Intel’s Core Ultra Series 3 platform, Panther Lake, and its champion, the Core Ultra X9 388H microprocessor, provide one thing distinctive: highly effective, gaming-class 3D efficiency with battery life that’s virtually unparalleled within the laptop computer area.
Intel positioned the Core Ultra Series 3 (Panther Lake) as a chip with the computational energy of its Arrow Lake platform, with the low energy consumption of the Core Ultra Series 2 (Lunar Lake). The chip maker additionally predicted that Panther Lake’s gaming efficiency is roughly equal to a laptop computer with an Nvidia GeForce 4050 laptop computer chip inside it. As I’ll present you, these are comparatively honest claims.
Instead of only a battery of exams, we’ll attempt to pull out the “story” of Panther Lake, demonstrating its strengths and weaknesses as we go. Let’s just hope you can buy one.
Mark Hachman / Foundry
Intel’s Core Ultra Series 3 chip stands alone, for now
In October, Intel originally positioned the Core Ultra Series 3 (Panther Lake) in one among three fundamental configurations, combining the brand new “Cougar Cove” P-core and the “Darkmont” E-core and Low Power E-cores. At the excessive finish was what Intel referred to then because the “16 core 12Xe” configuration, with 4 P-cores, 8 E-cores, and 4 LP E-cores, 12 Xe3 GPU cores, and 12 ray-tracing models. When it got here time for Intel to announce the Panther Lake chip lineup, that configuration included its formal identify, the Core Ultra X9 388H, with the “X9” prefix added to focus on the biggest Xe3 configuration.
Intel then let reviewers benchmark the Core Ultra Series 3 chip during CES, however solely utilizing video games. It was our first indication that Panther Lake might be one thing particular.
Intel prevented reviewers from testing CPU-specific benchmarks, nonetheless, most likely as a result of the variety of cores contained in the highest-end Panther Lake chip (16) are lower than these contained in the rival Qualcomm Snapdragon X2 Elite Extreme chip (18), which means that Intel would probably lose to the Snapdragon on paper in multi-threaded CPU-specific benchmarks.
The third contender would be the AMD Ryzen AI 400, an improve to the wonderful Ryzen AI 300, which AMD debuted final yr. The Ryzen AI 400 consists of simply 12 cores, however runs them at a most clock pace of 5.2GHz — the quickest pace of all three chips. But laptops with both the Ryzen AI 400 or Snapdragon X2 Elite aren’t but accessible.

Mark Hachman / Foundry
CPUs don’t suck any extra
Both Intel’s Core Ultra Series 1 (Meteor Lake) and Core Ultra 2 (Lunar Lake) had been surprisingly common in CPU efficiency, each in single-core and multi-core duties. (CPU-specific purposes embody net searching, apps like Excel, compiling software program, some video games, and decompressing recordsdata.) AMD’s Ryzen and Qualcomm’s Snapdragon used to carry the benefit. No longer — properly, not less than in contrast in opposition to its older rivals.
With Panther Lake, Intel has regained its management in CPU computations.
Intel supplied us an Asus ZenBook Duo (UX8407A) with an Intel Core Ultra X9 388H chip inside as a launch laptop computer for the Panther Lake platform. It was a barely odd alternative; the ZenBook Duo is a dual-screen laptop computer, with a gaming-class 99 watt-hour battery, which drastically inflated the battery life.
Intel additionally supplied a prototype Lenovo laptop computer, which we used as a actuality test for the estimated battery life and extra benchmarks. I left the ZenBook Duo in “clamshell” mode, solely utilizing one among its 2K screens to render information to provide outcomes I felt assured in evaluating to different platforms. I began utilizing the Cinebench 2024 and Geekbench artificial CPU exams.

Mark Hachman / Foundry
To deal with Intel’s declare that Panther Lake provides the CPU efficiency of the cell Arrow Lake chip: sure, that’s true. Intel launched the Core 285H chip final yr, and in our review of the Core 285H, I discovered that the Cinebench 2024 rating was 1,012 (multithreaded) and 128 (single-threaded), only a hair underneath Panther Lake’s efficiency. In Geekbench (measured beneath), the older 285H produced a rating of 16,755, once more barely lower than Panther Lake’s Core Ultra X9 388H.
But in case you’re a Windows fanatic like we’re, you is perhaps upset by the Core Ultra X9 388H’s exhibiting. Referring to the review of the Apple M5 MacBook Pro, our colleagues at Macworld report that the MacBook M4 Pro reported a rating of 1,010 in Cinebench 2024 and 14,763 in Geekbench 6. But the MacBook M5 Pro scored 1,126 in Cinebench 2024 and 18,013 in Geekbench 6, besting Intel’s present mainstream laptop computer chip.

Mark Hachman / Foundry
Perhaps an upcoming HX model can do higher?
Battery life is huge! But so is the battery
Intel has claimed that Core Ultra 3 laptops may have as much as 27 hours of battery life. That’s true — however, as is commonly the case, it relies upon. The two screens of the ZenBook Duo suck extra energy than a single show. However, Asus put in a 99Wh battery inside. That’s a gaming-class battery, and the biggest capability allowed on a aircraft by FAA guidelines. In this case, it’s like bolting a self-powered gasoline truck to a sedan.
So sure, the battery life was insane: about 22 hours on the ZenBook Duo working on a single display screen and 25 to 28 hours (1,704 minutes) on the Lenovo prototype laptop computer that we utilized in early exams at CES. Those exams had been carried out by looping a 4K video till the battery expired. When requested to do a bit extra work (simulating workplace work through the Procyon Office benchmark) battery life dropped to “just” underneath 14 hours on the ZenBook Duo — nonetheless mainly one of the best outcomes we’ve ever seen. We’ll break down the battery life a bit extra on the Asus ZenBook Duo in our upcoming, devoted evaluate.

Foundry / Mark Hachman
Again, Intel needs us to consider that the Core Ultra has the efficiency of its “Arrow Lake” chips with the ability draw of its Core Ultra Series 2 (Lunar Lake) chips. We can test that, form of, by monitoring the ability consumption of a Lunar Lake and a Panther Lake pocket book as they bear a benchmark. In idle, the Core Ultra 3 chip attracts about 5 watts, however can drop down underneath a watt. Lunar Lake averages about three watts or much less in idle.
It’s not apples to apples, although. Intel used TSMC’s N3 course of expertise to fabricate the CPU tile within the Series 2 Lunar Lake chip, whereas Panther Lake makes use of Intel 18A, with some tiles break up between the 2 corporations. In this case, Intel’s older Lunar Lake is a 17W TDP chip, whereas Panther Lake is 25W — extra energy to the chip usually means higher efficiency and worse battery life, however the bigger battery and Intel’s structure appear to offset this.
Here’s an influence graph exhibiting the 2 chips in idle, then working a benchmark, then dropping down into idle as soon as once more. This graph simply measures the ability going into the CPU bundle, not the complete laptop computer. That energy might fluctuate considerably, and is finest left to the battery-life comparisons you’ll discover in our particular person laptop computer evaluations. Still, Panther Lake is throwing much more energy and efficiency on the benchmark, and this graph demonstrates that if a Lunar Lake and a Panther Lake laptop computer contained the identical battery capability, the older Lunar Lake laptop computer might win.

Mark Hachman / Foundry
Still, if Intel convinces laptop computer makers so as to add bigger batteries to Panther Lake laptops, although, look out. Laptop battery life numbers might explode upwards!
Performance nonetheless drops whereas on battery
One of the attention-grabbing issues about Qualcomm’s Snapdragon chips is that they run at full energy on a regular basis. Intel’s Core Ultra chips don’t, clocking all the way down to decrease energy consumption, extending battery life.
I run all of our benchmarks on wall energy, battery energy, and at Windows’ most allowable settings, simply to see how efficiency varies in several person eventualities. As you’ll discover in our Cinebench 2024 benchmarks, the single-threaded efficiency normally related to OS duties stays unchanged between wall energy and battery energy, maintaining Windows as responsive in each eventualities.
But look right here: Intel’s Core Ultra Series 3 chips appear to keep up their efficiency on battery significantly better than Intel’s Core Ultra Series 2 or Series 1. We’re utilizing three real-world benchmarks to check this. First, right here is Procyon Office, which performs numerous duties in Microsoft Office / 365. Performance drops by about 20 % on battery.

Mark Hachman / Foundry
However, on our customized real-world Handbrake take a look at, the place the laptop computer is requested to transcode the open-source Tears of Steel film, efficiency dropped by simply three % between wall energy and battery.
Here, you may see how our take a look at Panther Lake laptop computer fared in comparison with the competitors. This is a customized take a look at, totally different than the one we run in as a part of our laptop computer evaluations. I additionally made certain to obtain an Arm-specific model of the app, however Qualcomm’s chip fared exceptionally badly right here. It normally performs fairly properly.

Foundry / Mark Hachman
Since we’re real-world benchmarks, we are able to see that Intel’s Core Ultra 300 / Panther Lake fares properly in PugetBench’s Photoshop take a look at. The take a look at makes use of the delivery model of Photoshop. Here, efficiency dropped nearly three % on my exams once I unplugged the laptop computer.
Keep in thoughts that CPU-specific exams are one among Snapdragon’s strengths. And with the Snapdragon X2 Elite generating exceptional CPU performance — preliminary numbers crush Panther Lake, and the Elite X1 nonetheless ranks extremely — this is perhaps an space the place Qualcomm catches up. This race ain’t over.

Mark Hachman / Foundry
Unfortunately, Puget Systems’ PugetBench benchmark hadn’t caught as much as the model of Adobe Premiere Pro (26.0) that Adobe makes accessible for obtain, so I used to be unable to check that software.
Panther Lake’s GPU efficiency is unimaginable
Remember, Intel’s flagship Panther Lake chip is the Core Ultra X9 388H–that is totally different. The Core 9 is now the Core X9, which implies the GPU has 12 Xe3 cores. Essentially, the “X” signifies that you’re getting one of the best Intel has to supply when it comes to graphics.
What does this imply? For years, built-in graphics has been capable of play video games: older, 2D sprite-based video games, and a few older 3D video games at decrease settings. They ran. And that was wonderful. With Panther Lake, we’re navigating a transition into built-in graphics performing virtually in addition to gaming-class discrete graphics — and whenever you add AI upscaling and body era to the combo, current top-tier titles are close to your grasp.
Some avid gamers check with these as “fake” frames, which is why it’s useful to look first at each conventional, non-accelerated exams. Here, we use UL’s 3DMark, particularly the Time Spy and Steel Nomad Lite benchmarks.

Mark Hachman / Foundry
A terrific enhance in gaming efficiency
This was one of many huge tales of CES 2026: Intel’s claims that Panther Lake supplied the ability of a gaming laptop computer with a discrete Nvidia GeForce 4050 GPU, however inside an built-in bundle.
This, for me, was the eye-opening second. A yr or so in the past, I used to be testing Intel’s Core Ultra 1 (Meteor Lake) and Core Ultra 2 (Lunar Lake) with customized runs of video games like Cyberpunk: 2077 at Low settings, which we present beneath.

Mark Hachman / Foundry
But these exams prompted me to “graduate” Panther Lake into our gaming benchmarks, too, with the settings that historically extra highly effective laptops now use. Even utilizing our aggressive gaming settings, a recreation like Shadow of the Tomb Raider reaches playable body charges. (Skip down to seek out these outcomes.) Yes, it completely is an older recreation, relationship from 2018. Yet Shadow was a top-tier AAA title, and built-in graphics has caught up. And that’s simply pure, unadulterated, farm-to-table frames, too.
Don’t get too excited, although. Metro: Exodus was launched in 2019, however its 4A Engine stays out of attain for Panther Lake. On our take a look at laptop computer, the sport averages 24 frames per second when run at 1080p on the Highest setting — 35 fps if Windows’ efficiency settings are cranked to their most.
AI frames make an infinite distinction, in the event that they’re supported
It feels very unusual to check Intel’s Core Ultra Series 2 chip utilizing dialed-down benchmarks centered round 1080p gaming at Low settings — usually a touch for a PC gamer that it’s time for a brand new machine or card. But the Core Ultra Series 3 hit the 60 fps threshold that signaled a “playable” recreation even with simply rendered frames. Panther Lake’s GPU additionally consists of two totally different strategies of artificially growing body charge — 2X upscaling, or rendering a body utilizing a decrease decision after which growing the decision to the specified stage — and XeSS 3, which might interpolate three extra frames utilizing AI. Naysayers name these “fake frames,” however Panther Lake permits for purists and extra aggressive avid gamers alike to seek out what they need.
Our take a look at laptop computer shipped with Intel Graphics Software, a customized Intel app that lets you management numerous points of your show and graphics — together with forcing on XeSS body era, or AI-generated frames that may inject as much as 4 interpolated frames for each body the GPU renders. That’s huge — or is it?
What I found is that, sure, turning on body era could make an infinite distinction. Simply turning on upscaling and XeSS 3 elevated the framerate to a whopping 140 frames per second! Dialing up the Windows power slider tacked on a number of extra frames. Both are included within the “Max” end result on the prime of the chart.

Mark Hachman / Foundry
The results appear to vary relying upon the picture high quality, although. When working Cyberpunk on our conventional 1080p Ultra settings, body charges jumped from 52 to 92. Pushing the Windows slider to most efficiency gave me body charges of 143 fps.
The distinction, although, is that Cyberpunk particularly helps XeSS modes. Metro: Exodus doesn’t — and “forcing” XeSS on utilizing the Intel Graphics Software app didn’t work. Modern video games appear extra forgiving of older {hardware}, and help for AI body era actually makes these video games playable by trendy laptops. Still, I’m wondering if there shall be a tier of AAA video games like Metro: highly effective sufficient that Panther Lake laptops gained’t be capable to run them, however sufficiently old that they gained’t be capable to help the body era that might in any other case bridge the hole.
I attempted a handful of different video games. Total War: Warhammer 3 crashed when working the “battle” benchmark, however its marketing campaign map benchmark performed again at 44 frames per second at 1080p High settings. The 2014 Thief remake produced a even 60 fps when performed at 1080p on the Highest settings. Neither supported XeSS or any body era. Forza Horizon 6 generated 62 fps on 1080p Ultra settings with body era pressured on, however with out specific help for it.
Can Panther Lake compete with a 4050 laptop computer?
This was essentially the most provocative declare that Intel made about Panther Lake at CES, proper earlier than we had an opportunity to check out the chip on a prototype Lenovo laptop computer. Using purely rendered frames, it falls a bit quick. When body era is included, it retains up.
Would a gamer with a desktop PC working a GeForce RTX 5090 flip off AI body era? Possibly. I feel that the majority lovers, already feeling the pinch of skyrocketing RAM, SSD, and GPU costs, will activate body era with out a lot thought. Again, right here’s the Core Ultra X9 388H working Shadow of the Tomb Raider, with out body era, dealing with off in opposition to a variety of current, however older gaming laptops.

Foundry / Mark Hachman
And right here is the Core Ultra X9 388H working Cyberpunk 2077 with body era enabled. This looks like a scene from an 1980s TV present, the place Voltron lastly pulls out his blazing sword or Ok.I.T.T. goes into turbo mode. The episode can be loads easier if each had occurred from the get-go.
Basically, setting apart the scorn some have for AI and “fake frames,” AI body era is the “win” button right here.

Foundry / Mark Hachman
AMD’s Ryzen AI Max is an alternative choice
AMD tried to work the refs (us) more durable than Indiana soccer coach Curt Cignetti complaining about private fouls throughout a halftime interview. The firm claims that we needs to be evaluating Intel’s Panther Lake to AMD’s Ryzen AI 400 chips in addition to its Ryzen AI Max processor as a substitute.
To that, we are saying, ship us one! We’re glad to evaluate the Ryzen AI 400 when laptops lastly can be found. As for the Ryzen AI Max, properly — we’ve reviewed it inside a (Framework) Desktop, and we’ve seen it in an HP ZBook Ultra G1a laptop, too. As our evaluate benchmarks present, the Ryzen AI Max outperforms the Core Ultra X9 388H handily, although we’d most likely put it in a tier that Intel’s eventual HX gaming processors will finally compete in opposition to, reasonably than a power-sipping laptop computer chip.
AI is much less necessary than earlier than
But the Ryzen AI Max does have some extent, so to talk. If folks do need to run non-public LLMs regionally, the Ryzen AI Max (Strix Halo) does present gobs of VRAM mandatory for such LLMs to run. An AMD driver allowed the Framework Desktop to assign 96GB for working LLMs. Our Asus ZenBook Duo evaluate unit, which has an Ultra X9 388H and 32GB of RAM, provided 18GB of VRAM for video games and AI purposes. That consists of an NPU that may present 50 TOPS, or 122 whole TOPS with the GPU roped in.
And let’s face it — AI has struggled on the PC, leaving us questioning a bit if the early emphasis on the NPU was worthwhile. What we do know is that the graphics chip is essentially the most highly effective AI processor. UL gives a number of benchmarks; I’ve ditched the summary “Vision” benchmark in favor of the Procyon image-generation (AI artwork) benchmark. (The take a look at is a piece in progress, excluding Arm and offering an odd implementation for AMD’s Ryzen processor.) But UL’s take a look at can typically run on both the NPU or the GPU, with some exceptions.
Basically, this take a look at displays the rating UL assigns to the method. In the actual world, it reveals that the ZenBook Duo with a Core Ultra 3 chip inside creates a 512×512 picture as soon as each 4.5 seconds utilizing the GPU, whereas our take a look at laptop computer with Intel’s Core Ultra 2 chip inside creates the identical picture as soon as each 9 seconds. But the Ultra X9 388H’s NPU efficiency suffers, and the AMD’s Ryzen AI NPU outperforms it, too.

Mark Hachman / Foundry
The identical goes for working UL’s LLM benchmark. Originally, this take a look at was one of many few that evaluated the NPU, and that was helpful. But as Procyon begins including help for the GPU, it does make you marvel why we’re utilizing an NPU when a extra highly effective different is true there.
Procyon’s take a look at masses and runs a number of fashions, then gives a sequence of prompts and generates a rating. Some exams merely don’t run on some processors (Arm, once more) and solely run on a few the others. Some exams will solely run on the NPU. This take a look at is de facto finest to match the three generations of Intel Core Ultra processors.
Again, the take a look at doesn’t do an awesome job in describing real-world outcomes. In this case, the plugged-in Core Ultra 3 system working the LLama 3.1 (8 billion parameters) on the NPU generated about 20 tokens per second, which would seem in your display screen at about 4 characters per token per second–a cushty studying pace for me. Using the GPU underneath the Windows balanced settings, the token output was about 25 tokens per second for a similar mannequin. Running Llama 2 (13 billion parameters), the token output was between 13 to 15 tokens per second, which is perhaps slightly gradual.

Foundry / Mark Hachman
I thought of noodling round with Intel’s AI Playground, however the app stalled out when making ready the llamaCPP-GGUF backend, so I deserted the undertaking.
2026 shall be an attention-grabbing yr
I actually thought that 2025’s crop of laptop computer processors had been one of the best ever — you may purchase a laptop computer whose processors had been made by both AMD, Intel, or Qualcomm and go away glad. But 2026 appears prefer it might be even higher.
Remember, although, that Intel is first out of the gate with this new era of chips. AMD will finally reply with the Ryzen AI 400, and Qualcomm’s Snapdragon X2 Elite Extreme is ready within the wings. Given that Intel historically instructions about 80 % of the pocket book PC processor market, an early leap might be a strong benefit, particularly with solely older chips to match it to. But we’re not performed but!
