery.

Mark Hachman / Foundry

That wasn’t always the case! Where I could, I ran each test on the default power settings, as well as with the Windows power slider cranked to the max. Finally, I tested on battery as well, again on the default efficiency settings. You’ll see these reflected in the charts below.

By default, the single-core CPU performance of the N1X isn’t great, and the Laptop Ultra felt laggy and slow in places, at least at the beginning. Pushing up the performance increased the single-core clock speed, making the laptop feel faster and more responsive. If you buy the Surface Laptop Ultra, dial up the performance using the power slider. It makes a significant difference.

I’ll give you another hint at my findings down below: This is an Arm machine, and Arm chips typically run cooler and more efficiently. I measured the power draw of the Surface Laptop Ultra versus several gaming laptops. I think a key selling point here is power efficiency.

Surface Laptop Ultra: Productivity tests

I chose to break down the performance in two ways: First, as a general productivity machine, and second, as a challenger for gaming laptops — though not so much for gaming per se, but for content creation and AI. I do include one gaming benchmark for comparison, but my colleague Michael Crider is preparing his own review with an eye toward gaming, so stay tuned for that.

Mark Hachman / Foundry

Because these are Arm laptops, some of our tests have to be swapped out for Arm-compatible versions. That includes replacing the PCMark 10 benchmark, which tests laptops against a variety of open-source applications, for Procyon’s productivity benchmark, which uses the three basic Microsoft 365 apps — Word, Powerpoint, and Excel — and adds Outlook to the mix. I guess it’s good that Microsoft is targeting AI and graphics, as opposed to productivity, because this clearly is not a good start. Well, unless you dial up the performance.

Mark Hachman / Foundry

The two synthetic CPU benchmarks restore the luster to the Surface Laptop Ultra and its N1X RTX Spark chip, with one exception.

I always point to the single-core performance as one response to the question, why does my laptop feel slow? Single-core performance and the close ties between the OS and the silicon is one of the key reasons that Macs feel quick. Microsoft may enjoy close ties to chipmakers, but the default single-core performance here is disappointing.

(For comparison, the base Apple M4 scores 175 with its single-core performance, and the Apple M5 scores 198, according to CPU Monkey.)

Remember, Nvidia’s N1X combines an Arm CPU (10 performance X-925 cores, 10 slower A-725 cores), which debuted as part of Nvidia’s DGX Spark workstation. That means that this chip platform has been around for about 21 months, and that’s not necessarily a good thing — “it’s old, which probably explains it,” as one analyst privately described the slow single-core performance.

Mark Hachman / Foundry

I’ve used the older Geekbench 6 scores for comparison’s sake, though the newer Geekbench 7 takes advantage of some of the RTX Spark’s capabilities. For reference, the Geekbench 7 single-core score is 2,008, while the multi-core score was 20,093, jumping to 2,687 and 23,256 with the Windows power slider set to “Best performance.” The GPU score is 84,107 on default settings, 140,100 with CUDA enabled, and 116,344 on default settings with the power slider maximized.

On battery, the Geekbench 7 CPU score was 1,899 (single-core) and 17,328 (multi-core), with a GPU score of 83,861 (OpenCL) and 138,212 (CUDA). If you’re a general consumer, you can ignore this; that information is just for productivity enthusiasts.

In the chart above, I tested single-core, multicore, and graphics performance on wall power and on battery on the older Geekbench 6.

Does the Surface Laptop slow down over prolonged use? While it doesn’t howl in protest, it does indeed falter. Under Cinebench 2024’s stress test, performance dropped by 6.3 percent after 10 minutes of use. It also failed 3DMark’s stress test, as graphics performance fell 6.8 percent over a series of looped testing runs.

Mark Hachman / Foundry

We use the Handbrake transcoding test as another way to test the laptop’s performance over prolonged periods. Since we’re measuring the time to complete a task, less is more. Note that this isn’t an Arm-specific test, though my testing in the past didn’t find a real advantage with versions coded specifically for Arm processors.

Mark Hachman / Foundry

When I can, I like to point to real-world applications as evidence of actual performance. But 3DMark’s venerable (too venerable?) benchmarks provide our first hints of how the graphics core will shape up, with more tests further down in this article.

Note that the Time Spy test is written for x86 processors; the Night Raid test was coded for Arm. I include both. I was very impressed with Intel’s Panther Lake chip (the X7 358H, below) in terms of graphics…but look how the N1X shapes up!

Surface Laptop Ultra: Battery life

Because of the early glitchiness associated with the Surface Laptop Ultra laptop, I wasn’t able to record a rundown test where I measure the laptop’s battery life by looping a 4K video over and over at a fixed brightness level. As video playback logic has been assimilated into various parts of the laptop silicon, video playback is more of an afterthought than ever. A video rundown test is still useful, since you’re perfectly welcome to copy or download a 4K video on to the laptop.

A more realistic way of measuring battery life is either streaming or trying to simulate a day’s work. I can do both. While Netflix appears to have implemented checks to prevent hours of streaming at a time, YouTube offers some 4K videos that last up to 24 hours. No laptop exceeds that, yet. The Surface Laptop Ultra lasted 7 hours, 57 minutes while streaming 4K video.

To simulate actual work, there’s the UL Procyon Office Productivity test, the longer version of the Office Productivity test I used earlier. Here, the laptop simply cycles through simulated tasks in Excel, PowerPoint, and other office apps until the battery runs out of power. The Surface Laptop Ultra lasted 10 hours and 3 minutes.

Note that the 95Wh battery in my review unit somehow charged to more than the design capacity, 89Wh. A battery of 99Wh is the most allowed on U.S. aircraft.

Surface Laptop Ultra: Content creation and gaming

Let’s switch gears. Until now, we’ve treated the Surface Laptop Ultra as just another Arm laptop, capable of handling typical office tasks. But that RTX GPU is just asking to be put to work.

The question is, what do we compare it to? Historically, devices like the Microsoft Surface Laptop Studio 2 were considered creator-class laptops, with a powerful CPU but a modest GPU aimed at video editing and light gaming. For a more apt comparison, I requisitioned some existing, pure gaming laptops. Here’s my question: Do you need to buy a monstrously pricey Surface Laptop Ultra when monstrously (?) pricey gaming laptops already exist?

My candidates included an as-yet unreviewed $1,899 updated version of the Acer Nitro V, with an AMD Ryzen AI 7 450, paired with an RTX 5060 creator-class GPU; an $3,799 MSI Raider, with an Intel Core i9-14900HX, and a last-generation RTX 4090; and a modern monster, $4,299 Asus ROG Strix gaming laptop equipped with a Core Ultra 9 275HX and an RTX 5090.

My tests included a Cinebench run to establish overall performance, then runs of Adobe Photoshop and Premiere to assess the laptops’ comparable creation chops. I made a brief segue into gaming with Cyberpunk: 2077, then dove into AI: Procyon’s text-generation benchmark measures large language-model performance. Finally, I crushed each laptop with a text-to-video application, governed by ComfyUI, with some on-device local LLM testing. Here, I didn’t fool around with battery mode; I tested with default settings but also at maximum power.

Here’s each laptop, as evaluated by Cinebench 2024 for total CPU performance. (I used the Arm version of the test for the Surface Laptop Ultra.) The Acer Nitro glitched and wouldn’t run this test. It’s impressively close.

Mark Hachman / Foundry

I then used PugetBench’s Photoshop and Premiere tests as a gauge to how the Surface Laptop Ultra fares as a content-creation machine. I was shocked by the first result.

At base levels, the Laptop Ultra finishes last. With performance dialed up, it’s…first? The PugetBench Photoshop test is finicky, so I only checked this result once. Still, the Photoshop test completes three runs before settling on a final score, so I suspect that it’s accurate.

Mark Hachman / Foundry

I didn’t see nearly as much improvement in the Premiere test, which exports several video files. I’m not sure why. Adobe will have new versions of its Creative Cloud applications to use with the RTX Spark platform, but Adobe hasn’t said when they’ll arrive. I used a beta version of Premiere for this test, which Adobe supplies as part of its Creative Cloud beta program.

Nevertheless: A solid showing here by Microsoft’s newest laptop.

Mark Hachman / Foundry

As mentioned above, my colleague Michael Crider is testing his own RTX Spark sample laptop with an eye towards gaming. Again, the RTX Spark platform is an Arm chip, which may have compatibility issues…or not. I present the Cyberpunk: 2077 test below as a representative sample, but expect more in terms of game testing soon.

I ran several tests to compare the gaming laptops to the Surface Laptop Ultra, and settled on using 1080p resolution at Ultra settings, a decent compromise between efficiency and quality. Initially, I turned off frame generation.

Mark Hachman / Foundry

But this is an AI laptop, and if there’s anything “AI”-related that Nvidia is known for, it’s DLSS AI frame generation. With those options enabled, frame rates jumped to 104 frames per second. And dialed up to maximum power with DLSS and frame generation on, the laptop ran that game at 130 frames per second — more than double the base rate.

I paused my testing for another angle: power. Even though the United States is blissfully headed into cooler temperatures, power bills keep rising. One of Arm’s arguments is power efficiency, so I hooked up a power meter to the laptops and reran the benchmark again.

Even though I didn’t measure prolonged power consumption, the differences between all four laptop platforms were immediately obvious:

• The Acer Nitro V laptop held steady at between 115 to 117 watts, touching 121W at peak.
• The MSI Raider blasted through 290 to 305 watts, hitting a peak of 335W max.
• The Asus ROG Strix fluctuated the most, consuming 261 to 290 watts and peaking at about 310 watts.
• As for the Surface Laptop Ultra? It didn’t consume the least amount of power, but 133 to 144 watts throughout the run holds its own. (I didn’t test the power output with DLSS on, which should increase efficiency.)

The point is that more power equals more performance, but more performance equals more power…and possibly a higher total cost of ownership.

Finally, I wanted to know if Microsoft expects people to work on an airplane, churning away at offline AI agents, with fans howling — and will the seat actually provide enough power? My research says no, but I can’t prove it definitively. (Microsoft said they’d get back to me.)

Surface Laptop Ultra: AI tests

Microsoft and Nvidia argued that the RTX Spark platform should usher in a new era of local AI, with local LLMs and agents busily chewing up compute cycles. That’s true, to a point. The argument in favor of that now is that local AI equals free tokens — cold hard cash is now the selling point, versus the privacy-first approach from a couple of years ago.

But there are still catches. Acer shipped its new Nitro V laptop with just 512GB of SSD storage on it, and I found myself spending my own human “compute cycles” just trying to find space on it for what I wanted to run. Photoshop, Premiere, text-to-video models…it all adds up. If you don’t have at least a terabyte of storage, you’re going to feel the pinch.

And that’s before you begin trying out AI models. Local AI requires local memory to run in. Dedicated gaming laptops with discrete GPUs have their own fixed pools of video RAM, which AI can tap for executing generative AI. Agents, however, lean more heavily on the CPU. Laptops like the Surface Laptop Ultra split memory between the GPU and CPU, and my review laptop came with a small utility to let Windows handle the division of labor. (To date, users have been able to manually configure the split between VRAM and system RAM via the BIOS or an app supplied by AMD or Intel.)

UL’s Procyon app already ships with a handy little test to measure how well the laptop will process queries using models with about 8 billion parameters. (“Better” models usually favor more parameters, though that can vary. Models can also be “compressed” via a process called quantization to run locally without taking up as many resources.)

Mark Hachman / Foundry

The performance via model varies, of course, but the overarching lesson is clear: The gaming laptops generated more tokens. (A token is a few letters, so that the word “few” might be a single token, and the word “somewhere” might be three.) An LLM spits out a response like an old dot-matrix printer, so reading that output is subjective. Using the Mistral 7B model, the Surface Laptop Ultra spat out about 52 tokens per second, about 13 words per second.

That’s not the only way tokens are processed, of course. A “token” can be part of a chart, an image, a video, or more. The most recent use of generative AI is AI-generated video, which has become rather scarily realistic. Is AI slop still AI slop when produced on your own local hardware? Regardless of how you land on the subject, text-to-video generative AI absolutely murders a laptop.

ComfyUI contains its own in-app benchmark to measure this. The Surface Laptop Ultra performs surprisingly well. (I ran it on its maximum performance settings by default.) It’s not noted in the cart below, but I used the WAN model to generate the output.

This benchmark doesn’t measure tokens, but the time to finish up a single run, supposedly generating several seconds of video. A lower score is therefore better. (ComfyUI too, now runs native on the RTX Spark platform.)

Mark Hachman / Foundry

I also tested token output via LM Studio, using the Qwen 3.8-27b model as an example. LM Studio’s settings allow you to only choose from models which will run on your PC, which meant that I was prevented from using some of the more powerful options. Would that make a difference? I honestly have no idea. Still, I could only run what I could run!

Using a consistent reasoning budget, zero temperature (for, ideally, fixed responses), and a fixed outcome length, I asked for a science-fiction essay using the same prompt for the two high-end gaming machines, plus the Surface Laptop Ultra. After a trial run, I averaged the token output over three times.

• Surface Laptop Ultra: 39.9 tokens/s
• MSI Raider: 15.2 tokens/s
• Asus ROG Strix: 61.9 tokens/s

I then ran a fourth test, working on a web browser in the background, while trying to test whether working with an LLM in the background would affect its productivity. It didn’t seem to, as I noticed no difference. I didn’t work hard in the background as I normally would, with Excel and Outlook and a web browser all competing for compute cycles. I have to believe that will affect it, but that will have to wait for further testing. Keep in mind that when you do that, you’ll be working while the laptop’s fans are cranked up high.

Nvidia debuted a new Nemotron 3.5 Bolt 74B-A9B 24GB model for RTX Spark laptops as I finished up this review, but I didn’t have a chance to try it.

Finally, I did try running NousResearch’s Hermes, one of the recommended AI agents which you can access via the Microsoft Store, on the Surface Laptop Ultra. I’m still somewhat reluctant to give it control of my PC (yet) but it did load on my PC. I found it interesting, though, that Windows assigned 18.9GB of the available memory to itself — and while it used every bit, the GPU was barely touched during the agent’s operation.

Still, I was right — memory is going to be at a premium in the AI age, and you’re going to want more, more, more.

Should you buy the Surface Laptop Ultra?

I wrestled with this, naturally. The version of the Surface Laptop Ultra we were offered for review isn’t a low-end machine, obviously. Nor is it the minimum Microsoft offers. It feels, though, that we’re exploring the “low end” of the Surface Laptop Ultra lineup for an insane $3,700. Whew.

We’re not done yet, though. PCWorld’s review strategy has been to examine the RTX Spark platform through different eyes: First mine, focused on productivity, then separate articles examining it from a gaming standpoint as well as from an AI perspective. And yes, we have a $7,000 RTX Spark laptop in house.

And I think that’s the crux of the matter. As a pure productivity machine, you can find better options — and, if you’re eager to use AI, you can turn to a cloud model. That’s sort of what executives were privately pushing at the RTX Spark launch: to run a local “harness” on 32GB laptops like the Surface Laptop Ultra, and tap the cloud instead.

And yet! This is a $3,699 PC that’s far lighter and more compact than the gaming PCs I compared it to, consumes far less power, and performs somewhat near the other, bulkier machines. You can certainly make an argument that it’s well out of reach of the ordinary consumer, yet worth it as a (sorry) “budget” AI workstation.

I still think, however, that there’s going to be FOMO if a new AI model comes out and this version of the Surface Laptop Ultra can’t run it. So be careful. That way lieth dragons.