Tested: Intel’s Lunar Lake wants you to forget Qualcomm laptops exist

Lunar Lake is Intel’s Snapdragon killer.

Intel’s Core Ultra Series 2 (Lunar Lake) was specifically designed to emphasize low power, but with competitive performance. In this it somewhat succeeds, though the Core Ultra 7 258V chip I tested can still run a distant second, or third, behind AMD’s mobile Ryzen processors. But Lunar Lake also provides incredibly good, Snapdragon-like battery life with a powerful, embedded GPU capable of playing yesterday’s top-tier games.

Intel supplied us with a Lunar Lake-powered Asus ZenBook S14 laptop for review, and we’ve spent the last week or so testing it to answer the question: Of the AMD Ryzen AI 300, Intel’s Lunar Lake, and the Qualcomm Snapdragon X Elite, which is the best laptop processor so far in 2024? And how does Lunar Lake compare to its predecessor, Meteor Lake?

We’ve run all three chips through a number of benchmarks — with separate tests run while plugged in and on battery power. We think it’s important that you know the maximum performance potential of these laptop processors as well as how much performance is lost to keep your laptop up and running when unplugged.

Intel launched Lunar Lake at the Computex 2024 show in Taiwan, where executives promised a “no-compromise experience.” The Asus ZenBook S14 and other Lunar Lake and Ryzen AI 300 laptops now qualify as Copilot+ PCs, even though a Windows update to enable their AI features won’t occur until October.

How we tested Intel’s Lunar Lake

This review uses benchmarks compiled by myself, my colleague Gordon Mah Ung, and other PCWorld laptop reviewers over current and previous testing. I’ll focus on the three key chips of this generation: Intel’s Core Ultra Series 2/Lunar Lake, the Qualcomm Snapdragon X Elite, and the AMD Ryzen AI 300, occasionally adding some older processors for additional context, such as a Core Ultra Series 1/Meteor Lake processor.

I’ll try to explain what each test covers, why it matters, and how test performance will reflect on the next laptop you buy. If you want even more depth, check out Gordon’s Lunar Lake video (below) that runs through extensive benchmarking and testing:

The laptop that Intel and Asus supplied includes a Core Ultra 7 258V. That’s about midway down the stack of the Core Ultra Series 2 chips Intel launched, and it’s a little unusual in that Intel isn’t handing us its top-of-the-line chip for comparison testing. Instead, the Ultra 7 258V includes four performance cores, four low-power efficiency cores, and a 4.8GHz turbo clock available when needed. All told, just eight threads are available for processing.

The nice thing is that Intel and Asus have consistently used similar laptop models for comparison between chip generations, specifically the Asus ZenBook 14 OLED, which is home to both the Lunar Lake and previous-gen Meteor Lake Core Ultra processors below. AMD used the Asus Zenbook S 16 as a debut platform for the Ryzen AI 300. On the Qualcomm side, though I reviewed the Microsoft Surface Pro 11th Edition, I commandeered the Surface Laptop 7th Edition Gordon tested and ran some additional benchmarks. It’s a fairer comparison for Qualcomm. In certain benchmarks, because of logistics, I used an MSI Prestige 16 AI Evo with an identical Meteor Lake chip inside.

Lunar Lake tested: CPU performance

I tested all of our test devices with two synthetic benchmarks: Cinebench R23 and Cinebench 2024. Both push the laptop’s CPU to its limits, first using every processing thread available and alternatively with just a single thread. The difference between R23 and and Cinebench 2024 is that the latter has a specialized Arm version, while the R23 version is run via the Snapdragon’s emulator software that translates X86 instructions into Arm code.

Two things stand out here: When you code for Arm, Snapdragon’s performance improves significantly, propelling it to the top of the heap. That pushes Lunar Lake (the Ultra 7 258V) down the rankings. But you’ll also note that Intel’s older Meteor Lake (the 155H) outperforms its latest chip. What this tells me is that Intel is prioritizing battery life, not performance.

According to our sister site, Macworld, the MacBook Pro and MacBook Air 15, both using Apple’s M3 chip, reported multithreaded scores of 693 and 658, respectively. In single-core performance, they reported a score of 141 for both platforms. So yes, Intel’s Lunar Lake is faster, but not in single-core performance.

One note: I test using the default performance settings that laptop makers provide, as that feels realistic. That means that the Snapdragon tests were performed at Microsoft’s “Recommended” settings (the lowest) while Lunar Lake’s tests were performed at the “Balanced,” or middle settings. However, I also test chips in performance mode; in this case, using the Windows settings for maximum performance and adjusting the Asus laptop settings to its performance settings, too. Occasionally you’ll see battery performance scores that outperform wall power; that can be due to statistical anomalies, but it can also just happen. We’ve seen this in our Snapdragon and Meteor Lake testing, too.

If you’d like Snapdragon’s performance settings for comparison, my Surface Pro (11th Edition) review includes those. (I don’t test “Max” settings on battery.)

I’m not a huge fan of the Geekbench suite, only because it runs in the background without demonstrating what it does as a real workload. But it’s a popular test, as it’s quick to run. It too has a dedicated Arm version, which propels the Surface Laptop 7th Edition to the top. I’ve organized the results hierarchically by the “everything goes” multithreaded CPU test, placing the powered test directly over the test running on battery for comparison’s sake.

One of Snapdragon’s strengths is how uniformly it runs, whether on battery or on wall power. Geekbench again shows how sharply Lunar Lake’s performance drops on battery…but also hints at something we’ll get to later: the excellent performance that its integrated GPU delivers. The other story is simply how well AMD’s Ryzen continues to deliver top-tier performance.

Another choice I made was to use Procyon’s Office Productivity benchmark for this round of testing. I prefer PCMark 10’s broad suite of apps (from office to CAD to simulated video calls) but Arm processors won’t consistently run all of the tests. These office benchmark tests are still legitimate, demonstrating how well the laptop chips work within Microsoft Office, including Outlook, using the real-world apps themselves.

Fortunately, or not, there’s not much difference between the various platforms. But again, look at that drop-off on the Core Ultra/Lunar Lake 258V when you run it on battery!

This becomes a real issue in certain situations.

One of the realities of working on the road is that you rarely “work” constantly: You browse to a web page, pause while you read it, author an email, take a sip of coffee, check your phone… your laptop is often sitting idle. But when you actually ask it to execute an application, the latest generation of mobile processors can have a hard time keeping up.

In our recent tests of the AMD Ryzen AI 300, we adjusted our Handbrake test to make it more intensive. Handbrake transcodes a video — in this case, the open-source film Tears of Steel — and it’s a computationally intensive process. How well the laptop cools itself plays a role here, but so does the battery: This test can take close to an hour to perform, and even a robust battery can give out. I couldn’t loop three of these tests together before the battery expired, which is about three hours of constant work.

Again, though, Intel’s latest and greatest…finishes last. This is a real-world application that you can download yourself, so it’s a meaningful test. It took Lunar Lake nearly 17 minutes more than the Ryzen to complete the task.

Lunar Lake tested: Graphics and AI performance

So far, Intel’s Lunar Lake hasn’t been that impressive compared to the competition. But we now turn to the integrated GPU, and the new Xe2 graphics engine that Intel says will debut in a second discrete GPU, code-named Battlemage, later.

Given AMD’s historical strength in GPUs, Intel’s resurgence is a bit surprising. But in two tests — Time Spy, an older benchmark geared toward desktop and mobile gaming, and its replacement, Steel Nomad Light — Intel’s Lunar Lake shines.

Can you edit photos on the go? Puget Systems, a workstation developer, authors its own benchmarks to try to answer that question. The test uses a scripted benchmark to apply various filters and adjustments to images opened within Adobe Photoshop, and then generates a score.

Here, all three mobile platforms are bunched closely together, which probably indicates that all three are okay for this task…but not on battery. Here’s some additional context: According to the PugetBench site, an AMD Ryzen 7 7800X3D desktop system with a GeForce RTX 3080 generates a score of 9,959. A massive MSI Raider notebook with an Intel Core i9-14900HX and GeForce RTX 4090 laptop GPU generates a score of 9,959. And an Intel Core i7-9750H desktop paired with a GeForce GTX 1660 Ti produces a score of 4,549.

We normally don’t run gaming benchmarks on top of what are usually considered to be productivity machines. But in this case Intel has been telling us that, yes, these machines are good enough for light gaming, especially on Low graphics settings. That’s good enough for us.

Typically, 60 frames per second is considered the minimum for gameplay — any less, and the game can stutter. In this case, we tried out Shadow of the Tomb Raider, run at 1080p and Low settings. Any you know what? It looks and runs pretty dang well. All of our gaming tests were performed on wall power.

After its disastrous (?) launch, Cyberpunk: 2077 has surged back to become one of the more playable games on PCs. And in this case, it runs pretty well on Lunar Lake. However, AMD’s Ryzen still outpaces it. Again, these tests were run at 1080p, at Low settings. Frame generation was turned on for the Ryzen processor; that’s not offered for the other two chips.

Instead, Cyberpunk offers XeSS resolution scaling, which, combined with dialing up the Windows and laptop settings to performance mode, allows for a playable frame rate of about 65fps, though I saw 69fps on a cold run at the beginning of the day.

Finally, we move to AI. AI has become one of the hottest topics in computing, although developers still haven’t flooded the market with AI applications that run locally on the PC. (Windows 11’s upcoming 24H2 update should have more, including generative art for Paint and Photos, and eventually Windows Recall.)

For now, there aren’t too many benchmarks that tap the TOPS that these new platforms integrate. Remember, Intel’s Lunar Lake, AMD’s Ryzen AI 300 (up to 50 TOPS), and Qualcomm’s Snapdragon X Elite chips (45 TOPS) are all Copilot+ PC-qualified, even if they won’t receive their new feature set without a Microsoft update. Intel’s Lunar Lake chips provide a varying amount of TOPS; in this case, the Core Ultra 258V provides 47 TOPS.

UL’s Procyon app measures the AI output by testing each chip for how well it executes various AI frameworks. Like the Cinebench or 3DMark tests, the idea is that you’re getting a generic picture of how well the chips will perform AI functions.

While I don’t have a Procyon result for the AMD Ryzen chip, the test shows that both Lunar Lake and the Snapdragon X Elite are pretty evenly matched; Intel’s chip does better with AI on wall power, but Qualcomm is the superior AI chip on the go.

Procyon also offers a more practical test: generative AI image generation, which uses AI to generate a series of eight images. (This test doesn’t yet run on an Arm processor.) I ran two tests, just to indicate the practical power of the NPU. First I used the test to create a series of eight images using just the NPU, and then I ran the same test with a toggle that allowed me to use the integrated GPU as an assist. It’s this “total TOPS” that companies like Intel are espousing as the future of AI, though one that really isn’t here yet, either.

The test uses an 8-bit integer model, which generates less complex images than a floating-point model. But they’re quick: about 8 seconds per image when the NPU and iGPU are combined. Here, I compare the Procyon scores.

Lunar Lake tested: Battery-life performance

It’s easy to measure the battery life of a laptop. For a chip? Well, that’s a lot messier. How long your laptop will last depends a lot on the size of the battery and how big and how efficient your laptop’s screen is. The type of screen (IPS, OLED) also makes a difference.

I think the best way to answer this particular question is to refer you to my colleague Gordon Mah Ung’s forthcoming video, in which he takes two virtually identical Dell laptops — one with a Qualcomm Snapdragon X Elite chip inside, the other with an Intel Core Ultra Series 2 / Lunar Lake chip — and does a direct comparison. That’s about as good as you’re going to get in answering which chip is most efficient.

We do have previous test results of the Asus ZenBook S 16 (with a Ryzen AI 300 chip inside and a 78Wh battery), the 13.8-inch Surface Laptop 7th Edition (with a Snapdragon X Elite chip and a 54Wh battery), the Asus ZenBook 14 OLED (with a first-gen Core Ultra Core 155H and a 75Wh battery) versus the current Asus ZenBook OLED (with its Lunar Lake chip and a 73Wh battery).

Running these laptops on our standard video rundown test, where we loop a 4K video over and over at a fixed screen brightness until the laptop’s battery expires, is kind of ludicrous these days. The Lunar Lake laptop lasted 21 hours!! That’s basically like taking a laptop on a weekend “camping” trip or a prolonged sales conference.

It makes more sense to ask it to do a little work to simulate a work environment. UL’s Procyon battery-life test moves from Office app to Office app, performing “tasks” in each to simulate you working on your laptop for hours on end.

In this case, here’s how it worked out. Remember, Meteor Lake, Lunar Lake, and the Ryzen AI tests were run in “balanced” mode, while the Snapdragon X Elite was left in its “recommended” low-power mode. As they say, your mileage will vary…but at least 10 hours of computing power from all three competitors? That’s fantastic.

• Intel Lunar Lake: 17 hours, 7 minutes
• Qualcomm Snapdragon X Elite: 16 hours, 20 minutes
• AMD Ryzen AI 300: 10 hours, 42 minutes
• Intel Meteor Lake: 10 hours, 35 minutes

Conclusion: Which laptop chip is better?

Old chip nerds will recall Transmeta, a laptop processor startup that could translate its instruction set from X86, much like Arm. Transmeta’s legacy isn’t so much what it accomplished, but what it drove Intel to do: finally zero in on the laptop market and develop a competitive laptop processor during the first years of the new century. Now, it’s happened again. Intel’s first Core Ultra, Meteor Lake, downplayed performance for long battery life; Lunar Lake takes that to the next level.

So is Intel’s Lunar Lake a Snapdragon killer? Not exactly. In terms of efficiency, there’s a strong case to be made that Qualcomm’s Snapdragon X Elite is still the most efficient processor out there. AMD’s Ryzen chips win at many performance tasks, but battery life suffers.

But in some sense, that doesn’t matter. Dumping in a few extra grams of battery essentially “eliminates” Qualcomm’s advantage, as our test Asus laptop demonstrates. But the Snapdragon still outperforms it under several scenarios, save for gaming. There, lingering compatibility issues associated with the Arm instruction set and the impressive 3D engine give the Core Ultra Series 2 a solid advantage. All apps just work on Intel and AMD machines, but can be hit or miss on Snapdragon laptops, though Qualcomm has managed to get many of the most popular apps running on Arm.

The bottom line is that for any task that depends on the CPU, Intel’s Lunar Lake is middling at best. When the GPU or NPU enters the equation, it surges higher. Ditto for battery life, where it competes well against the Snapdragon. (At a certain point, do people stop worrying about battery life, or does long battery life equal a longer-life notebook?)

A few years back it was easier to identify the “best” laptop chip. Now, the question is what’s best for you. And it’s not completely answerable, either: With Arrow Lake lurking in the background, Intel could have a separate performance processor up its sleeve. For Intel, this is a simple sell: Lunar Lake is a safe bet, a Snapdragon without the compatibility concerns.