Snapdragon X2 Elite vs Intel Core Ultra 300: Laptop Chips Compared

Qualcomm's Snapdragon X2 Elite and X2 Elite Extreme and Intel's Core Ultra 300 ("Panther Lake") are the two high-end chip platforms now shipping in flagship Windows laptops: Qualcomm's 3nm Oryon-based Arm chips top out at 18 cores, 80 TOPS of NPU AI performance and 5.0GHz boost clocks, while Intel's 18A-built Panther Lake tops out at 16 cores, a 50 TOPS NPU paired with a much larger Arc B390 GPU, and up to 27 hours of claimed battery life on its flagship X9 SKU. Neither company has published independent third-party benchmarks against the other — the figures below are each company's own published specifications and performance claims, laid out side by side so you can compare the numbers that actually matter for a laptop purchase.

What Snapdragon X2 Elite and Core Ultra 300 actually are

Snapdragon X2 Elite and Snapdragon X2 Elite Extreme are Qualcomm's second generation of Arm-based PC chips, built around a third-generation Oryon CPU core and announced on September 25, 2025 at Qualcomm's Snapdragon Summit in Maui, Hawaii. They are the direct successor to the original Snapdragon X Elite and X Plus chips that launched the Copilot+ PC category in 2024. Qualcomm ships three SKUs in the family: the flagship X2 Elite Extreme (part number X2E-96-100), a second 18-core X2 Elite (X2E-88-100), and a lower 12-core X2 Elite (X2E-80-100), according to Qualcomm's own Snapdragon X2 Elite product brief, detailed in Qualcomm's September 2025 Snapdragon Summit announcement.

Intel Core Ultra Series 3, codenamed Panther Lake, is Intel's answer: the company's first client processor built on its own Intel 18A manufacturing process, using RibbonFET transistors and PowerVia backside power delivery. Intel launched the lineup at CES 2026 in Las Vegas on January 5, with laptops going on sale globally on January 27, 2026, and edge/embedded variants following in the second quarter of the year. Unlike Qualcomm's compact three-SKU stack, Intel ships a wide range of Core Ultra 5, Core Ultra 7, Core Ultra 9, and new high-end Core Ultra X7/X9 parts, letting laptop makers dial in cores, graphics and price across a much broader lineup — a strategy similar to how Intel has long segmented its desktop parts, as we covered in our Core Ultra 7 270K+ vs Ryzen 7 9800X3D desktop comparison.

CPU cores and clock speeds compared

Qualcomm's Snapdragon X2 Elite Extreme pairs 12 "Prime" cores (the high-performance Oryon cluster) with 6 "Performance" cores. The Prime cluster runs at a 4.4GHz multi-core frequency and boosts to 5.0GHz on single- and dual-core workloads; the Performance cluster tops out at 3.6GHz. Total cache across the chip is 53MB. The step-down 18-core X2 Elite (X2E-88-100) uses the same 12+6 layout but at lower clocks — 4.0GHz Prime multi-core, 4.7GHz boost — while the smallest X2 Elite (X2E-80-100) drops to 6 Prime and 6 Performance cores with 34MB of cache, according to Qualcomm's product brief.

Intel's Panther Lake uses a different core recipe entirely. Its "Cougar Cove" performance cores and "Darkmont" efficiency cores replace the Lunar Lake generation's Lion Cove/Skymont pairing, and the top-tier X9 and X7 SKUs add a third core type — low-power E-cores — for a maximum of 16 total cores and threads. According to Intel's own Core Ultra Series 3 Quick Reference Guide, the flagship Core Ultra X9 388H hits a 5.1GHz P-core max turbo with an 18MB last-level cache, the X7 368H reaches 5.0GHz, and the mainstream Core Ultra 9 386H and Core Ultra 7 366H/356H top out around 4.7–4.9GHz. Entry Core Ultra 5 SKUs (338H down to 322) scale from 6 to 12 total cores with clocks between 4.4GHz and 4.7GHz.

In short: Qualcomm's Oryon design leans on more total cores (18 vs 16) built from just two core types, while Intel leans on a three-tier core hierarchy and slightly higher single-core clocks at the very top of its stack. Intel says the Core Ultra X9 388H delivers up to 60% better multithreaded performance than its own previous-generation Core Ultra 9 288V ("Lunar Lake") chip, measured with Cinebench 2024 Multi-Core at 25W — a generational claim, not a head-to-head test against Qualcomm's silicon.

GPU and graphics performance

Qualcomm's Adreno GPU in the X2 Elite Extreme is the X2-90, clocked up to 1.85GHz, supporting DirectX 12.2 Ultimate, Vulkan 1.4 and OpenCL 3.0; the two lower X2 Elite SKUs use the same X2-90 die at 1.70GHz or a smaller X2-85 part, also at 1.70GHz. Qualcomm says the Adreno architecture in this generation delivers roughly 2.3x the performance-per-watt of its prior-generation GPU, though it has not published an absolute TFLOPS figure.

Intel's graphics story is more segmented by SKU. The X9 and X7 tiers get a full Intel Arc B390 GPU with 12 Xe3 cores running up to 2.5GHz — the same "Arc" branding and Xe3 architecture Intel uses on its desktop and mobile discrete cards. Mid-tier Core Ultra 9/7 chips step down to an unbranded "Intel Graphics" configuration with 4 Xe3 cores at similar clocks, and the entry Core Ultra 5 338H uses a 10-core Arc B370. Intel says its top Arc B390 configuration delivers over 77% faster gaming performance than the Core Ultra 9 288V it replaces, measured as a geometric mean across 45 game titles at 1080p with upscaling enabled where supported. That is a large generational jump specifically because Panther Lake's top SKUs get roughly 3x more graphics cores (12 Xe3 vs 4 Xe2 cores) than the outgoing Lunar Lake flagship.

Because Intel varies GPU core count so widely by SKU — from 2 Xe3 cores on the cheapest Core Ultra 5 up to 12 on the X9/X7 — the graphics gap between a Snapdragon X2 Elite laptop and an Intel-based one will depend heavily on which specific Core Ultra 300 chip is inside the machine you're looking at, not just the "Core Ultra 300" branding on the box.

NPU and on-device AI TOPS

Both platforms clear the 40 TOPS NPU bar Microsoft set for Copilot+ PCs, but Qualcomm's Hexagon NPU is rated noticeably higher: all three Snapdragon X2 Elite SKUs carry an 80 TOPS (INT8) Hexagon NPU with a 64-bit architecture, plus a separate dual micro-NPU on the always-sensing hub for low-power tasks like wake word and presence detection, per Qualcomm's product brief.

Intel's fourth-generation NPU in Panther Lake is rated up to 50 TOPS on its highest-end SKUs (X9, X7 and the 366H/356H tier), scaling down to 46–49 TOPS on Core Ultra 5 parts, according to Intel's Quick Reference Guide. Intel also publishes a combined "platform AI TOPS" figure that adds NPU, GPU and CPU AI throughput together — Intel has cited up to 180 platform TOPS on Panther Lake, of which roughly 120 comes from the Arc B390 GPU's matrix engines and 50 from the NPU itself. That combined number is not directly comparable to Qualcomm's NPU-only 80 TOPS figure, since it folds in GPU compute that Windows' Copilot+ NPU requirement doesn't count. For workloads that specifically route through the NPU — background blur, Windows Studio Effects, on-device Copilot+ features — Qualcomm's 80 TOPS rating is the higher of the two chips' dedicated NPU numbers; for GPU-accelerated AI workloads like local image generation, Intel's larger Arc B390 has more raw compute to draw on.

Manufacturing process node

Snapdragon X2 Elite and X2 Elite Extreme are built on a 3nm process node, per Qualcomm's official spec sheet (Qualcomm does not name the specific foundry process in its public product brief). Intel Core Ultra Series 3 is built on Intel 18A, Intel's own next-generation node that introduces RibbonFET gate-all-around transistors and PowerVia backside power delivery — and Panther Lake is explicitly the first Intel client processor manufactured on it, a milestone Intel has framed as central to closing its long-running process-technology gap with rival foundries.

Memory support and bandwidth

Every Snapdragon X2 Elite SKU uses LPDDR5x memory at a 9523 MT/s transfer rate. The flagship X2 Elite Extreme uses a wider 192-bit memory bus for 228GB/s of bandwidth and supports 128GB+ of configured memory, while the two other X2 Elite SKUs use a 128-bit bus capped at 152GB/s. Because Qualcomm's memory is packaged on-die rather than in swappable modules, capacity and speed are fixed by the OEM at build time.

Intel's Panther Lake supports LPDDR5X up to 9600 MT/s — which Intel calls an industry first for that memory speed — and, on select SKUs, DDR5 up to 7200 MT/s or LPCAMM2 modules, also up to 9600 MT/s. Maximum capacity is 96GB with LPDDR5/X or 128GB with DDR5, according to Intel's Quick Reference Guide. That LPCAMM2 support is notable: unlike Qualcomm's fixed on-package memory, some Panther Lake laptops built around DDR5 or LPCAMM2 configurations can theoretically be reconfigured or upgraded after purchase, though many thin-and-light designs will still solder LPDDR5X for power and space reasons. For more on how next-generation memory standards are evolving industry-wide, see our explainer on DDR6 RAM and how it compares to DDR5 — a technology neither of these laptop chip families uses yet, but one that gives useful context for where laptop memory bandwidth is headed next.

Battery life and efficiency implications

Efficiency is the headline pitch for both chip families, but the two companies frame it differently. Qualcomm's Arm-based Oryon architecture has built its reputation on efficiency since the original Snapdragon X Elite, and Qualcomm says the X2 generation extends that lead: the company states the X2 Elite family delivers up to 31% higher CPU performance than the original Snapdragon X Elite at the same power draw, or the same performance using up to 43% less power, plus a roughly 2.3x gain in GPU performance-per-watt.

Intel, migrating an x86 architecture onto its new 18A node, makes its own specific real-world claim: up to 27 hours of battery life on a Lenovo reference laptop running a Netflix streaming test, and up to 17 hours in an office-productivity battery test, on the flagship Core Ultra X9 388H. Both companies' numbers come from internal testing on reference or partner hardware rather than independent lab results, so real-world battery life in any specific laptop will vary with screen size, resolution, battery capacity and software — but the efficiency architecture underneath each chip (Arm's inherently lower per-instruction power draw on Qualcomm's side, versus Intel's new 18A process and three-tier core hierarchy on its own) is the reason both companies are willing to make aggressive battery-life claims for the first time in years.

Connectivity and other platform features

Both platforms have converged on largely similar connectivity. Snapdragon X2 Elite pairs with Qualcomm's FastConnect 7800 system for Wi-Fi 7 and Bluetooth 5.4, plus an integrated Snapdragon X75 5G modem rated for up to 10Gbps peak download speeds — a built-in cellular option Intel's chips don't offer natively. Panther Lake instead integrates Intel Wi-Fi 7 (5Gig) R2 with dual Bluetooth Core 6.0, along with up to 4 integrated Thunderbolt 4 ports and, on higher SKUs, discrete Thunderbolt 5 support — a wired connectivity standard Qualcomm's platform doesn't match. Both chips support USB4-class high-speed I/O and up to three external displays. If you're shopping for a new laptop and want to understand what the "Wi-Fi 7" spec sheet line actually gets you on either platform, we've broken it down separately in our Wi-Fi 7 explainer covering MLO and real-world speeds.

On the software side, Windows on Arm compatibility has historically been Qualcomm's weak point versus Intel's native x86 support, though Microsoft and Qualcomm have continued expanding native and emulated app coverage since the original Snapdragon X Elite launch. Buyers considering either platform for professional software, older games, or niche Windows utilities should still check app compatibility for their specific use case before choosing Arm over x86, or vice versa.

Snapdragon X2 Elite vs Core Ultra 300: full spec comparison

The table below lines up the flagship SKU from each family — Snapdragon X2 Elite Extreme and Intel Core Ultra X9 388H — alongside Qualcomm's mid-tier X2 Elite and Intel's mid-tier Core Ultra 7 366H, using each company's own published specifications.

Source data: Qualcomm's Snapdragon X2 Elite product brief and Intel's Core Ultra Series 3 Quick Reference Guide, both cited in full below.

Which chip should you buy?

If your priority is the single highest dedicated NPU TOPS number, the widest core count, or you specifically want cellular connectivity built into the chip, Snapdragon X2 Elite (particularly the Extreme tier) is the stronger spec sheet on paper — 80 TOPS beats every individual Core Ultra 300 NPU, and its 18-core Oryon design plus 3nm process is built around Qualcomm's efficiency-first Arm pedigree. If your priority is graphics performance for gaming or GPU-accelerated creative work, native x86 app and game compatibility, Thunderbolt 5 peripherals, or a wider range of price points and configurations across Core Ultra 5, 7, 9 and X7/X9 tiers, Intel's Panther Lake lineup — especially the Arc B390-equipped X9 and X7 SKUs — is the more flexible platform, backed by Intel's own claims of up to 77% faster gaming performance and 27 hours of battery life on its flagship configuration.

Neither company's performance claims have been independently verified against the other's chip, so treat every percentage figure in this piece as a vendor claim, not a lab result — and if you're comparing specific laptop models rather than the chips in the abstract, check what specific SKU (not just "Snapdragon X2 Elite" or "Core Ultra 300") is actually inside the configuration you're buying, since both families span a wide range of core counts, clocks and GPU tiers under one marketing name. Qualcomm and Intel both publish their full SKU listings and marketing pages directly — see Qualcomm's own Snapdragon X2 Elite product page and Intel's Core Ultra Series 3 product page — before comparing specific laptop listings against either spec sheet above.

Written byMara Lindqvist

Mara Lindqvist edits the hardware desk. She covers graphics cards, processors, memory and storage, the foundries and chip designers behind them, and what the numbers on a spec sheet mean for people choosing a PC. Specifications in her stories come from manufacturer spec pages and datasheets.

More from Mara LindqvistAll Hardware coverage