Intel's Panther Lake: A Leap Forward in Laptop Chip Technology

Intel's next-generation Panther Lake laptop chips could be a return to form

They say there's no such thing as bad publicity, but the press that Intel has generated in the last year has certainly been testing the boundaries of the aphorism. Is it good when your company posts an annual loss for the first time in almost 40 years? When you're doing multiple rounds of mass layoffs? When your board pushes your CEO into leaving, and starts arguing with the new CEO about major strategy decisions almost immediately? When you're either losing or failing to gain market share in areas critical to your bottom line? When you need to explain to the president why your CEO should keep his job, and then explain to investors the many possible downsides of a mercurial president deciding he wants a stake in your company? I feel like the answer to these questions is "mostly no." Even Intel's recent investment from and partnership with Nvidia came with a tacit admission that Intel was mostly failing to make a dent in AI hardware and software and the gaming and workstation GPU markets. (Reports that Intel could start manufacturing chips for AMD would be good news, as bizarre as that arrangement would have been at any other point in the two companies' history with one another, but those talks could still fall apart.) But before it was a company that primarily manufactured bad news, Intel was mainly known for making and selling computer processors. And Intel is hearkening back to that heritage today by taking the wraps off of a new series of chips, codenamed Panther Lake, that will form the basis of its next-generation Core Ultra laptop CPU lineup. Intel says that systems with these chips in them should be shipping by the end of the year. In recent years, the company has launched a small handful of ultraportable-focused CPUs at the end of the year, and then followed that up with a more fully fleshed-out midrange and high-end lineup at CES in January -- we'd expect Intel to stick to that basic approach here. Panther Lake draws near Panther Lake tries to combine different aspects of the last-generation Lunar Lake and Arrow Lake chips. Intel Panther Lake tries to combine different aspects of the last-generation Lunar Lake and Arrow Lake chips. Intel Panther Lake combines three functional chiplets using Intel's Foveros packaging technology. Intel Panther Lake combines three functional chiplets using Intel's Foveros packaging technology. Intel The three Panther Lake dies use the same package design, making for easy interchangeability for PC manufacturers. Intel The three Panther Lake dies use the same package design, making for easy interchangeability for PC manufacturers. Intel Panther Lake combines three functional chiplets using Intel's Foveros packaging technology. Intel The three Panther Lake dies use the same package design, making for easy interchangeability for PC manufacturers. Intel Intel's first Core Ultra chips, codenamed Meteor Lake, were introduced two years ago. There were three big changes that separated these from the 14th-generation Core CPUs and their predecessors: They were constructed of multiple silicon tiles, fused together into one with Intel's Foveros packaging technologies; some of those tiles were manufactured by TSMC rather than Intel; and they added a neural processing unit (NPU) that could be used for on-device machine learning and generative AI applications. The second-generation Core Ultra chips continued to do all three of those things, but Intel pursued an odd bifurcated strategy that gave different Core Ultra 200-series processors significantly different capabilities. The most interesting models, codenamed Lunar Lake (aka Core Ultra 200V), integrated the system RAM on the CPU package, which improved performance and power consumption while making them more expensive to buy and complicated to manufacture. These chips included Intel's most up-to-date Arc GPU architecture, codenamed Battlemage, plus an NPU that met the performance requirements for Microsoft's Copilot+ PC initiative. But Core Ultra 200V chips were mostly used in high-end thin-and-light laptops. Lower-cost and higher-performance laptops got the other kind of Core Ultra 200 chip, codenamed Arrow Lake, which was a mishmash of old and new. The CPU cores used the same architecture as Lunar Lake, and there were usually more of them. But the GPU architecture was older and slower, and the NPU didn't meet the requirements for Copilot+. If Lunar Lake was all-new, Arrow Lake was mostly an updated CPU design fused to a tweaked version of the original Meteor Lake design (confused by all these lakes yet? Welcome to my world). You'll still see multiple iterations of the new Panther Lake architecture, which will presumably come to market under the "Core Ultra 300" banner, but the chips will no longer use a hodgepodge of mixed-and-matched technologies. All Panther Lake chips are still assembled with Foveros; all Panther Lake chips get the same NPU capable of 50 trillion operations per second (TOPS), the same CPU and GPU architectures, the same media encoding and decoding capabilities, and external RAM (either soldered down or in SODIMM slots). The main difference is how many CPU and GPU cores you get, not what kind of cores you get. Intel is introducing three distinct Panther Lake chips, all using the same package design. Comparing the three different Panther Lake configurations. Intel Comparing the three different Panther Lake configurations. Intel The 8-core Panther Lake. Intel The 8-core Panther Lake. Intel Comparing the three different Panther Lake configurations. Intel The 8-core Panther Lake. Intel The 16-core Panther Lake. Intel The 16-core Panther Lake. Intel A version of the 16-core chip with less I/O but a bigger GPU. Intel A version of the 16-core chip with less I/O but a bigger GPU. Intel The 16-core Panther Lake. Intel A version of the 16-core chip with less I/O but a bigger GPU. Intel Each of these three chips is tailored toward a different kind of laptop. The 8-core version will clearly be the mainstream workhorse chip for most midrange Ultrabooks. The 16-core version's extra PCI Express lanes mean we'll probably see it the most frequently in bulkier workstations and gaming laptops with external GPUs. And the 16-core 12Xe version is aimed at high-end thin-and-lights without dedicated GPUs (it will also require soldered-down LPDDR5X, runs at faster speeds, and will maximize the larger integrated GPU's performance). Intel is sticking to broad and high-level performance comparisons for its new "Cougar Cove" P-core and "Darkmont" E-core architectures and the new Xe3 GPU architecture -- the company will probably get more specific when it's actually announcing the specific products, rather than the architecture. But Intel claims we can expect up to a 10 percent improvement in single-core CPU performance compared to Lunar Lake, and up to 50 percent better multi-core CPU performance compared to both Lunar Lake and Arrow Lake. The GPU is said to be roughly 50 percent faster. And Intel says the chip consumes 10 percent less power than Lunar Lake, and 40 percent less power than Arrow Lake. New CPU architectures. Intel New CPU architectures. Intel The media encoding engine makes some tweaks. Intel The media encoding engine makes some tweaks. Intel New CPU architectures. Intel The media encoding engine makes some tweaks. Intel On the manufacturing side, Intel has returned to make many of the components of Panther Lake in-house, but as with Meteor/Lunar/Arrow Lake, TSMC is still handling some of the silicon tiles. Each of these chips uses a total of three functional tiles, not counting the Foveros base tile that binds them together or the "filler tile" that makes them rectangular. * The compute tile houses all the CPU cores, the NPU, and the media engine and is built on the new Intel 18A process. * The platform controller tile that handles most I/O, including PCI Express and USB, is still built at TSMC. * The four-core version of the graphics tile is made using the Intel 3 process (mostly unused in consumer chips, though Intel does use it for server processors). The 12-core version of the GPU tile is still being outsourced to TSMC. Assuming Intel's performance figures all work out in the real world, early indicators suggest that Panther Lake should be something of a return to form. In older generations, Intel used consistent CPU and GPU architectures all up and down its desktop and laptop lineups, ensuring that chips had similar capabilities even though they targeted much different power and performance levels. Panther Lake returns to something approximating that level of simplicity, and as such requires a whole lot less explaining than the Lunar/Arrow Lake split did. High-level performance and power comparisons for Panther Lake. Intel High-level performance and power comparisons for Panther Lake. Intel A collection of Panther Lake features. Intel A collection of Panther Lake features. Intel High-level performance and power comparisons for Panther Lake. Intel A collection of Panther Lake features. Intel One thing that remains to be seen is how Intel scales some version of this architecture up to the desktop, where the company has (in the words of CFO David Zinsner) "fumbled the football" in recent generations. That's not what Panther Lake is about, but hopefully we'll see these CPU, GPU, and NPU cores remixed into a compelling high-end desktop chip sooner rather than later.

Next-Gen Intel Panther Lake Mobile Chips Target Faster Graphics on Lower Power, Better Battery Life

When we're treated to the annual deluge of laptop announcements at CES in January, chances are Intel's Panther Lake processors will be inside a lot of them. They'll probably show up in the top-end thin-and-light models, which tend to have the most cutting-edge integrated graphics and best NPU performance. Panther Lake debuts Intel's 18A 2nm fab process in its consumer chips, and a switch to a smaller process generally results in performance gains and new, denser chip layouts. Basically, we should see an increase in power and power efficiency. For comparison, the Apple M4 and upcoming Qualcomm Snapdragon X Elite are 3nm and AMD's Zen 5 is 4nm. The 18A process also introduces a new transistor architecture, dubbed RibbonFET, which lays the groundwork for future generations of chips. I find it interesting that Panther Lake has scaled up its AI performance not via the neural processing unit, which all the CPU makers have been hyping as the most efficient way to implement it, but via the GPU. (GPUs have always delivered better performance for AI-heavy calculations, they just use a lot of power for it.) Intel's new NPU 5 only bumps performance from a maximum of 48 trillion operations per second to 50 TOPS. That's a lot less than the Qualcomm, which goes up to 80. It does add support for FP8, a data type that's become common for having low overhead but better performance than its integer predecessor. Instead, the new Xe 3 graphics architecture in Panther Lake (for its third-generation Arc graphics) adds more, optimized cores (up to 6 per render slice compared to 4 on Xe 2), resulting in a boost from 67 TOPS in Xe 2 to up to 120 TOPS with Xe 3. For gaming, Intel claims the new Arc graphics is a lot faster than Lunar Lake's Xe 2-based Arc generation for a given power draw. Play longer, faster: the Holy Grail. Another notable update is with the image processing used by webcams. Intel's added support for "staggered" HDR acceleration -- a retronym for the original method of expanding a photo's tonal range by algorithmically combining bright and dark exposures. There's also improved noise processing. Overall, it means that photos and videos should have better low-light quality. There will be a few variations of Panther Lake chip configurations: up 8- or 16-core versions with up to 4-core GPUs and an up to 16-core 12Xe with a 12-core GPU. They all support faster memory (albeit at different speeds) as well as Wi-Fi 7 (R2) and Bluetooth Core 6.

Why Intel Panther Lake's 'boring' updates could mean big wins for gamers

Some of the more impactful features aren't as flashy, but translate to better connectivity and improvements for gamers. Intel recently hosted its Tech Tour 2025 event in Arizona, showing off the next generation of processor technology to press and analysts, codenamed "Panther Lake". The new line of chips is slated to power a wide range of upcoming laptop models with improved battery efficiency and performance with the new A18 process technology. Also: I saw Intel's new 'Panther Lake' chip first-hand, and it could be a turning point I had the opportunity to attend the event in Arizona, a short distance from the company's manufacturing plant, and what really impressed me wasn't just the performance upgrades. Some of the less-discussed technology could be game-changers when they finally land on laptops early next year. Here are a few technologies I think will be significant for the everyday user. I didn't expect to walk away excited by Panther Lake's improvements to wireless connectivity, but that's exactly what happened. First off, the chip will introduce support for 6GHz band to laptops. According to Intel, a 6GHz band is twice as fast as a 5GHz channel, with a bandwidth spectrum up to 1,200 MHz wide. Laptops running the new hardware will support a new flagship Wi-Fi 7 feature called Multi-Link Operation (MLO), which allows devices to connect to a access point across multiple channels, hopping between bands in order to ensure the best connection. Also: What Nvidia's stunning $5 billion Intel bet means for enterprise AI and next-gen laptops For example, let's say you're trying to transmit a big video file across a 5GHz channel, but it's congested because a lot of other devices are on the same band. MLO will automatically move you to the 6GHz channel, which, in this scenario, is congestion-free, improving connection speeds. Panther Lake will also add support for Wi-Fi 7 Release 2 (R2), which supports Multi-Link Configuration. This allows access points to detect if a Wi-Fi channel is being underutilized, and if so, will move all client devices over to a different channel, shutting down the previous band down to conserve energy. Panther Lake promises a major leap forward in gaming performance with Intel's new Xe3 GPU architecture. Compared to the older Xe2 design, Xe3 boasts 25% more threads, upgraded 12-bit vector engines, and an enhanced ray tracing unit to deliver realistic lighting. "These micro-architectural improvements translate into real-world performance," Intel Fellow Tom Peterson said during the gaming technology session. It's a common theme, with the improved A18 process technology resulting in better performance and efficiency. Also: Firefox just made it easier to separate your work and personal browsing - here's how Intel performed early benchmarks with Panther Lake, and the results were impressive. A single frame was rendered in just 22.84 milliseconds on the new hardware, compared to 45.44 milliseconds on the previous Lunar Lake architecture, so it's a significant leap in speed. These performance gains come from precise optimizations, such as a larger L1 cache plus variable register allocation. Panther Lake advances Intel's approach to graphical rendering via a new software technology called XeSS Multi-Frame Generation (or MFG for short). The chip creates a base image via rasterization techniques and then utilizes AI to generate up to three extra frames per native frame, resulting in ultra-smooth gameplay that feels like the frame rate is much higher. During the event, we saw Panther Lake running the game "Dying Light: The Beast" natively at around 30 FPS, but thanks to all the extra frames from XeSS MFG, the FPS presented itself to be much higher -- at over 130FPS. It maintained consistently smooth gameplay without any tearing or other graphical artifacts. Whether or not this means Intel will surpass AMD in the gaming remains to be seen and depends on plenty of other factors. The first laptops with Panther Lake chips will arrive in the first months of 2026, at which time we'll be going hands on to see what they can do.

Panther Lake sets stage for Intel's 2 nm comeback

Notebook chip promises 8 to16 cores and up to 180 TOPS of total AI performance when it hits shelves in January Intel has begun clawing back production from TSMC with the introduction of its Panther Lake processors, the company's first chip based on its long-awaited 18A process tech. The x86 giant began volume production of the processor's compute tiles at Fab 52 located at its Ocotillo campus in Arizona this summer, with the first notebooks and computers based on the SoC slated for release around the time of CES early next year. If Intel is to be believed, the process and architectural improvements introduced with Panther Lake will be well worth the wait. Chipzilla claims the process will deliver up to 10 percent single-threaded and 50 percent in multithreaded uplift in performance per watt over last gen's Lunar and Arrow Lake processors. Much of Panther Lake's efficiency and performance gains can be attributed to the chip's new process tech, which, in addition to a node shrink from Intel 3 to a 2 nm-class process it calls 18A, shifts power delivery to the back of the wafer. According to Intel, 18A's denser RibbonFet transistors account for more than a 15 percent increase in performance per watt, while backside power boosts transistor density by up to 30 percent. However, Intel hasn't quite quit TSMC just yet. Only Panther Lake's compute die, which contains the CPU, NPU, and image processors, is actually taking advantage of Intel's 18A. Much like Meteor Lake, Intel's latest client CPUs use a heterogeneous, multi-die chip architecture, which combines silicon from various wafer fabs and process nodes into a single package. For Panther Lake, the chip comprises three key components: These three chips are packaged alongside a few filler tiles on a base die that facilitates power and communications between them using Intel's Foveros advanced packaging tech. Speaking of SKUs, Intel will offer Panther Lake in at least three configurations: a base model with eight cores and two higher-end options with 16 CPU cores and the option of either additional PCIe connectivity for dedicated graphics or a beefier iGPU. Working up the stack, the base model features a compute tile with eight CPU cores. Half of those are Intel's new Cougar Cove performance (P) cores while the four remaining cores are arranged in a low-power efficiency (LP-E) core cluster based around its Darkmont microarchitecture. If Darkmont sounds familiar, that's because it's the same underlying architecture that underpins Intel's datacenter-centric Clearwater Forest Xeons we looked at during Hot Chips earlier this year. Located alongside the P and LP-E core clusters is a 50 TOPS neural processing unit and an upgraded image processing stack, which are shared across the entire line-up. I/O functionality is handled by a Platform Controller Tile that offers four Thunderbolt 4 ports and two USB 3.2 ports, Wi-Fi 7, Bluetooth 6.0, four lanes of PCIe 5.0 for storage, and eight lanes of PCIe 4.0 for additional connectivity and peripherals. The eight-core chip comes equipped with a quad-core GPU tile fabbed on the chipmaker's older Intel 3 process tech. Xe3 is new for the Panther Lake generation and represents a modest but not insignificant upgrade over the Xe2 cores found on last gen's Lunar Lake processors. Step up to the 16-core part, and things get a bit more interesting. In addition to the P and LP-E cores, the part picks up another eight E-cores which are connected directly to the chip's L3 cache. Intel has also upgraded Thread Director, an application designed to optimize workload provisioning across the various core types to function a bit like a waterfall. As the LP-E cores reach capacity limits, they're pushed into the more powerful E-cores and then onto the P-cores as needed. The idea here is that, for power constrained environments - a thin and light notebook, for example - it may be preferable to keep processes on the more efficient cores and only fire up the beefier P-cores when absolutely necessary. Moving on to graphics, there are two paths. The first is a larger, more powerful GPU tile from TSMC that triples the number of Xe graphics cores to 12. According to Intel, the GPU is up to 50 percent faster than the iGPUs found on its last-gen Arrow Lake and Lunar Lake parts. Alternatively, you can keep the quad-core GPU tile and opt instead for an upgraded platform control tile with eight additional lanes of PCIe 5.0 connectivity. This additional I/O should open the door to some interesting system designs dedicated to Arc, Nvidia, or AMD graphics, or even external GPU docks. Despite the improved CPU and GPU processing, one area that Panther Lake falls short of the competition is in AI PC performance. Panther Lake's all-new neural processing unit (NPU) manages to achieve 50 TOPS of AI compute - that's exactly two more than Lunar Lake, and far fewer than the competition. AMD's Strix Halo, which is already shipping, boasts 60 TOPS, while Qualcomm's upcoming X2 Elite will feature a 75 TOPS NPU. With that said, all three exceed Microsoft's 40 TOPS Copilot+ PC requirement, and when it comes to AI/ML workloads, software can have a bigger impact on performance than what the spec sheet might suggest. Speaking of TOPS, between the CPU, GPU, and NPU, Intel boasts its latest chip will deliver 180 platform TOPS. As great as that might sound, you can't harness all of that performance at the same time, at least not yet. Arguably just as important to AI workloads - and graphics-heavy applications like gaming - are memory capacity and bandwidth. Panther Lake supports up to 128 GB of DDR5 memory at up to 7,200 MT/s or 96 GB of LPDDR5 at 9,600 MT/s, which works out to between 112.5 and 150 GB/s. Memory bandwidth has major implications for graphics-heavy workloads like gaming as well as local AI inference on large language models, like Meta's Llama 3.1 8B. This is one of the reasons why we see GDDR or HBM used on dedicated GPUs rather than slower DRAM. While the memory specs represent a decent jump, particularly in capacity, over Lunar and Arrow Lake, they make it one of the slowest upcoming mobile processors in terms of bandwidth. Qualcomm's top-specced X2 Elite Extreme parts manage 228 GB/s, while AMD's Strix Halo processors manage 256 GB/s. Meanwhile, Apple's notebook processors can be specced for up to 546 GB/s of memory bandwidth. Speaking of the competition, Intel isn't ready to compare its latest AI PC parts against the best AMD, Apple, and Qualcomm have to offer just yet. As such, it's hard to say just how big a difference maker Intel's new process tech and CPU/GPU architectures really are. In fact, Chipzilla hasn't shared full details on things like pricing, clock speeds, or power consumption either. We're told to expect parts ranging from 15 to 45 watts, which suggests we may see a wider range of SKUs than the three main configs disclosed during the Intel Tech Tour. With so much riding on the chip, we can understand why Intel might be hesitant to share more until we get closer to CES. Intel hasn't had a great track record when it comes to executing on its promises, a bad habit that newly minted CEO Lip Bu Tan has made it his mission to squash. Getting Panther Lake right is especially important because, in addition to being its flagship mobile processor, it's a showcase for what Intel's Foundry division is capable of. Promise too much too soon and Intel could end up scaring away potential Foundry customers. We look forward to learning more about Panther Lake, and how it stacks up to the competition, as we get closer to CES. ®

Intel unveils 18A chips in major push to revive US semiconductor edge

Both chips are being manufactured at Fab 52 in Chandler Arizona.l's Intel has unveiled its most advanced client and server processors yet, marking a major leap in semiconductor technology. The new Intel Core Ultra series 3 and Xeon 6+ chips promise higher performance, energy efficiency, and AI capabilities, all built on Intel's groundbreaking 18A process. Panther Lake, the codename for Intel Core Ultra series 3, is set to power consumer and commercial AI PCs, gaming devices, and edge solutions. It will begin high-volume production this year, with the first units shipping before the end of 2025.

Panther Lake unveiled: A deep dive into Intel's next-gen laptop CPU

Take the low power of Lunar Lake, combine it with the performance of Arrow Lake, and you've got Panther Lake, Intel says. Intel's next-generation mobile processor, "Panther Lake," builds incrementally on the excellent "Lunar Lake" chip populating laptops right now. But there's something odd afoot: a "16-core, 12 Xe graphics cores" version that could be Intel's answer to AMD's Ryzen AI Max, complete with multi-frame graphics generation powered by AI. Intel has officially revealed its new Panther Lake architecture, and it gives enthusiasts a lot to chew on: a return to the performance (P-cores), efficiency (E-cores), and low-power efficiency (LP E-cores) cores of Intel's first Core Ultra chip, Meteor Lake. Intel has returned with a fifth-generation NPU capable of 50 TOPS and an image processing unit (IPU) that actually uses AI for some functions. Intel's "Xe3" GPU is worth some discussion all by itself, with its awkward branding and powerful multi-frame generation that will potentially elevate frame rates three or four times what they were before. Can Panther Lake address mainstream laptops, handheld PCs, and this new breed of "AI workstations" AMD's Strix Halo is aiming for? Well, with three separate processors, maybe. Intel's confidence in its 18A manufacturing process may be a bit overblown, however, as several portions of Panther Lake are still being manufactured overseas, including the 12Xe version of the new, disaggregated (separate) GPU tile. In terms of performance, we know some further details. Intel says that Panther Lake's single-threaded performance should be 10 percent higher than Lunar Lake at the same power. Compared to both Lunar Lake and Meteor Lake, Intel's Panther Lake offers more than 50 percent better multithreaded performance, Intel says. Intel is also claiming that the total Panther Lake system-on-chip will consume 10 percent less power than Lunar Lake, and demonstrated a trio of laptops running a Core Ultra chip from each generation to back that up. Some of this will be due to the design of the chips; some will come from the process technology, including the 18A manufacturing process that's become a critical part of Intel's future. Intel launched Lunar Lake at Computex 2024, and we had Lunar Lake benchmarking completed by late September. (Intel's been talking up Panther Lake for over a year, too.) Intel representatives were quite clear that CES 2026 in January will be the launch event for Panther Lake laptops, and they will come to market soon after. I suspect that Panther Lake will be marketed as the Core Ultra Series 3, but that's just an educated guess. "Panther Lake literally combines the power efficiency of Lunar Lake and the performance of Arrow Lake in a product family, and we're using Intel 18A to bring these architectures together," Jim Johnson, the senior vice president in charge of Intel's Client Computing Group, said at Intel's launch event in Phoenix. Intel is also quite proud of its 18A process, the foundation of Panther Lake, which Intel fellow Tom Petersen called the most expensive die Intel has ever made. "We can confidently say Intel Foundry is in production on the only two nanometer class process that was developed and will be manufactured here in the United States," Kevin O'Buckley, the senior vice president and general manager of Intel Foundry Services. Intel typically describes the architecture of its upcoming processors during its "Tech Tour" conferences like the one it held to describe Panther Lake near Phoenix, then provides the speeds, features, and prices of the actual processors you'll be able to buy at launch. This was a combination of the two: some details we know, and some we don't. Contrast that to Qualcomm's launch event of the rival Snapdragon X2 Elite: we know how fast each chip is and how many cores are inside them, but not the deep details of their architecture. Right now, Intel's "Panther Lake" consists of three chips, primarily consisting of the new "Cougar Cove" P-core and the "Darkmont" E-core and LP E-cores: We don't know how much power Panther Lake will consume; one executive on Intel's marketing team said that Panther Lake is "expanding on the segment that we addressed with Lunar Lake," and that the chip power or TDP will go "up and down" from there. That will obviously affect which products Intel's Panther Lake chip will fit into. Gameplay demos of an updated Painkiller game running on the chip used both a 45-watt reference platform as well as a 30-watt laptop, and Intel's Petersen referred to a "max" of 44 watts. In total, Panther Lake can address up to 96 GB of LPDDR5x memory or 128GB of DDR5 memory. Intel's Panther Lake is divided into tiles, or what rival AMD calls "chiplets": a disaggregated design that allows tiles to be placed and swapped in and out. (Intel even provided a knockoff Lego kit of Panther Lake as a tchotchke.) In Panther Lake, there is a compute tile, a GPU tile, a platform controller tile, and a "base tile" that the other tiles are mounted upon. All of the tiles are connected together via a second-gen scalable I/O fabric. The base tile connects to the active tiles via Foveros 2.5D packaging technology, allowing Intel to stack die one on top of the other. According to Johnson, the disaggregated tiles means more PC segments and more price points. "Panther Lake will be the most broadly adopted, globally available AI PC platform Intel has ever delivered," he said. Right now, none of these Panther Lake chips have product names, and Intel isn't telling us how fast they'll run either in sustained or in turbo mode. But they will not have hyperthreading; from Meteor Lake on, Intel rearchitected its performance cores to deliver high enough single-thread performance without the need to add an additional thread. Intel chief executive Lip-Bu Tan has called this decision a mistake, but Intel's plan to eliminate hyperthreading was already baked into the design. The GPU tile is kept separate. Separating the GPU tile theoretically means that Intel could just "drop in" a replacement, though it's a bit more complicated than that. Still, it's natural to see this and the eventual RTX chiplets being talked about as part of Intel's Nvidia investment as two points on the same line -- though Intel carefully declined to confirm this. Here's a detail you may find intriguing: the 16-core, 12 Xe3 chip isn't designed to connect to a discrete GPU, and that chip only contains a total of 12 PCI Express lanes as a consequence. The 16-core, 4 Xe3 version includes 20 PCIe lanes, which certainly implies that this version may appear in gaming laptops. Intel executives said that there's nothing there preventing 16-core, 4Xe3 chip from connecting to a discrete GPU, so that seems likely. "The 8-core will service the thin-and-light [market], the second [16-core version] with the small GPU is going to be attached to a discrete GPU, and the big one will have a life of its own, said Fuad Abazovic, principal at ACAnalysis. "It's whatever you like with Lunar Lake, but with double the cores and much wider, much bigger GPU, but you still keep the battery life." The 8-core Panther Lake chip will have 12 PCIe lanes (8 PCIe 4, 4 PCIe 5), while the 16-core chip will have 20 PCIe lanes (8 PCIe 4, and 12 PCIe 5). The 16-core, 12Xe Panther Lake option drops back down to the 12 PCIe lanes and configuration of the 8-core chip. Also of interest: integrated Thunderbolt 4, but not Thunderbolt 5. That means another year's worth of Thunderbolt 4 docks, with discrete Thunderbolt 5 controllers probably only attached to pricey gaming PCs. Enthusiasts shouldn't really care which process technology Intel will use to make its chips, though Intel has been nearly desperate to win customers for Panther Lake's 18A manufacturing process and the subsequent 14A process which will follow it. It had to be embarrassing for Intel to make Lunar Lake's key tiles at rival TSMC. Now, Intel has started to bring its manufacturing in house once again. All of the compute tiles are manufactured on Intel's 18A process, but all of the platform controller tiles are made at TSMC, as well the 12 Xe3-core GPU tile. The other 4 Xe3 GPU tiles will be manufactured on Intel's Intel 3 process. Inside the CPU tile are the P- and E-cores, the NPU 5, plus what's known as the Image Processing Unit 7.5 as well as the memory interface. Panther Lake is a system on a chip, so referring to it as a "CPU" isn't exactly correct. But onboard Panther Lake are two separate CPU architectures: the "Cougar Cove" performance cores, and the "Darkmont" efficiency cores. Stephen Robinson, its chief CPU architect and an Intel fellow, traced the roadmap: Lunar Lake had two types of cores, but it couldn't scale to a high frequency. Arrow Lake's Skymont efficiency cores improved performance, but didn't offer all-day battery. Lunar Lake offered dedicated power delivery for the E cores, and added additional cache memory, for a total of up to 18MB shared level-3 cache. Essentially, Panther Lake takes the low-power islands of Lunar Lake, combines them with the additional cores previous generations offered, and upgrades the performance cores for improved single-thread performance. (The latter aspect is usually what gets attributed to how quick and responsive an OS is, at least in the MacOS world.) Intel's new 18A process also offers what's called "backside power delivery," or PowerVia, which routes power away from the signal logic. That improves frequency while reducing idle power loss -- both exactly what Intel wanted to achieve. According to Robinson, Cougar Cove is "optimized" for 18A, with better branch prediction and memory disambiguation. In Darkmont, Intel achieved better prefetching. It also implemented what it calls "nanocode," where Intel has taken some microcode and added the ability to decode in each of its parallel front-end clusters, Robinson said. It all works out to 40 percent lower power for the entire Panther Lake chip at the same performance of Arrow Lake, or 10 percent more performance at the same power, Intel executives said. Intel also exposes how tasks will be routed through each type of core through a technology called Thread Director. In Panther Lake, Thread Director should feel familiar: a workload will first land on the low-power E-cores, then move to the full-power E-cores if it proves too much for the low-power cores. From there, it will be pushed to the performance cores, Rajshree Chabukswar, the Intel fellow in charge of the technology, said. Intel's technology is also smart enough to auto-assign some threads based on the application, so Microsoft Teams will always begin on the LP E-cores and likely remain there. A benchmark like Cinebench, which asks for all cores and threads, will get them across all three different types of cores. In games, however, Panther Lake can do something interesting: assign the game to E-Cores, then to P-cores...but then use some of the leftover power and give it to the GPU. "Using OS containment zones and some of our graphics driver hints and our power management, we are able to deliver 10 percent [more] frame rate because we are making power headroom available to graphics," Chabukswar said. Normally, users can adjust the Windows 11 performance by adjusting the Windows power slider. With Panther Lake, Intel is going a bit further with what it's calling the Intelligent Experience Optimizer. Intel already has the Dynamic Tuning Utility, which uses AI to dynamically optimize the system for performance, battery life, or thermals -- like the "AI Mode" in an MSI laptop's system settings. Most gaming laptops already have "turbo" or "silent" modes, but the impression that Intel gives is that users will have more input into how the laptop performs. "It's purely load-driven. When I see that I need more performance, it will move there," Chabukswar said. "When I see that I need more efficiency, it will move there automatically." In two examples, UL Procyon's Office Productivity benchmark and the single-threaded Cinebench 2024 benchmark, turning on Intelligent Experience Optimizer boosted performance by 19 percent on both tests. Intel's integrated graphics path continues: Meteor Lake included Xe, Lunar Lake added a Xe2 GPU, and Panther Lake moves to Xe3. The killer selling point? Multi-frame generation, which injects AI-generated frames between actual rendered frames. However, there's an issue with branding. Intel's "Battlemage" discrete GPU was considered part of the Xe2 generation, while Intel's Panther Lake contained "Xe3" cores. Yet both Battlemage and Panther Lake are also both part of the Intel Arc B-series of products, which doesn't make much sense. Intel did show a roadmap that signals a "Xe3P" variant is coming, however. Panther Lake, from what Intel showed, will have both 4Xe and 12Xe variants. Not everything scales linearly; while the 32 XMX engines on the 4Xe variant triples to 96 in the 12 Xe version, the amount of level-2 cache quadruples from 4MB to 16MB. Who is the 12Xe version of Panther Lake aimed at? "The cheesy answer is 'everybody," said Daniel Rogers, vice president and general manager of PC products at Intel, in an interview. "I think you'll see it show up in a few ways, One, maybe the most obvious is for gamers. In some gaming designs and some handhelds as well -- that's a good fit." Rogers said the 12-Xe Panther Lake will also be the "flagship solution for commercial notebooks on the AI side," Rogers said. Is it Intel's answer to AMD's Strix Halo? "Everybody's chasing high-performance local AI, and we will certainly do the same," Rogers said. The Xe3 engine also improves the ray tracing, eliminating backups in the pipeline through a new thread sorting unit. It also doubles the anisotropic filtering on die, Petersen said. The Xe3 GPU was re-designed for scalability, Petersen said. In the GPU's render slice, Intel increased the number of Xe cores from four to six, increased the level 1 cache from 192 kilobytes to 256KB, and upgraded the level-2 cache from 8 megabytes to 16MB, reducing the need to access the local memory and increasing performance. Intel also moved to a variable allocation strategy when assigning threads, with "dramatic" effects on performance, Petersen said. While Intel declined to show actual benchmarks, Petersen did show how traffic to the SOC's memory fabric dropped by 17 percent to 36 percent on key games. He also demonstrated some "micro benchmarks," or internal tests that Intel itself uses to determine generation-over-generation performance in specific tasks. The X3 architecture also improves the way in which a frame is rendered, including DirectX calls and pre-rendering, but Petersen said real gains were granted by the larger level-1 cache. All told, Intel was able to cut the time to execute one frame to about 22.84ms on the 12Xe chip, versus 45.44ms for Lunar Lake. Essentially, Intel can generate about 50 percent higher GPU performance than Lunar Lake, or about 40 percent more performance per watt than Lunar Lake, Petersen said. The Xe3 GPU also supports cooperative vectors, which Intel showed off with Microsoft this June. Cooperative vectors use AI as a way to replace the overwhelming requirement for shaders. Microsoft has also proposed storing them in tbe cloud. Cooperative vectors, in Petersen's words, essentially "replace that render pipeline, that raster process, with a 3D model." "So effectively, we can completely eliminate the render pipeline and replace it with a per pixel AI," Petersen added. "You can see this might be kind of a look forward into a future where graphics is really substantially different, and it is primarily AI," Petersen added. That future is already here...sort of. XeSS 2, launched alongside Intel's "Battlemage" GPUs, launched XeSS Super Resolution, XeSS Frame Generation, and Xe Low Latency, all designed to speed up the graphics powering video games. XeSS Frame Generation injected an AI-generated frame between two "real," rendered frames, and used the low-latency technology to offset the delay that engendered. "Hybrid rendering is kind of where we are for most titles today," Petersen said. "Things have indeed moved forward, and we are now into the world of hybrid rendering, plus AI. And in this world, not all pixels are rastered. In fact, most pixels are not rastered. They are generated." Now, Intel has launched XeSS-MFG, or XeSS Multiframe Generation, which can inject as many as three additional interpolated frames, in conjunction with the existing upscaling and multiframe generation technologies. You'll be able to control it as paert of Intel's existing Intel Graphics Software package, where you'll have the option of setting the additional frames, or letting the application itself make that decision, Petersen said. That software will also offer options such as specifying how much system memory is shared with the GPU, a feature Intel announced earlier and that AMD has exploited for better AI. The idea is simple: more frames means a smoother gaming experience, and happier gamers with PCs which can exceed playable frame rates. Even games designed to benefit from XeSS 2 will support XeSS 3, Petersen said. Yes, those frames will introduce lag. Each generated frame will add about 1.5 milliseconds of latency per frame, with about 1 ms of setup, Petersen said, or about 6 ms per three generated frames. That's bad when the frame rate is always high, he said, but when frame rates are at 25 to 30 fps, the latency isn't as bad versus the value of the added frames. According to Petersen, the latency or lag that bothers gamers isn't really the "click to photon" latency where there's a delay in the player's action. The more noticeable lag is what he called "motion to photon," where the latency between a user's input and perceived motion creates nausea in VR and wobble when mousing. Eventually, that could be solved by AI prediction of mouse movements or a variable rate of rasterized to AI- rendered frames. "There's a whole bunch of technologies that have not been announced, that are not done, that could make that [latency] better, but right now you can see some," Petersen added. "And if you're more latency sensitive, depending on the game type, I would say, turn off frames and just run regular frames." In addition to working toward cooperative vectors, Intel is also launching precompiled shader distribution, where your PC doesn't have to wait to compile shaders, it just downloads precompiled shaders from the cloud. It is also developing Intelligent Bias Control version 3, first launched this summer, in which the CPU and GPU talk to one another and route power to the appropriate logic. It's the complement to Thread Director, where GPU performance could be increased by 10 percent by that technology alone. Finally, Intel's PresentMon software will differentiate between rasterized and rendered/generated frames, and provide an indication whether a given frame was real. It will also demonstrate animation stutter, if frames are out of sync. According to James Sanders, a senior analyst at TechInsights, said that there's potential for creating a market for thin-and-light gaming laptops with the 16-core, 12Xe3 Panther Lake part. "Games that are like Marvel Rivals I would imagine would be a really good fit for that type of hardware," Sanders said. "If you're doing esports, that's something that would be a good fit. But if you're playing Final Fantasy, where you're worried more about the graphics performance, you'll want to go to a discrete GPU. But it is clear that that communicating what that market is going to be something that Intel and their OEMs need to work on, and that's going to take time." The story of Panther Lake's NPU: it's about the same as Lunar Lake, just more efficient. Lunar Lake's NPU 4 supported 48 TOPS. The NPU 5 inside Panther Lake supports 50 TOPS in a much smaller package. An NPU basically performs a ton of multiply-accumulate operations, performed by what's known as a MAC array. Lunar Lake and Panther Lake both have the same number of MACs -- 12,000 -- but Lunar Lake spread those out into six neural compute engines, while Panther Lake has only three. Panther Lake's NPU 5 also has 4.5MB of scratchpad RAM, 256KB of level-2 cache, and 6 SHAVE DSPs. Panther Lake's NPU can perform 4,096 MACs/cycle at int8 (8-bit integer), 4.096 MACs/cycle at FP8 (8-bit floating-point), and 2,048 MACs/cycle at FP16. Those numbers become more significant and more familiar when you're performing generative AI functions, where you're specifying the complexity of a particular task. "What we really wanted was more MACs and less of the other stuff," Petersen said. Some of the improvements are just Intel trying to be smarter about how it handles data. Running Stable Diffusion's AI art generator, the algorithm doesn't really need the more complex FP16 granularity. Instead, FP8 can be used; in Petersen's example, energy usage dropped 35 percent from 108 joules to 70 joules. Panther Lake also includes what's known as the Image Processing Unit 7.5, responsible for taking the image your PC's webcam sees and making it look its best. "That means handling the full spectrum of lighting from bright outdoor sunlight streaming in through a window to dim corners of a warehouse at night," said Tomer Rider, the IPU product marketing manager. "It means capturing and preserving fine details and high resolutions without introducing the noise, so what you see is crisp and accurate. It also means creating a life like natural image, vivid colors, true to life, skin tones and high frame rate that makes everything feel real wherever you are and whatever the line is, the IPU makes sure that the image looks its best." For Panther Lake, the IPU does three things: it adds enhanced HDR for wider dynamic range, and includes AI-based noise reduction and AI-based tone mapping. The "staggered HDR" takes both a short exposure and a long exposure and merges them together. All of these work with both integrated laptop webcams as well as standalone devices. Up to 5 Mpixel sensors are supported, which translates to 2560 x 1920 pixels in the real world, all whie shaving off 1.5 watts or so of power. AI noise reduction actually uses AI to filter out noise, especially in low-light environments, Rider said. And tone mapping divides up an image or video into different regions, then uses AI to improve the visuals. In this case, the IPU may be actually asking the NPU to perform these functions. "Either the NPU or the GPU" will handle the AI, Rider said. "We're working with all the NPU and GPU teams in order to make sure that the system makes the right choice when selecting whether it's the GPU or the NPU." Finally, Panther Lake also includes a wireless module, which places the Bluetooth and Wi-Fi MAC onto the chip and separates the remainder into a separate die inside the package, known as Whale Peak 2. Panther Lake supports both Wi-Fi 7 and Bluetooth Core 6.0, but with some neat twists. The most interesting wireless features Panther Lake adds are part of Bluetooth. Panther Lake now supports Auracast, a wireless technology launched in 2024 that supports simultaneous playback across multiple devices, so that you don't have to share an earbud with a friend. Intel's chip also supports platform sounding, which uses distance- and phase-change modeling to track the distance between two devices, to better locate them quickly. Bluetooth also uses both wireless antennas, and not just one, which should lengthen Bluetooth connectivity. Intel fellow and wireless CTO Carlos Cordeiro used 52 meters as an example distance. Panther Lake actually supports what's known as Wi-Fi 7 Release 2, which haven't been officially certified. Basically, the R2 features facilitate better communication between the access point and your PC, resulting in fewer dropped connections and improved communication speed and latency, Multilink reconfiguration allows the router to shift your PC from one channel to the other, while restricted TWT allows the two to figure out which applications (and devices) deserve the most priority. Single-link eMLSR allows one of the two antennas on your laptop to "sniff" out a different channel, and quickly shift if it opens up. It also allows the laptop to signal of it's using a channel for peer-to-peer communication. You probably don't use Intel's Connectivity Performance Suite, a small Windows app which allows you to see the various access points available to your PC and allows it to prioritize voice and video calls or streaming apps. Intel is adding AI to this list of options, so if yor PC is having a prolonged session with a cloud service like ChatGPT, you can ensure that that traffic is given priority. Intel's deep dive in the desert certainly opened the doors to reveal a number of features about Intel's next-gen mobile chip. What we still don't know is how many versions of the chip itself will eventually ship, and what they'll be called. Reports from Asia now indicate that the 12Xe version will be branded as the Core Ultra X, to differentiate them from the "vanilla" version of the Core Ultra chip itself. Naturally, we'll have to wait even longer for the first performance testing of the chip itself, as well as the subsequent announcements of laptops using the chip. Those should happen at or near the time of the CES 2026 show in January, we're told. It's there we should expect customer announcements, as well. It was a little surprising not to see any future Intel CPU roadmaps, at all, after Panther Lake has been talked about for over a year. "Nova Lake" should be the next step. "Meteor Lake served its purpose," said Mario Morales, the group vice president and general manager of semiconductors at IDC. "You needed to have a dog in the race. That can be easily relegated into the lower end. Lunar Lake came in and hit the sweet spot, and it did emphasize a lot more performance efficiency, which I think was another important message for them as they introduced new P-cores and E-cores into the...PC space. Panther Lake is important for them, as it's the next step. Nova Lake is what gets them to par -- with AMD for sure, but also maybe with Apple." Still, we now know details of Qualcomm's Snapdragon X2 Elite as well as Intel's Panther Lake. Waiting in the wings: AMD. Is Gorgon Point the next name you'll need to know about in laptop processors for 2026?

Intel's Panther Lake chip is ready to roar, promising faster AI and cooler laptops

Intel's upcoming 18A-based chip could be its biggest leap in years, promising faster AI performance and improved efficiency. What's happened? Intel is preparing to lift the lid on its Panther Lake laptop processor, a chip that could mark a major turning point for the company. Slated for a technical reveal soon, the new CPU is built on Intel's long-awaited 18A process and is expected to headline the next generation of AI-powered laptops. Panther Lake will be the first Intel chip built entirely on the 18A node, part of its "five nodes in four years" strategy. Internal tests point to up to 30% energy efficiency gains and up to 50% better performance in select workloads. The chip integrates improved AI cores and upgraded graphics for smoother on-device AI experiences. Yield rates are around 10% (up from 5%) based on reports, suggesting steady progress in 18A manufacturing. Why this is important: This isn't just another processor launch, but it's Intel's latest move to reclaim leadership from AMD and TSMC in the laptop performance race. While AMD and Qualcomm have sprinted ahead with purpose-built AI silicon, Intel's betting that Panther Lake's new 18A process will finally help it close the gap. Panther Lake also serves as a proof point for Intel's in-house manufacturing ambitions, after years of delays and missed milestones. Reinforces Intel's commitment to compete head-on with advanced foundries like TSMC. Positions the company to power a new generation of AI PCs focused on smarter, faster workflows. Success with 18A could restore investor confidence and set the tone for Intel's future chip designs. Recommended Videos Why should I care? For laptop buyers, this could mean a tangible jump in both speed and efficiency. Not to forget, it also hints at better AI integration for creative and productivity workflows. Additionally, Panther Lake might finally bring longer battery life and cooler operation. And if early leaks are to be believed, the lineup could also feature improved integrated GPUs under the new Core Ultra X series, promising more balanced all-around performance even without a discrete GPU. Expect laptops that handle complex AI-driven tasks, from editing to automation, with less reliance on cloud computing. Improved thermal design and efficiency could make thin-and-light notebooks more capable for everyday multitasking. Upgraded iGPUs may offer a meaningful boost for casual gaming and GPU-accelerated workloads, reducing the need for dedicated graphics on mid-tier systems. Okay, so what's next? Intel is expected to share more about Panther Lake's architecture, benchmarks, and rollout timeline during its Thursday event. The first wave of laptops featuring the new chip is rumored to arrive in early 2026, from brands like ASUS, HP, and Dell.

Intel Bets Recovery on Panther Lake AI Chip as Foundry Bleeds Billions - Decrypt

AMD's OpenAI GPU deal, Apple Silicon, and Qualcomm's PC chips tighten the noose. Intel unveiled its Panther Lake processor architecture Thursday at its Arizona campus, staking the company's future on a chip that CEO Lip-Bu Tan calls critical to reclaiming lost ground in the AI-powered PC market. The Core Ultra Series 3 processors, code-named Panther Lake, mark Intel's first client products built on its 18A manufacturing process. Intel says the chips will begin shipping later in 2025 from Fab 52 in Chandler, Arizona, with broad availability in January 2026. It's a big geopolitical move Intel is attempting to play up. "The United States has always been home to Intel's most advanced R&D, product design, and manufacturing, and we are proud to build on this legacy as we expand our domestic operations and bring new innovations to the market," Lip-Bu Tan said in a statement. The announcement comes as Intel's foundry division continues to bleed cash. For Q2 2025, Intel's foundry business posted an operating loss of $3.2 billion on $4.4 billion of revenue, per summarized figures from the company's earnings coverage. Intel's previous Q2 2025 results show the broader restructuring and cost controls related to the turnaround. Last month, the White House announced a strategic agreement with Intel, in which the government acquired a 9.9% stake in the company by purchasing 433.3 million primary shares, investing a total of $11.1 billion to provide Intel with sufficient cash to invest in new technology and remain operational. Competition is fierce. Apple's latest Apple Silicon laptops routinely post strong multithreaded and single-core results versus prior-gen Intel i9s in public benchmarks. Meanwhile, AMD is making serious inroads in desktop CPUs. As of Q2 2025, AMD captured 32.2% of desktop CPU unit share and 39.3% of revenue share, narrowing Intel's dominance to roughly 2:1 from estimates of around 9:1 in prior years. AMD is not just competing in PCs; it's also forming large AI partnerships, including a confirmed multi-year 6 GW GPU supply deal with OpenAI this week. Qualcomm's PC push, meanwhile, centers on efficiency, with early reviews and databases highlighting strong performance-per-watt for Snapdragon X Elite. On the technology front, Intel's 18A process brings RibbonFET (gate-all-around) transistors and PowerVia backside power delivery. Intel claims up to 15% better performance per watt and 30% higher chip density compared to its Intel 3 node, making the chips a sub-2 nm-class alternative for advanced client and data-center silicon. In other words, these new chips use highly advanced technology and are designed to be efficient in AI tasks. The company needs Panther Lake to land. With Lip-Bu Tan acknowledging there are" no quick fixes" for Intel's situation and analysts openly debating structural options, Intel's next-gen client platform is basically a bet that can either save or ruin the once dominant American chipmaker.

Panther Lake revealed: Intel's pocket powerhouse is a big step up from Lunar Lake, with performance and efficiency gains all around

But is the cost of multiple tiles and Foveros the answer to AMD's monolithic approach? After months of rumours, leaks, and tidbits from the company itself, Intel finally announced its Panther Lake CPUs at its regular Technology Tour event, this time in Arizona, US. It was an appropriate location to do so, as the US state is also home to multiple Intel foundries, and Fab 52 is one of two that operate its newest process node: Intel 18A. This particular semiconductor chip production method is responsible for making the key tile, or chiplet, that lies at the heart of every Panther Lake processor. But where Intel's previous tiled architectures, Meteor Lake (MTL) and Arrow Lake (ARL), sported one design apiece, Panther Lake or PTL for short, comes in three varieties. With Meteor Lake, Lunar Lake, and Arrow Lake, Intel created a single processor design and used binning strategies to create broad processor families from them. For example, one LNL chip can be found in nine different Core Ultra 200V CPUs. Not so with Panther Lake, because while binning will still be used to produce all the different models (i.e. some will have disabled cores, others will just be clocked lower), there are three chip designs in total. This isn't just one chip with disabled bits to make the others; Intel is using multiple tiles to create these configurations. Making things more complicated to explain is the fact that Intel has also created two different tiles for each of the primary sections: compute, graphics, and platform controller. To better understand how each PTL variant differs, let's consider the new tiles separately. Starting with arguably the most important chiplet in Panther Lake, Intel's new compute tile is going to be under close scrutiny in the coming months. That's because no small part of what makes the compute tile special is down to the 18A process node, with its gate-all-around RibbonFET nanoscale switches and PowerVia backside power delivery. Both compute tile versions house P-cores and E-cores, along with Intel's latest NPU (neural processing unit), an updated IPU (image processing unit), media and display engines, plus the DRAM controllers and interfaces. The last aspect is a notable improvement over Arrow Lake, which has the memory system in a separate tile from the compute one, which isn't good for achieving low latencies. There's also 8 MB of 'memory-side' cache that's accessible by all cores. It's a bit like an L4 cache, but it's better to think of it as being an ultra-fast buffer for handling data transfers between RAM and the various clusters of cores, rather than an all-encompassing cache. It's designed to help reduce traffic on the memory bus, as well as improve overall latencies within the processor. Intel doesn't have a codename for the different compute tiles, so I'll just refer to them as 'small' compute and 'large' compute, although there's not a huge size discrepancy between them. Both of them house four Darkmont Low Power E-cores in a separate 'island' within the tile, but where small compute only sports a further four Cougar Cove P-cores, big compute boasts an additional eight Darkmont E-cores, grouped right next to the P-cores. It's fair to say that the new P- and E-cores aren't a revolution in CPU architecture; instead, they're a small, evolutionary step up from previous designs, sporting a variety of improvements to achieve more performance, at the same clock speeds and power, as before. For example, Cougar Cove gets a 50% larger TLB and a better branch prediction unit. The former is a translation lookaside buffer, which speeds up memory requests by storing the most recent address translations (virtual to physical), whereas the latter reduces thread stalls by calculating what instructions and data are most likely to be processed next. It's a similar story with Darkmont: better branch prediction and prefetch, more instructions covered by the nanocode cache, and a wider data bus to the L2 cache. By themselves, none of the changes are major, but taken altogether, Cougar Cove and Darkmont give Panther Lake roughly 10% better single-threaded performance, compared to LNL and ARL-H. The NPU is a lot more compact, thanks to the 18A process node and a fresh design, though your average PC user is perhaps still waiting for that 'killer app' to make an AI-capable CPU a must-buy. However, in the edge market (Intel describes this as everything that isn't PC or server), the 50 TOPs peak output of the NPU, in such a small, low-power package, will be greatly appreciated. Combined with the cores and the GPU, Panther Lake as a whole can achieve up to 180 TOPs (trillions of operations per second). That's 50% more than Lunar Lake and a massive 400% better than Arrow Lake, though the latter only has a tiny NPU and iGPU. Like every other chip company, Intel is betting heavily on AI, so improvements in this area were to be expected. This is also why the compute tile houses the DRAM controllers and interfaces, and the most capable PTL variant supports up to 96 GB of LPDDR5x-9600 or 128 MB of DDR5-7200. That's a little better than AMD's best APUs, the Ryzen AI 9 HX 375 and Max+ 395, which support up to LPDDR5x-8000, but the best news is that, unlike Lunar Lake, the DRAM is no longer on package, so more vendors will be willing to use it. Adding better cores, more cache, and faster RAM is all well and good, but PTL's fundamental layout demands top-notch thread scheduling to maximise performance. Where AMD's processor cores are all the same in each architectural generation, Intel's P- and E-cores are very different. Fortunately, the Thread Director part of the new compute tile has been given a thorough overhaul and Intel is claiming some pretty significant gains in how well threads are scheduling on its hybrid chip. This is especially true for gaming, where better power management also kicks in, ensuring that a game's threads are correctly sent to the right core and the iGPU gets the lion's share (or should that be panther's share?) of the available power. Following on from Meteor Lake and Arrow Lake, Panther Lake's integrated graphics processor is a separate tile all by itself, though the media and display engines are within the compute tile to improve latencies and power efficiency when running videos. I've covered all the important details about the new GPU architecture, Xe3, in a separate article, but to summarise briefly here, it's a bit like the CPU cores: small improvements rather than a wholesale change in direction. Just as with compute tile, Intel has created two GPU tiles. The smallest is fabricated on its own Intel 3 process node, and it really is small, just 4 Xe3 cores (512 shaders) hooked up to 4 MB of L2 cache. That's the same size as Arrow Lake's iGPU (which is half as small as Lunar Lake's), but the base PTL configuration isn't really designed to be a gaming powerhouse. That honour goes to the other GPU tile, as it boasts 12 Xe3 cores (1536 shaders) and a whopping 16 MB of L2 cache. Per core, that's 48% more cache than in any Battlemage graphics card. Along with the architectural improvements, better memory latencies, and faster RAM support, and the fact that it's made on TSMC's N3E process node, the 12Xe PTL chip is going to be a serious contender for the best processor for handheld gaming PCs. Unlike the compute and graphics tiles, the two platform controller designs are very similar and differ only in terms of the number of PCIe lanes. Packed inside both of them are Thunderbolt 4, USB 3.2, and USB 2.0 controllers for up to four, two, and eight ports, respectively. It might seem disappointing to see that USB4 doesn't feature, but TB4 is fully compatible with USB4 anyway, and is just as fast (if one ignores USB4's 80 Gbps mode, which is barely supported at the moment). When it comes to PTL's PCI Express features, the smaller platform controller only offers 12 lanes, four of which are PCIe 5.0 and the remaining eight being Gen 4. The bigger tile has 20 lanes in total, adding a further eight Gen 5. The base PTL layout and the 12Xe version use the small tie, with the 16-core 4Xe one using the larger. Why? Because it's expected to hand over eight PCIe lanes to connect to a discrete GPU. The platform controller tile also houses a second-generation Intel Wi-Fi 7 module, along with Bluetooth Core 6.0 and Bluetooth LE. These are just the MACs (medium access controls) and the physical interfaces for both wireless connections are housed in the motherboard chipset. Intel claims its new Wi-Fi 7 module is significantly better than its first iteration, though with so many variables affecting wireless performance, you may find your new Panther Lake laptop no better in this aspect. You should, however, find Bluetooth to be a lot better. Intel claims its new system now has twice the range of the previous BT modules, along with better features for audio streaming over BLE. It might come as a surprise to learn that the platform controller tiles are manufactured on TSMC's N6 process node, something that's five years old and essentially nothing more than a tweaked N7. However, the tile contains a lot of analogue circuitry for all the PCIe and USB signals, and that scales very poorly with node shrinks. Had Intel used something like N3 to make the platform controller tile, it would only be a fraction smaller, run no better, and cost an awful lot more to make. In other words, N6 is the sensible choice. In many ways, Panther Lake as a whole is a sensible choice. Sure, we'd all like more P-cores in the compute tile, but rather than that, Intel has instead focused on improving its thread scheduler and power management system to ensure that threads go to the right cores. A USB4 controller would be neat, too, but that would make the platform controller tile bigger and more costly, and not every laptop needs that kind of port at the moment. For those that do, it can be implemented via the motherboard. Intel could have copied AMD and made the whole thing monolithic, either on Intel 18A or some cutting-edge TSMC node, but that would have made it much harder to offer physically different variants to suit the demands of different markets, and it would have also served to reduce potential wafer yields, too. Panther Lake has a lot of promise, and while we don't officially know what Core Ultra names the chips are going to field (leaks point to it being Core Ultra 300 for the Xe4 variants and X300 for the 12Xe models), what I saw running at the Technology Tour event honestly impressed me. That big GPU chip handled the Painkiller reboot at 1080p well enough to run at 50 fps natively, all while using less than 45 W of power. I saw an engineering sample churn through unoptimised AI code, utilising the CPU cores, NPU and GPU, at more than twice the speed of an ARL-H chip, while using a fraction of the power. A demonstration of the image processing unit, handling video streams from multiple cameras at an intersection, and then pushing it all through an AI algorithm that determined traffic flow speeds in the blink of an eye, would be impressive on a full-power desktop CPU, let alone an energy-sipping mobile chip that's primarily aimed at laptops. Intel could have gone totally left-field with Panther Lake; it could have equally gone down well-trodden paths from days past. Instead, it's been sensible and created a processor the market actually wants: powerful, flexible, and energy efficient. There's only one big question mark hanging over it all, and it's about how much Panther Lake chips are going to cost. The new processors won't be officially launched until next year, at the CES 2026 event, but Intel's Fab 52 is already working flat out, producing 18A wafers for Panther Lake and the enormous Clearwater Forest server chips. Both the foundry and the process node are new, and cost billions of dollars to develop and make. Each PTL processor is a complex affair: five tiles (though two of which are just blank fillers) on top of a large passive base tile, mounted via Intel's Foveros 2.5D technology. All of this adds to the expense of making the chip. Intel faces stiff competition from AMD, Qualcomm, and less directly, Apple. These already have great APUs and laptop processors in the market, most of which are monolithic designs, which therefore makes them cheaper to package. Many of them are made on TSMC's N4 node, which will have been in mass production for over a year, by the time Panther Lake comes to market. A company flush with cash and bountiful operating margins might be willing to sell such processors for little in the way of profits, to ensure it makes rapid traction in the market. That's not something Intel has typically done in the past, but its finances aren't so hot at the moment, because it has spent huge sums of money on foundry projects that have been subsequently closed down. Then again, it's recently shaken hands with Nvidia to co-develop new processors and enjoyed a healthy injection of capital as part of the deal. Arrow Lake launched with high price tags but has since undergone some significant price cuts because of a dearth of sales. So did Lunar Lake, but that's because its on-package DRAM makes it costly to package and thus forces a high price. I suspect that Intel won't sell Panther Lake to partners and other vendors at a low price or at the very least, not to begin with, and that's because the processor itself is significantly better than any Core Ultra 200V (Lunar Lake) or 200H/HX (Arrow Lake) model. You're getting the same amount of performance for a lower power consumption, or more performance for the same power; the feature set is broader and newer, and the AI capabilities are superior, too. Panther Lake is also proof that the move to a tiled, modular processor design was the right decision for Intel. Meteor Lake felt like nothing more than an experiment, and as good as Arrow Lake is, it's not a comprehensive package. This new CPU feels like the real deal, and if the final product is as good as what I've seen so far, then it will have a real winner on its hands. This bodes well for Nova Lake, and if that's just as good, Intel may well have done enough to wipe the slate clean of the numerous mistakes it's made over the past few years.

Intel Unveils Panther Lake Architecture: First AI PC Platform Built on 18A

Today Intel revealed the architectural details for the company's next generation client processor Intel Core Ultra series 3 (code-named Panther Lake) which is expected to begin shipping later this year. Panther Lake is the company's first product built on Intel 18A, the most advanced semiconductor process ever developed and manufactured in the United States. Intel also previewed Xeon 6+ (code-named Clearwater Forest), its first Intel 18A-based server processor, which is expected to launch in the first half of 2026. Both Panther Lake and Clearwater Forest, as well as multiple generations of products built on Intel 18A, are being manufactured at Fab 52, Intel's new, state-of-the-art factory in Chandler, Arizona - a key milestone as Intel invests in strengthening American technology and manufacturing leadership and building a resilient semiconductor supply chain. "We are entering an exciting new era of computing, made possible by great leaps forward in semiconductor technology that will shape the future for decades to come," said Intel CEO Lip-Bu Tan. "Our next-gen compute platforms, combined with our leading-edge process technology, manufacturing and advanced packaging capabilities, are catalysts for innovation across our business as we build a new Intel. The United States has always been home to Intel's most advanced R&D, product design and manufacturing - and we are proud to build on this legacy as we expand our domestic operations and bring new innovations to the market." Panther Lake: Scalable AI PC Performance Built on 18A Set to power a broad spectrum of consumer and commercial AI PCs, gaming devices and edge solutions, Intel Core Ultra series 3 processors are the first client system-on-chips (SoCs) built on Intel 18A. Panther Lake introduces a scalable, multi-chiplet architecture that offers partners unprecedented flexibility across form factors, segments and price points. Highlights include: Lunar Lake-level power efficiency and Arrow Lake-class performance. Up to 16 new performance-cores (P-cores) and efficient-cores (E-cores) delivering more than 50% faster CPU performance vs. previous generation. New Intel Arc GPU with up to 12 Xe cores delivering more than 50% faster graphics performance vs. previous generation. Balanced XPU design for next-level AI acceleration with up to 180 Platform TOPS (trillions of operations per second). Beyond the PC, Panther Lake will extend to edge applications including robotics. A new Intel Robotics AI software suite and reference board enables customers with sophisticated AI capabilities to rapidly innovate and develop cost-effective robots using Panther Lake for both controls and AI/perception. Panther Lake will begin ramping high-volume production this year, with the first SKU slated to ship before the end of the year and broad market availability starting January 2026. Clearwater Forest: Efficiency and Scale for the Modern Data Center Clearwater Forest is Intel's next generation E-core processor. Branded Intel Xeon 6+, this processor is the most efficient server processor the company has ever created and is built on Intel 18A. Intel plans to launch Xeon 6+ in the first half of 2026. Highlights include: Up to 288 E-cores. 17% Instructions Per Cycle (IPC) uplift over prior generation. Considerable gains in density, throughput and power efficiency. Tailored for hyperscale data centers, cloud providers, and telcos, Clearwater Forest enables organizations to scale workloads, reduce energy costs, and power more intelligent services. Intel 18A: U.S. Technology Setting New Industry Standards Intel 18A is the first 2-nanometer class node developed and manufactured in the United States, delivering up to 15% better performance per watt and 30% improved chip density compared to Intel 35. The node was developed, qualified for manufacturing and began early production at the company's Oregon location and is now ramping toward high-volume production in Arizona. Key innovations on Intel 18A include: RibbonFET: Intel's first new transistor architecture in over a decade, enabling greater scaling and more efficient switching for improved performance and energy efficiency. PowerVia: A groundbreaking backside power delivery system, enhancing power flow and signal delivery. Additionally, Foveros, Intel's advanced packaging and 3D chip stacking technology, enables the stacking and integration of multiple chiplets into advanced SoC designs, delivering flexibility, scalability and performance at the system level. Intel 18A forms the foundation for at least three upcoming generations of Intel's client and server products. Fab 52: Building on Intel's Five Decades of U.S. R&D and Manufacturing Investment Fab 52 is Intel's fifth high-volume fab at its Ocotillo campus in Chandler, Arizona. This facility produces the most advanced logic chips in the United States and is part of the $100 billion Intel is investing to expand its domestic operations. With advanced R&D and production in Oregon, high-volume fabrication in Arizona, and packaging operations in New Mexico, Intel is uniquely positioned to support key national priorities and provide strategic capacity for Intel Foundry customers. Fab 52 builds upon Intel's 56 years of U.S. R&D and manufacturing advancement and marks a major milestone as the company builds a trusted leading-edge U.S. foundry for the AI era. Source: Intel

Intel's next-gen Panther Lake CPUs launching at CES 2026, available in Q1 2026

TL;DR: Intel's upcoming Panther Lake "Core Ultra Series 3" processors, launching at CES 2026, feature advanced P-cores, E-cores, and Xe3 GPUs with up to 50% improved CPU and graphics performance. Built on Intel's 18A and TSMC's N3E nodes, Panther Lake offers enhanced AI acceleration and significant battery life improvements for laptops and AI PCs. Intel will be launching its next-generation Panther Lake "Core Ultra Series 3" processors at CES 2026, available inside of new laptops in Q1 2026. The new Panther Lake CPUs will feature new technologies including new P-Cores, E-Cores, Xe3 GPU cores, and NPU/IPU. Intel is combining all of this technology into a multi-tile solution that will not just be for laptops, but new AI PCs, and other products. One of the more exciting things about Panther Lake is that the CPU tile is being fabbed in-house on Intel's new 18A process node, while the Xe3 GPU will be fabbed at TSMC on its N3E process node. Intel hosted the media at its recent Tech Tour 2025 event where it teased its new Panther Lake CPUs, but didn't provide concrete details on the new processors. However, in a few months' time at CES 2026, we'll get the full skinny on Panther Lake including performance and more, as well as Intel partners showing off their new Panther Lake-powered laptops, Mini-PCs, and more. Intel says it will begin ramping high-volume production later this year, and that the first Panther Lake SKU will be ready to ship before the end of 2025. As the production ramp continues, the first "full-on availability" for Panther Lake CPUs will kick off in January 2026, with more platforms shipping in Q1 2026. During the Intel Tech Tour 2025 event, the company was asked whether we'll see Panther Lake laptops paired with discrete (higher-end) GPUs, to which the company said it doesn't expect to see Panther Lake CPUs paired with discrete GPUs. But, we have already seen Lunar Lake CPUs paired with discrete GPUs, so it could happen given more time... just not in Q1 2026. On the GPU side of things, Panther Lake ships with Intel's next-gen Xe3 GPU with 12 cores that reportedly offer up to 50% more performance over the previous Lunar Lake generation. We'll have to wait and see just how good the performance is from Xe3, as AMD has been nailing it with its new RDNA 3.5-powered "Strix Halo" APU that offers damn good gaming performance out of an APU. Intel says that Panther Lake has massive battery life improvements that are provided through compute adjustments in the P/E-core configuration, more specifically, which core type handles which workload. Intel says that with the hybrid architecture, which started with Lakefield, refined with Alder Lake, Meteor Lake, and current-gen Lunar Lake, has reached a point where Intel looks "really" confident with Panther Lake. Intel Panther Lake "Core Ultra Series 3" CPU features:

Intel details Panther Lake architecture, its first AI PC platform built on 18A process

The Economic Times was on the ground at ITT covering the eventIntel has officially detailed its upcoming Panther Lake architecture, the next-generation platform that marks the company's first major leap into AI-first computing built on its Intel 18A process. The announcement, made from Intel's newly operational Fab 52 facility in Chandler, Arizona, signals a turning point - not just for Intel's product roadmap, but for U.S. semiconductor manufacturing as a whole. The Intel Core Ultra Series 3, codenamed Panther Lake, represents Intel's first client processor line produced on the Intel 18A node the company's most advanced semiconductor process to date. Intel says the platform is designed from the ground up for AI PCs, integrating new architectures for CPU, GPU, and NPU workloads. Panther Lake chips are already in production and are expected to begin shipping later this year, with full-scale, high-volume production ramping up in Arizona. Intel claims Panther Lake will become the most widely adopted PC platform in the industry once it rolls out commercially. Built to succeed the current Meteor Lake and Lunar Lake series, Panther Lake emphasizes on-device AI performance, better power efficiency, and improved multitasking all critical for the expanding wave of AI-powered software running locally on laptops and desktops. The Intel 18A process (short for 1.8 angstrom) is central to the company's ambition to reclaim process leadership by 2025. It features two key innovations: Together, these enable smaller transistors, faster switching speeds, and greater performance per watt crucial for both AI workloads and next-gen server applications. Intel says 18A is the most advanced semiconductor node ever developed and manufactured in the U.S., reinforcing the company's commitment to local innovation and supply chain resilience. Alongside Panther Lake, Intel also previewed Xeon 6+, codenamed Clearwater Forest -- its first server processor built on the same Intel 18A node. Slated for release in the first half of 2026, Clearwater Forest targets massive performance-per-watt gains for data centers, marking a major step in Intel's roadmap to compete more aggressively with AMD and ARM-based cloud processors. Intel has hinted that the Xeon 6+ lineup will deliver double-digit performance improvements and significantly lower power draw a key consideration as hyperscalers and enterprises push for sustainable, AI-ready data center infrastructure. Intel's Fab 52 in Chandler is now fully operational and is expected to reach high-volume production later this year using the Intel 18A process. It joins Fab 62, also located in Arizona, as part of Intel's multibillion-dollar investment to expand advanced chip manufacturing capacity in the U.S. The site will produce not only Panther Lake and Clearwater Forest, but also future product generations on the 18A and beyond. Intel's investments in Arizona align with the broader U.S. CHIPS Act objectives to strengthen domestic semiconductor production and reduce reliance on overseas fabs. Intel CEO Lip-Bu Tan called this transition "an exciting new era of computing," highlighting how the company's advances in process technology, AI integration, and domestic production will shape the next decade of computing. "Our next-gen compute platforms, combined with our leading-edge process technology, manufacturing and advanced packaging capabilities, are catalysts for innovation across our business as we build a new Intel," Tan said. With Panther Lake and the 18A process, Intel appears intent on reclaiming its position at the forefront of semiconductor innovation not only competing with rivals like TSMC and AMD but also anchoring the next chapter of AI computing firmly on U.S. soil

Intel Unveils New Processors To Power AI, Gaming Devices - Intel (NASDAQ:INTC)

Intel Inc. (NASDAQ:INTC) has unveiled the architectural details of its next-generation Intel Core Ultra series 3 processors, code-named Panther Lake, which are set to begin shipping later this year. It marks the U.S. chipmaker's product debut under Intel 18A, its most advanced semiconductor process ever developed and manufactured in the country. Panther Lake will power AI PCs, gaming devices, and edge solutions, featuring a scalable multi-chiplet architecture that provides flexibility across different form factors, segments, and price points. Also Read: Taiwan Semiconductor Shuts Down Rumors Of Intel Deal Intel previewed the Xeon 6+ (code-named Clearwater Forest), its first server processor based on Intel 18A, which is expected to launch in the first half of 2026. Intel will manufacture Panther Lake and Clearwater Forest at Fab 52 in Chandler, Arizona, marking a significant milestone in Intel's strategy to enhance American tech and manufacturing leadership. The Intel 18A process offers up to 15% better performance per watt and a 30% improvement in chip density compared to Intel's previous 35nm node. Innovations like RibbonFET, a new transistor architecture, and PowerVia, a novel backside power delivery system, drive these advancements. Intel's Foveros packaging technology also enables advanced system-on-chip (SoC) designs by stacking multiple chiplets. Intel is ramping up production of Panther Lake, with the first shipments expected by the end of 2025, followed by broader availability in January 2026. The Clearwater Forest processors, designed for hyperscale data centers and cloud providers, are scheduled to launch in 2026, offering significant gains in efficiency and scalability. Shares of Intel surged nearly 40% in September and have risen around 90% year-to-date. This boost came following news of significant investments in the company from the U.S. government, Nvidia Corp. (NASDAQ:NVDA), and SoftBank (OTC:SFTBY). Price Action: INTC stock is up 0.96% at $37.81 at last check on Thursday. Read Next: Trump's Chip Plan Could Mean Big Trouble For Apple And Dell Photo: Tada Images / Shutterstock INTCIntel Corp$37.10-0.87%OverviewNVDANVIDIA Corp$193.322.23%SFTBYSoftBank Group Corp$74.988.46%Market News and Data brought to you by Benzinga APIs

Inside Intel's Panther Lake: Future of AI, Gaming & XPUs Explained | TechPulse

At the Intel Tech Tour in Phoenix, we sat down with two of Intel's top minds to decode what the upcoming Panther Lake processors mean for the future of AI, gaming, and everyday computing.Featuring:Daniel Rogers, VP & GM, Products at Intel on how Panther Lake will redefine AI performance, productivity, and the rise of XPUs in modern computing.Tom Peterson, Fellow at Intel - breaking down the basics of CPUs, TOPS, and the difference between NPU, CPU, and GPU, while also sharing what gamers and everyday users can expect from Panther Lake.From AI-driven workflows to next-gen gaming performance, this episode dives deep into how Intel's Panther Lake architecture is shaping the next leap in PC innovation.

Inside Intel's Panther lake reveal in Phoenix | Next-Gen AI for PC & Edge | TechPulse

The Economic Times TechPulse team went to Phoenix for the Intel Tech Tour, where Intel unveiled its upcoming Panther Lake chip designed to power the next wave of AI-driven computing.From Adobe Premiere Pro's new smart tools and a real-time teleprompter demo to AI-enhanced gaming and Edge innovations like robotics and order accuracy algorithms, we got an exclusive hands-on look at what Intel's future of performance and intelligence looks like.Watch now for a deep dive into how Intel is reimagining both PC experiences and Edge AI applications with Panther Lake.

Panther Lake explained: How new Intel chip will impact laptops in 2026

2026 PCs promise smarter multitasking, better battery life, and richer media There's a version of PC progress where you get a bigger number on a spec sheet and not much else - Panther Lake or Intel Core Ultra series 3 doesn't seem to be anything like that. It's Intel's attempt to make next year's laptops and small desktops feel faster, quieter, and smarter in the moments you actually notice: 20 Chrome tabs, a Zoom call, Lightroom exporting in the background, fans barely audible. The trick isn't one monster core, according to Intel, but a platform that reshuffles the whole deck - CPU, GPU, AI, memory, even the camera pipeline - to scale performance without torching your battery. Microarchitecture highlights of Intel Panther Lake According to the company, the Panther Lake or Intel Core Ultra series 3 family of chips is said to deliver Lunar Lake-level power efficiency and Arrow Lake-class performance. Panther Lake or Intel Core Ultra series 3 does that by utilizing a next-generation hybrid core architecture, with two distinct core types, both designed and optimized for the Intel 18A process node. * P-core (Performance-core): The new P-core microarchitecture is called Cougar Cove. * E-core (Efficiency-core): The E-core and low-power E-core microarchitectures are called Darkmont. Disaggregated tiles (compute, GPU, and I/O) live on a common Foveros 2.5D package and talk over Intel's Scalable Fabric Gen 2. What this means is that Intel can mix-and-match parts for thin-and-lights, creator rigs, or compact gaming boxes without breaking coherency or software - like Lego, but for silicon. It's flexibility first, performance second, and power efficiency to the max - at least that's what Intel claims for now. Panther Lake brings fresh core IP - Cougar Cove (P-core) and Darkmont (E-core + LP E-core) - built for Intel 18A, with efficiency and IPC uplift as the headline goals. Two cluster styles match work to the right silicon: a Performance cluster that puts P- and E-cores behind a shared L3 for fast throughput, and an Efficiency cluster of E-cores with its own memory hooks for lean, scalable power draw. Darkmont's E-core cluster gets 4 MB of shared L2, giving the scheduler (now with an enhanced Intel Thread Director) more room to keep "busy but realistic" multitasking on the efficient cores - so responsiveness stays high while the fans stay low. Also read: Intel unveils Panther Lake: First AI PC chip built on 18A node A dedicated 8 MB Memory-Side Cache sits at the SoC level, acting like a fast valet for frequently accessed data - from CPU bursts to I/O engines - cutting trips to DRAM, improving effective latency/bandwidth, and trimming power spikes. In human terms that translates to snappier app switches and fewer stuttered moments when everything piles on at once when you're working on your laptop. Three reference builds available Intel's own examples show how it scales the same package: an 8-core part for mainstream laptops, a 16-core for heavier multitaskers, and a 16-core paired with a larger GPU tile for creator/gaming machines. All share the same fabric and I/O tile, so features like PCIe Gen5 and Thunderbolt don't go missing when you move down the stack. * 8-core: 4 P-cores + 4 LP E-cores, up to LPDDR5X-6800/DDR5-6400; GPU tile with 4 Xe-cores/4 RT units; up to 12 PCIe lanes, 4x Thunderbolt 4, Wi-Fi 7 (R2), BT 6.0. * 16-core: 4 P-cores + 4 LP E-cores + 8 E-cores; up to LPDDR5X-8533/DDR5-7200; GPU: 4 Xe-cores; up to 20 PCIe lanes (12 Gen5 + 8 Gen4), 4x Thunderbolt 4, Wi-Fi 7 (R2), BT 6.0. * 16-core / 12Xe: same CPU, bigger GPU tile (12 Xe-cores/12 RT units), memory up to LPDDR5X-9600; I/O includes 12 lanes (4 Gen5 + 8 Gen4) and 4x Thunderbolt 4. Intel flags TB5 support across the family as well. Across the range, Intel claims you'll see up to 12 lanes of PCIe Gen5, integrated Wi-Fi 7 (R2), dual Bluetooth 6.0, and as many as six integrated Thunderbolt 4 ports depending on model and OEM choices. Graphics, AI and IO are enhanced The GPU moves to its own tile and debuts the Xe3 architecture - up to 12 Xe-cores and 12 ray-tracing units with 16 MB L2 on the tile - so premium systems can scale graphics without roasting the CPU/NPU neighborhood. Intel also cites up to 120 TOPS worth of AI/graphics math on the GPU side, which matters for the growing wave of local AI effects (upscaling, denoising, background removal tasks). On the neural side, NPU 5 targets up to 50 TOPS, with a >40% TOPS/area bump over Lunar Lake and ~3.8x more performance than Arrow Lake-H chips. It's aimed squarely at on-device AI that doesn't spin your fans like a ceiling vent. What this means is that in 2026, expect snappier copilots on your laptops, better live captions/translations, and smarter photo/video tools that don't chew through your battery. And because your webcam is now your office, Intel's IPU 7.5 brings hardware-accelerated staggered HDR (Intel pegs ~1.5W lower power consumption than Lunar Lake), AI noise reduction, local tone mapping, support for up to three concurrent cameras, and capture up to 16MP still photos or 120fps slow-mo videos - quality-of-life upgrades you will actually notice on calls and creator workflows on laptops releasing in 2026. Thread Director now spans P-, E-, and LP E-cores with tuned models per core type, while more power decisions move into the SoC hardware layer. The aim is simple but long overdue here, which is to provide a consistent behaviour across Windows, ChromeOS, and Linux, and across battery vs. plugged-in states - historically a sore spot for laptops. With I/O centralized on a platform controller tile, OEMs can light up the ports you actually want - Thunderbolt 4 today (family support for TB5), PCIe Gen5 for GPU docks and scorching external storage, plus integrated Wi-Fi 7 (R2) and dual Bluetooth 6.0 - without respinning the CPU/GPU tiles. What this means for laptops in 2026 * Thinner, quieter, longer-lasting laptops: The combination of efficiency-first cores, Memory-Side Cache, and smarter scheduling is designed to keep you in lower power states more often - even while doing "a lot of little things at once." Expect fewer fan ramp-ups and more "all-day" that actually survives your day. * AI that's useful, not gimmicky: With ~50 TOPS on the NPU and up to 120 TOPS available on the GPU tile, expect local copilots, background effects, transcription, and image enhancement that don't demand the cloud (or your charger). That's privacy, latency, and battery life all getting a big boost -- at least on paper. * For creator laptops: They should target the 16-core with 12Xe GPU tile. Pair it with fast LPDDR5X RAM of between 16GB to 32GB. Intel Xe3's media/display engine covers modern codecs and higher-end workflows, while the beefier GPU tile option gives you more headroom for timeline scrubs and encodes - without needing a discrete GPU in every chassis.. Thunderbolt + PCIe Gen5 lanes make external RAID and eGPU setups practical. * For compact gamers: The 12Xe GPU tile systems should comfortably handle 1080p/1440p with modern upscalers; TB4/TB5 plus PCIe Gen5 let you dock a bigger GPU at home if you want to swing between portable and powerful. * Better webcams by default: IPU 7.5's smarter, lower-power pipeline means clearer, cleaner video in bad lighting and less heat during long calls. Multi-cam support opens up interesting creator/meeting setups on mainstream machines. * Port and wireless sanity: Wi-Fi 7 (R2) is standard; Thunderbolt 4 is common; TB5 exists in the family; PCIe Gen5 shows up where it counts. External GPU docks and ultra-fast SSDs won't feel like science projects. In a nutshell, Panther Lake or upcoming Intel Core Ultra series 3's efficiency gains vs. Lunar Lake & Arrow Lake chips of the past couple of years can be summed up as follows: * CPU: Greater than 10% more Single-Thread (ST) performance at similar power vs. Lunar Lake, and 30% lower power at similar Multi-Thread (MT) performance vs. Arrow Lake. * SoC: Up to 10% lower power vs. Lunar Lake and 40% lower power vs. Arrow Lake. * GPU: Greater than 50% more performance vs. Lunar Lake and Arrow Lake. If Intel's claims are true and they meet their timelines in shipping devices by early next year, the 2026 PC will feel different in the ways that matter - snappier, steadier, and strangely quiet while doing more. That's what Panther Lake's going for, according to Intel, to deliver AI-PC excellence at scale, without giving up on endurance.

Intel unveils Panther Lake: First AI PC chip built on 18A node

Intel took the wraps off its third-generation of AI PC chip, Intel Core Ultra series 3 - codenamed Panther Lake. It's Intel's next processor update to feature in laptops starting in 2026, and the first to be built on its new 18A semiconductor process. Set to power a wide range of AI-enabled PCs and edge devices, Intel Core Ultra series 3 or Panther Lake chips will begin shipping later this year, with high-volume production ramping at Intel's newest manufacturing facility, Fab 52, in Chandler, Arizona. Intel also offered an early look at Clearwater Forest - its first Xeon server processor based on 18A - scheduled to launch in the first half of 2026. These announcements were made as part of the company's 2025 Tech Tour, which Digit attended, marking a key milestone in Intel's multi-year roadmap to regain leadership in process technology and semiconductor manufacturing. Both chips are being produced at Fab 52, Intel's most advanced factory to date and part of its $100 billion US expansion strategy. "We are entering an exciting new era of computing, made possible by great leaps forward in semiconductor technology that will shape the future for decades to come," said Intel CEO Lip-Bu Tan. "Our next-gen compute platforms, combined with our leading-edge process technology, manufacturing and advanced packaging capabilities, are catalysts for innovation across our business as we build a new Intel." Panther Lake: Not just another SoC The star of the show is Panther Lake, Intel's first client processor built on its 18A node - a 2-nanometer-class manufacturing process developed and produced entirely in the United States. Panther Lake is the new anchor of the Intel Core Ultra 3 series of SoCs The chip features a new scalable multi-chiplet architecture and is designed for a broad spectrum of AI PCs, gaming systems, and edge computing devices. This is no minor spec bump, but an architectural overhaul, according to Intel. We're looking at up to 16 performance and efficiency cores with more than 50% CPU performance and graphics gains over the previous generation Lunar Lake chips, alongside up to 180 TOPS of AI compute acceleration packed into a balanced XPU layout - which also includes a new integrated Intel Arc GPU with up to 12 Xe cores. According to the company, the Intel Core Ultra series 3 family of chips is said to deliver Lunar Lake-level power efficiency and Arrow Lake-class performance. Because Panther Lake isn't just for spreadsheets and gaming rigs. Intel is positioning it as the foundational silicon for a new breed of AI PCs - a phrase that, not long ago, felt like industry vaporware. But with Apple, Qualcomm, AMD and Nvidia all doubling down on on-device AI, Intel's hand was forced. With Panther Lake and Intel Core Ultra series 3, it looks like it's finally answering. Also read: Panther Lake explained: How new Intel chip will impact laptops in 2026 Beyond consumer devices, Panther Lake will also target industrial and robotics applications. Intel announced a new Robotics AI software suite and reference board (robotics development kit) to support partners building AI-enabled robots with the new chip. According to Intel, high-volume manufacturing of Panther Lake has begun, with initial SKUs of Intel Core Ultra series 3 processors shipping by year-end and broad availability starting January 2026. Clearwater Forest and the Xeon 6+ era Intel also previewed Clearwater Forest, the upcoming data center platform based on Intel 18A process, expected to launch later in 2026. Branded under the Xeon 6+ family, the processor features up to 288 E-cores and a 17% uplift in instructions-per-cycle (IPC) compared to the prior generation. Designed for hyperscale and cloud environments, Clearwater Forest promises substantial gains in throughput, density, and power efficiency, helping data center operators scale workloads while reducing energy consumption. Inside the 18A node Intel 18A is Intel's first 2-nanometer-class process node, representing a major leap in the company's manufacturing roadmap. The "A" stands for angstrom, indicating a shift in Intel's naming convention to reflect advanced transistor scaling beyond traditional nanometer metrics. Compared to TSMC's 3nm (N3) and upcoming 2nm (N2) nodes, Intel 18A introduces two defining innovations: RibbonFET, Intel's gate-all-around transistor design (its first new transistor architecture in over a decade), and PowerVia, a backside power delivery system that routes power through the rear of the chip - freeing up the front for improved signal routing and reducing voltage droop. Intel claims these combined advances deliver up to 15% better performance per watt and 30% higher chip density compared to Intel 3, putting 18A on competitive, if not leading, footing with TSMC's N2 process in terms of transistor technology. Crucially, 18A is also being manufactured domestically in the United States., giving it geopolitical and supply-chain relevance that TSMC - still reliant on Taiwan for leading-edge production - currently lacks, according to Intel. Intel says 18A will serve as the foundation for at least three future generations of client and server products, with Panther Lake and Clearwater Forest setting the tone for what the company calls a "new era" of compute innovation that's made in America for the world. Located in Chandler, Arizona, Fab 52 is Intel's fifth high-volume fab on its Ocotillo campus and will serve as the lead manufacturing site for Intel 18A. The node was first qualified for manufacturing in Oregon and is now ramping up toward full production in Arizona. The fab represents a key pillar in Intel's ambition to build a leading-edge, US-based foundry for both internal products and external customers. "We are proud to build on this legacy as we expand our domestic operations," said Tan, referencing Intel's 56-year history of United States R&D and manufacturing. Intel's advanced packaging technology, Foveros, is also being used to integrate chiplets into system-level SoC designs, offering flexibility for future product scaling.