Intel Unveils Next-Gen Arrow Lake CPUs: Balancing AI and Efficiency for Desktop and Laptop Markets

Mobile Arrow Lake Core Ultra 200H and 200HX processors coming early next year

Serving tech enthusiasts for over 25 years. TechSpot means tech analysis and advice you can trust. Something to look forward to: Intel has introduced two Core Ultra 200 CPU lineups across desktops and mainstream laptops this year. That leaves the high-end laptop processors for mobile gaming and content creation devices, about which the company has said little. At the very least, we now know that they won't launch in 2024. Toward the end of its presentation introducing the Arrow Lake S Core Ultra 200K and KF desktop processors this week, Intel confirmed plans to reveal the H and HX chips for high-end laptops in the first quarter of 2025. Details on the upcoming processors are scarce, but specs for the HX CPUs leaked recently. The final slide in Intel's Arrow Lake S brief describes Core Ultra 200H and HX as the company's first mobile AI PC for gamers and creators. They aim to offer high-end laptop performance with improved power management, echoing the efficiency gains Intel has touted with Core Ultra 200. A report from last month indicated that six HX chips align closely with their desktop counterparts' core counts with slightly lower clock speeds. Regarding integrated graphics, most will feature 64 Xe2 Battlemage EU cores at clock speeds ranging between 1.8GHz and 2GHz. Less is known about the Core Ultra 200H processors, but they might resemble the 14th-generation Raptor Lake 14000H CPUs. Arrow Lake Halo remains another uncertainty on Intel's roadmap. Supposedly positioned against AMD's ultra high-end mobile SoC, Strix Point Halo, Intel's rumored enhanced iGPU features 320 EU cores with a 14-core processor. With the newly unveiled Core Ultra 200K and KF CPUs, Intel is comprehensively revising its architecture. The new lineup removes hyperthreading, adds an NPU for on-board generative AI, transitions to the company's first multi-tile desktop layout, and is Intel's first desktop chip series manufactured mainly by TSMC. Although the desktop Arrow Lake products might not break performance records, they're more power-efficient and carry reasonable price tags. Intel's next generation of laptop processors - codenamed Panther Lake (Core Ultra 300) - is expected to follow in the second half of 2025. Early leaks describe five H-class chips with core counts ranging from eight to 18. The lineup marks the debut of the company's Xe3 Celestial graphics architecture and 18A (1.8nm) semiconductor node. Intel CEO Pat Gelsinger promised that Panther Lake will also significantly boost AI performance.

Intel's Arrow Lake high-performance chips are less about AI and more about gaming

If you'd rather avoid generative AI in favor of pure computing power, Intel may just have the right chip for your next gaming laptop. Thursday, Intel released the next chips hitting the Intel Core Ultra 200 family, the 200S and 200HX desktop chips. The company also offered some early details on the 200H mobile processors expected early next year. The "Arrow Lake" Intel Core Ultra 200S, Intel Core Ultra 200HX, and Intel Core Ultra 200H processors favor high power efficiency, stronger multi-thread performance, and better graphics. The Intel Core Ultra 200S chips are intended for desktops and will launch this month, but we do have info on the 200HX, and 200H processors that will becoming early next year. These chips will likely wind up in the next wave of gaming laptops and workstations, so let's get into what Intel's "Arrow Lake" can do for laptops. The Intel Core Ultra 200H processor features integrated Intel WiFi 7 support, integrated Thunderbolt 4 support, Bluetooth 5.4 support, PCIe Gen 5 connectivity, DDR5 support, all in Foveros 3D packaging. Intel's Platform Marketing Manager Greg Boots referred to the Core Ultra 200H chip as a "standout platform for those who need premium portable graphics and AI capabilities" during a hardware briefing on Arrow Lake at IFA Berlin 2024. The Intel Core Ultra 200H uses the same Lion Cove performance P-cores and Skymont efficiency E-cores as the other Intel Core Ultra 200 series chips. The exact core count hasn't been confirmed, though we will likely get more details about the Intel Core Ultra 200H chip in the coming months. The main standout of the Intel Core Ultra 200H mobile processor family is its new Xe with XMX integrated GPU. This graphics tile offers four times the AI throughput of the Intel Core Ultra 100 "Meteor Lake" processor and supports the BFP16, FP16, int8, and int4 AI datatypes. It also offers twice the ray tracing capacity of the Meteor Lake Xe GPU with two traversal pipelines for RTU ray tracing workloads, and also has twice the cache size for an 8MB L2 cache. The Intel Core Ultra 200H features an Intel AI Boost 3 NPU which offers up to 13 TOPS (trillions of operations per second) of AI performance. The entire chipset offers up to 99 TOPS of AI computing power, with up to 77 TOPS on the GPU, 13 TOPS on the NPU, and 9 TOPS on the CPU. A representative for Intel tells Laptop Mag that the Intel Core Ultra 200H chips are expected to launch around CES 2025 in January. The Intel Core Ultra 200HX processor features integrated Intel WiFi 6 support, integrated Thunderbolt 4 support, Bluetooth 5.3 support, PCIe Gen 5 connectivity, DDR5 RAM support, all in Foveros 3D packaging. Manufacturers can upgrade the Core Ultra 200HX chipset to support Thunderbolt 5 and BlueTooth 5.4. Intel's VP of the Client Computing Group Josh Newman claimed Arrow Lake will "bring the AI PC to the desktop and mobile markets" when discussing the chipsets at IFA Berlin 2024. With 16 Skymont efficiency E-cores and 8 Lion Cove performance P-cores, the Intel Core Ultra 200HX processor offers a multi-thread performance increase of about 10% over the Intel Core i9-14900K chip. The Intel Core Ultra 200HX also sees an increase of up to 30% more power efficiency compared to the Intel Core i9-14900K desktop chip. Intel has made it easier to overclock the Intel Core Ultra 200HX chip with the new Intel eXtreme Tuning Utility with AI tool that offers granular core clock controls, a dual base clock, tile-2-tile and fabric overclocking, DLVR bypass, and AI recommendations for helping set the best clock speeds to optimize your CPU. The Intel Core Ultra 200HX also features a larger GPU than the Intel 14th-gen chipsets. The new Xe GPU offers twice the graphics compute resources with higher clock frequencies, better architecture efficiency, and full Direct X 12 Ultimate support built in at the silicon level. The new Xe GPU features 4 Xe cores with 64 Vector engines, 1 Geometry pipeline, 4 samplers, 2 pixel backends, 4 ray tracing units, and a 4MB L2 cache. Lastly, the Intel Core Ultra 200HX offers an Intel AI Boost 3 NPU with up to 13 TOPS of AI performance, with a total of up to 36 TOPS across the platform. Intel rates the "Arrow Lake" chips with up to 8 TOPS on the GPU, 13 TOPS on the NPU, and 15 TOPS on the CPU. A representative for Intel tells Laptop Mag that the Intel Core Ultra 200HX chips are expected to launch around CES 2025 in January. Intel's Core Ultra 200 series chips first launched in September with the Core Ultra 200V "Lunar Lake" series of laptop processors. The Core Ultra 200V series of chips focused on AI performance and power efficiency, thus taking a bit of a hit to pure CPU power. Intel's Core Ultra 200 "Arrow Lake" series takes an opposing stance. These processors focus on power efficiency, enhanced graphics, and multi-thread performance, and have a less-powerful NPU based on the Intel Core Ultra 100 series "Meteor Lake" architecture. The Intel Core Ultra 200H series of chips will launch for mobile, and we'll likely see it hit gaming laptops and workstations in the future. Though the more high-end desktop-replacement laptops will likely opt for the Intel Core Ultra 200HX desktop processor variant for increased performance. We've seen that trend hold in the past, all the way up to last-gen, and there isn't much reason to expect that to change in the future. While NPU performance is helpful for getting the most out of your laptops battery, a lot of AI is still programmed to run through the GPU. And "Arrow Lake" has some solid GPU AI performance, plus better integrated graphics performance than the Intel 14th "Raptor Lake Refresh" series. So you can still run generative AI on an "Arrow Lake" Intel Core Ultra 200H or HX system. You'll just get worse battery life than the Intel Core Ultra 200V "Lunar Lake" systems, though power efficiency should still be better with the Intel Core Ultra 200H and HX compared to the previous generation. So that should get you som extra battery time. So for the folks confused that "Lunar Lake" didn't come in any gaming laptops, that's because "Arrow Lake" was coming just a few months later.

Intel Arrow Lake-HX "Enthusiast" & Arrow Lake-H "Mainstream" Laptop CPUs Launching In Q1 2025, Xe-LPG+ Gets XMX Support

Intel's Arrow Lake CPUs will be getting their HX "Enthusiast" & H "Mainstream" variants in Q1 2025 with the launch of next-gen laptops. Intel Arrow Lake-HX Enthusiast CPUs For Laptops Offer Up To 24 Cores, Arrow Lake-H Gets Alchemist Refresh With 8 Xe Cores, XMX Support & Up To 77 TOPs While Intel is launching its Arrow Lake-S "Core Ultra 200S" Desktop CPUs today, the company is also providing an update on the laptop front which will see two major launches in the first quarter of 2025. The first one is going to be the most major platform update that replaces the Core Ultra Series 1 "Meteor Lake" lineup, the Core Ultra Series 2 "Arrow Lake" series. The Intel Arrow Lake-H series is going to feature all the major architectural changes including new CPU P-Cores in the form of Lion Cove, new E-Cores in the form of Skymont, an upgraded iGPU, and lots of new I/O capabilities. Starting with the CPU specifications, the Intel Arrow Lake-H "Core Ultra 200H" CPUs will feature a max configuration of up to 6 P-Cores and 8 E-Cores, totaling up to 14 cores and 14 threads. There will be various configurations and two different packages, one tuned for high-performance and one for low-power platforms. One of the biggest changes incorporated within the new laptop lineup is the update to the Xe-LPG+ graphics architecture which is a slightly modified version of the Xe-LPG "Alchemist" cores featured on the Meteor Lake CPUs. This new iGPU comes with support for XMX and offers up to 8 Xe cores. These aren't the same Xe2 cores as the one featured on the Lunar Lake parts. The iGPU has its own 8 MB L2 cache, 128 XMX engines, and 8 ray tracing units. With the new iGPU, the AI TOPs on the Arrow Lake CPUs reach an astonishing 99 platforms TOPs, much higher than the competition and almost on par with the Lunar Lake offerings which do end up faster with over 100 TOPs. The launch of the first Arrow Lake-H laptops is planned for Q1 2025 which is around the same time as the next-gen laptop GPU launches. The second major laptop launch will be the Intel Arrow Lake-HX "Core Ultra 200HX" lineup which will be aiming at the high-end workstation and enthusiast gaming laptop segment. These CPUs will retain the same core configurations as the Arrow Lake-S lineup with 24 cores and 24 threads. The major change will come from the fact that Arrow Lake CPUs will sip much lower power & run very cool, making them a great upgrade for mobility platforms. Another interesting detail is that Arrow Lake-HX CPUs will feature a different package, offering a 33% reduction in size versus the Arrow Lake-S Desktop CPUs. The platform will feature up to 36 TOPs of AI performance and retain memory support in the form of up to DDR5-6400 speeds with 192 GB capacities in SODIMM and CSODIMM flavors. Once again, the Arrow Lake-HX lineup is being positioned as a Q1 2025 launch so we can expect some systems to display during CES 2025. Lastly, Intel also confirmed that Arrow Lake will be getting vPro versions in the future and they have a bunch of announcements for next year so stay tuned.

Intel's highly anticipated Arrow Lake is finally here with new AI CPUs for gaming

Intel announced a new family of desktop and laptop processors codenamed Arrow Lake. We've been following the leaks and slow trickle of official information on this new generation of Intel chips for some time, replacing the hot, power-hungry, and somewhat disappointing Raptor Lake Refresh. Competing against AMD's AM5 platform with Zen 4 and Zen 5 processors, Intel has been on the backfoot these past few generations, largely due to the company's struggle with shrinking its manufacturing process. This resulted in considerably less efficient processors launched to the general public. With Arrow Lake, Intel claims a power reduction of up to 30% compared to 14th-gen processors, and AI buzzword marketing is everywhere. Although these chips will suck less power through the socket, performance is claimed to be up to 10% higher for multi-threaded workloads and graphics have seen a bump of two times what's found in Raptor Lake Refresh. Not bad. Intel did manage to shift the focus away from AI and touched on performance for gaming. Related Intel 15th-gen Arrow Lake: Everything we're expecting from Team Blue Here's everything we know about Intel's next-gen processors. Meet the new Intel Arrow Lake processors Representing the Arrow Lake platform on the desktop is the mighty flagship Intel Core Ultra 9 285K, essentially the direct successor to the Intel Core i9-14900K. It has 24 cores, consisting of 8 P-cores and 16 E-cores, 24 threads (Hyperthreading is out the window), 4 GPU cores, and a maximum boost speed of 5.7 GHz. It's a slight reduction compared to the outgoing 14900K, which had 32 threads and a 6.0 GHz boost speed, though that processor almost hit more than 300W at maximum load. Intel is pushing efficiency gains here without sacrificing performance, so results should be similar. CPU Cores Threads Speed TDP GPU (Cores) NPU (TOPS) MSRP Intel Core Ultra 9 285K 8P + 16E 24 5.7 GHz 250W 4 13 $589 Intel Core Ultra 7 265K 8P + 12E 20 5.5 GHz 250W 4 13 $394 Intel Core Ultra 7 265KF 8P + 12E 20 5.5 GHz 250W - 13 $379 Intel Core Ultra 5 245K 6P + 8E 14 5.2 GHz 159W 4 13 $309 Intel Core Ultra 5 245KF 6P + 8E 14 5.2 GHz 159W - 13 $294 The rest of the SKUn line-up consists of an Intel Core Ultra 7 265K with 8 P-cores and 12 E-cores and an entry-level Intel Core Ultra 5 245K with 6 P-cores and 8 E-cores. A "KF" variant of the Intel Core Ultra 5 and Core Ultra 7 will also be available, which strips away the GPU for a slightly more affordable MSRP. Every CPU within the Arrow Lake-S family of desktop chips has a tera operations per second (TOPS) score of 13. This has been handled by a dedicated neural processing unit (NPU) within the PC but Intel is combining the NPU with the CPU and GPU on the package to leverage more computing power. Close Thermal design power (TDP) tops at 250W for the Intel Core Ultra 9 and 7 series with the Core Ultra 5 CPUs at 159W. Intel shared a reduction of up to 58% in power consumption compared to previous-generation chips. Intel has consistently been the leader in single-core performance and Arrow Lake is seemingly no exception. While we have to get these chips in for our testing, the chip maker claims a 1T performance gain of 8% over Raptor Lake Refresh and a 4% win over the AMD Ryzen 9 9950X with the Intel Core Ultra 9 285K -- nT gains are a similar story with Intel claiming a 13% win over AMD's flagship Zen 5 CPU. Related Intel is back: Core Ultra Series 2 is here to take on Snapdragon X Elite and Apple Silicon It's promising to win in both battery life and performance In released official test data, Intel showed 165W less power being used by the Intel Core Ultra 9 285K over the 14900K in games with similar performance results across a wide selection of tests. Some showed lower performance, others showed notable gains, so it worked out to an even ground with the Arrow Lake CPU drawing considerably less electricity. It's also good to see Intel pushing the Intel Core Ultra 5 series of CPUs as the essential picks for pure gaming systems. The Core Ultra 7 is a better match for 4K and more intensive gaming with productivity in mind, and the Core Ultra 9 is reserved for PC enthusiasts. More changes under the hood Arrow Lake marks a considerable upgrade for Intel. This isn't a simple refresh like we've seen with the previous couple of generations. The company is going all-in with AI, much like the rest of the industry, and Arrow Lake is Intel's first with CPU AI, GPU AI, and NPU. At the surface, we've got support for integrated Wi-Fi 6E, Thunderbolt 4, PCIe 5.0, a revamped Xe GPU, new Lion Cove P-cores and Skymont E-cores, and a silicon security engine. The primary goal for Intel with Arrow Lake is efficiency and power reduction. The changes are substantial with some interesting results from Intel. Arrow Lake HX chips have seen a 33% reduction in package size, yet still include up to 16 E-cores and 8 P-cores like the flagship Arrow Lake-S Intel Core Ultra 9. The new P-cores have been "re-architected" for efficient performance. In other words, Intel pushed through architectural changes for lowering power draw. There's AI-based power management, wider predict and scheduling, and optimization for power, performance, and area (PPA). E-cores saw changes over the previous generation with increased workload coverage, double vector and AI throughput, and scalability. Compared to previous-gen cores, the new Lion Cove and Skymont cores should offer considerable gains in efficiency, allowing Intel to reduce the amount of power drawn by the package, resulting in a cooler and more stable system. Because we've seen issues with how everything is scheduled on these new Intel chips, the company worked on new E-core prediction, P-core telemetry infrastructure, and prediction model. These changes with the Intel Thread Director should allow the CPU to more effectively handle various workloads across P-cores and E-cores. Close The all-new Intel 800-series chipset joins the Arrow Lake desktop processors with plenty of I/O support and up to 48 PCIe lanes when combined with a 15th-gen CPU. Thunderbolt 4 and 5 are supported with Wi-Fi 6E and Wi-Fi 7 both on the specs sheet (Thunderbolt 5 and Wi-Fi 7 being discrete). Intel has also revamped the overclocking capabilities of its platform and processors, offering 16.6MHz steps for granular core clock changes, DLVR bypass, features in the Intel eXtreme tuning utility, the ability to bypass voltage limits at lower temperatures, and new XMP and CUDIMM DDR5 memory support. Speaking of RAM, Arrow Lake can support up to DDR5-6400 at 48GB per DIMM for a 192GB total maximum. For the graphics, Intel added full DirectX 12 support throughout the Xe GPU with up to two times the available compute resources, higher clock frequencies, and architectural efficiency gains. This should provide us with the claimed two times better performance over previous-gen processors and be most noticeable on laptops without an available discrete card. Available soon with laptop chips arriving in 2025 Intel is wasting no time in getting these new desktop chips to market. To battle AMD and attempt to stop the loss of market share to Team Red, Intel will make the new desktop processors available for pre-order on October 24. For those of you who prefer using a portable device, Intel Core Ultra H and HX series mobile processors for laptops will arrive in Q1 2025.

Intel Unveils Next-Gen 'Arrow Lake' Desktop CPUs: They're All About Efficiency

The unsurprising part of Intel's pitch for its "Arrow Lake" desktop processors is its extensive emphasis on AI. The less-obvious new feature in a series of CPUs built for computers that don't run on batteries? Energy efficiency. "It's one of the largest power reductions and efficiency gains that Intel has seen in quite some time," said Greg Boots, platform marketing manager for Intel's client computing group, at a September 4 preview event at IFA in Berlin. Arrow Lake -- formally known as the Core Ultra 200S series, with sales starting October 24 -- introduces some significant architectural changes that Intel began hyping up two years ago. Many of the changes map to the "Lunar Lake" series of laptop chips that the company introduced at IFA 2024. However, Arrow Lake aims at a market where Intel doesn't face intense competition from Qualcomm's Snapdragon X series of laptop processors. The Heart of Arrow Lake: Core (and Cost) Competition Like previous Intel processors going back to the Alder Lake series introduced in 2021, Arrow Lake farms out operations to a hierarchy of specialized processing cores. That maintains a serious design distance between Arrow Lake and AMD chips like the Ryzen 9000 Zen 5 series introduced in mid-2024, which constitutes Intel's sole competition in desktop CPUs: Intel's Santa Clara, Calif., neighbor has yet to adopt a hybrid architecture combining power -- and efficiency-optimized cores. As you'll see below, Intel is also ahead of AMD on pricing. Two examples: Its flagship Core Ultra 9 285K has a suggested retail price of $589 versus $649 for AMD's leading Ryzen 9 9950X desktop chip. Likewise, Intel's Core Ultra 7 CPU outpaces AMD's second-tier Ryzen 9 chip on core count while costing more than $100 less. The initial lineup comprises five CPUs on the new LGA1851 socket, with the two ending in "F" indicating a lack of integrated graphics. As rumored, Intel has dropped support for its traditional thread-doubling Hyper-Threading with these chips. Arrow Lake features versions of Intel's "Lion Cove" Performance (P) cores and "Skymont" Efficient (E) cores that debuted in Lunar Lake but includes more of them, depending on the configuration of individual processor models. That ranges from six to eight of the Lion Cove P-cores and from eight to 16 of the Skymont E-cores. Intel backs those cores with a deeper bank of caches in Arrow Lake than in Lunar Lake. Intel has outfitted 4MB of L2 for each E-core cluster, 3MB of L2 for each P-core, and as much as 36MB of L3. The P-cores further benefit from AI-based power management and an expanded scheduling framework to handle more transactions. At the same time, the E-cores see upgrades such as doubled vector and AI throughput, as well as hardware-based task prediction from Intel's Thread Director. Intel says these power-saving moves will pay off in improvements such as reduced gaming power consumption by as much as 165 watts (W) and a drop in CPU package temperatures during active gaming of as much as 17 degrees C. Another Arrow Lake Addition: AI, of Course Arrow Lake is also Intel's first desktop processor architecture to integrate a neural processing unit (NPU), the company's latest nod to the trend everybody else is chasing in personal computing. As Boots said at IFA, "AI is rapidly becoming a cornerstone of PC technology." Intel says this module, built on its NPU 3 architecture and packaging two neural compute engine tiles and 4MB of scratchpad memory, can hit 13 trillion operations per second (TOPS) at AI tasks, combined with up to 15 TOPS from the CPU proper and 8 TOPS from the GPU. That amounts to far fewer TOPS from the NPU than the up to 48 TOPS available from Lunar Lake's NPU. During the Q&A part of the IFA presentation, Josh Newman, general manager and vice president of product marketing and management at Intel's Client Computing Group, chalked that up to the company's design-process cycles getting slightly out of sync. "What we did in Lunar Lake with the new NPU, we'll obviously bring to other segments in the future," Newman said. Arrow Lake's integrated graphics processor (IGP), in turn, is built on the Xe IGP that Intel launched with its Meteor Lake laptop processor line. The Xe IGP combines four Xe cores with four ray-tracing units, 4MB of L2 cache, and DirectX 12 support. Arrow Lake -- to be manufactured mostly by TSMC on a variety of processes depending on the tile, another break from Intel's past practice -- uses Intel's Foveros 3D packaging technology to combine all these parts as a series of tiles in one compact processor module. Smarter Overclocking Features for Power Users Intel's focus on traditional desktop-PC enthusiasts with Arrow Lake includes multiple features to ease overclocking, such as acceleration steps as small as 16.67MHz, dual base clocks to allow different frequencies for the CPU's tiles, and the option to override the chip's internal voltage management. Intel plans to add an AI-based overclocking feature to the Intel Killer performance-tuning suite sometime later. The suite can already automatically optimize app priority, bandwidth management, and access-point selection. Asked at IFA about the potential of that to add to the computing industry's growing catalog of AI fails by making a warranty-voiding move, Boots said Intel would exercise caution in designing that feature. "I don't expect for our users to click a one-click AI overclocking and have that void their system," he said. We asked Intel if that expectation had changed since early September but did not get a response before publishing. More -- make that more important -- automated help will come on the security side via three built-in engines to monitor threats: First is linear address space separation to ward off side-channel attacks like Hertzbleed, second is Intel Boot Guard, and finally, the Intel Silicon Security Engine that shipped with Meteor Lake. Lunar Lake also sticks with the previous version of Thunderbolt but is current on the two wireless connectivity standards. Intel says it's leaving those upgrades as options for computer vendors and builders to add as discrete hardware. "It basically comes down to timing," Newman said at IFA in a nod to the magnitude of changes in Arrow Lake. And desktops are a more forgiving environment for bolt-on upgrades. "They tend to be able to accommodate a lot more discrete options." What's Next? Arrow Lake Heads to Laptops Intel will, however, add those standards to the forthcoming Arrow Lake-H portable-gaming processors, briefly teased during the IFA presentation and projected for the first quarter of 2025. That will bring the essential bits of this architecture into a smaller package, with Foveros 3D allowing for up to a 33% shrink in its dimensions while still delivering faster performance because of a much higher-powered Xe GPU. Intel predicts that this graphics chip, doubling the Xe cores and L2 cache of the desktop Arrow Lake's GPU, will hit up to 77 TOPS, with its NPU contributing the same 13 TOPS as in the desktop version and the CPU stepping back slightly to 9 TOPS. Boots' prediction at IFA: "Arrow Lake-H is shaping up to be a standout platform." That may give Intel new goals for its next line of desktop processors if true.

Intel pins hopes on Arrow Lake desktop processor refresh

New silicon, new architecture, and loads of new motherboards rise to support it, but will power be anchored down? Back in September 2023, Intel unveiled its newly designed Meteor Lake SoC for the mobile market, which was the first disaggregated chip for mobile using multiple tiled packaging. While in consensus opinion indicates Meteor Lake flopped, it did pave the way for Intel to try new things in the consumer space. Its next generation of Lunar Lake processors appear to increase efficiency and performance in the low-end to mid-range mobile segment, and while Intel has been working on bolstering its mobile portfolio, they also just announced its next client focused desktop platform. Enter Arrow Lake, Intel's new desktop platform, which aims build on what works, and no doubt also to move on from the controversy surrounding the instability of its previous 14th Gen Raptor Lake Refresh chips. Intel is integrating its AI-focused neural processing unit (NPU) into its Arrow Lake desktop products - which is the first time it has done so outside of its mobile focused SoCs. Arrow Lake is using a disaggregated and scalable tiled approach to manufacturing and design, with separate tiles for compute, graphics, I/O and the SoC. When CEO Gelsinger announced Intel's aggressive and ambitious "five nodes in four years" roadmap in 2021, it was at a critical juncture for the tech giant. Since then it has been on a rocky path and more recently, has been posting relatively weak financials. Touching more on the nodes, a lot of the focus around Intel's latest Arrow Lake processors for desktop was expected to be built on its 20A node, which is the first Intel node to move from the relative nanometer measurement to an angstrom-based measurement - the dawn of the angstrom era. Unfortunately Intel recently cancelled 20A, which in reality was only meant to be a stepping stone to its 18A node. As we know, Intel is expected to deliver its upcoming Panther Lake client chips and their Clearwater Forest for server on 18A. Although many expected Intel to use all of its fabbing and manufacturing flexibility on the 20A node to manufacture Arrow Lake, the company has actually outsourced all of the tile manufacturing for this generation to TSMC, while Intel will still be packaging the disaggregated architecture using its own Foundry. All of the tiles are packaged onto a base tile using Intel's Foveros 3D stacking packaging, which integrates all of the tiles onto one package, as we've seen previously with their disaggregated Meteor Lake SoC architecture. Let's dive into Arrow Lake, first by focusing on Intel's new compute tile for the device. This particular tile is manufactured using TSMC's N3B node, with Intel introducing two new cores to its desktop portfolio, which the company concurrently uses for its Lunar Lake mobile SoCs: Lion Cove and Skymont. The Lion Cove performance (P) cores primarily focus on delivering IPC and single-threaded performance gains. Intel has claimed previously that Lion Cove brings up to a 9 percent improvement in performance over the previous Raptor Lake P-cores. As always, take in-house performance figures with a tiny tiny grain of salt. Architecturally, the biggest enhancements in P-core (Lion Cove) are in the execution engine. Intel has expanded its OoO window by increasing both the allocation/rename and retirement stages within the pipeline. This wider execution pipeline should give the core creedence to issue more instructions per clock cycle, which should theoretically present itself through a strong increase in throughput during compute-heavy tasks. Intel greatly expanded the number of execution ports, which should provide a substantial increase in the types of parallel instructions being handled by Lion Cove. This means benefits should come to the ILP are most pronounced in workloads such as rendering, AI inferencing, and physics simulations. Branch prediction, as you know, is one of the key elements used in modern processors to maintain high throughput. The redesigned Lion Cove branch predictor is designed to cut down on mispredictions. These are quite expensive in terms of pipeline stalls, but by making improvements in both prediction accuracy and the recovery latency from mis-predicted branches, it should lead to significant enhancements in efficiency, especially those involving dynamic workloads like AI-driven tasks or gaming. Intel has also increased the size of the reorder buffer, which tracks the status of every in-flight instruction as it works its way through the pipeline. This should in theory bolster the overall out-of-order executions from when the processor is waiting on a branch or memory access that has not been resolved yet. In terms of cache hierarchy of the new performance cores, Intel has completely redesigned the structure compared to previous generations. Lion Cove features a multi-level data cache consisting of a 48KB L0D cache with 4-cycle load-to-use latency, a 192KB L1D cache with 9-cycle latency, and an expanded L2 cache. One area where it has improved the L2 cache per core in Lion Cove is by increasing it to 3 MB. A larger cache means fewer memory accesses need to spill over onto the slower L3 cache or even the main memory (DRAM), and this is something that is quite welcomed under mostly heavy loads such as video editing, large-scale simulations, and high-framerate gaming. Intel further extends the shared L3 cache to 36 MB shared; this means the core can share much more data between threads. Clearly, for multi-core workloads, inter-core communications have to be both low-latency and high-bandwidth, which are both important attributes to cache hierarchy improvements that will directly translate to smooth multi-threaded performance, especially in several workloads that involve multiple cores accessing the same data sets. The Skymont E-cores mark a notable upgrade over Gracemont, at least on paper, which we saw present in 12th, 13th, and 14th Gen Core processors. The Skymont cores will be much more efficient than their previous counterparts, with Intel promising a gain of up to 32 percent in integer performance and as high as 72 percent in floating-point performance over Gracemont at ISO -frequency. Again, take these performance claims with a pinch of salt, but in theory, its plausible. The E-cores reduce power consumption in multi-threaded tasks, which is their job after all: efficiency cores free up the higher-performing P-cores by dealing with background processes, all forms of parallel computation tasks, and light-to-moderate work. Intel also removes hyperthreading, so each core is parallel in count to each thread, e.g. one thread per core. This is designed to reduce the overall power envelope and give more power and thermal headroom overall. One of the major architectural enhancements in Skymont is that it has increased vector throughput with more lanes for the SIMD units, which should allow for more executions per data cycle in each E-core. This is essential in workloads such as multimedia, AI, and scientific computing etc., Secondly, L2 cache per cluster has increased to 4 MB from 2 MB, which should remove some memory access bottlenecks. The larger cache, in conjunction with deeper instruction queues and wider instruction dispatch, should technically make the Skymont cores better at parallel workloads and improve the execution efficiency across multi-threaded applications. On paper, these improvements over make Skymont E-cores an improvement over Gracemont (the E-cores in Raptor Lake), and should handle content rendering, AI inference, and background system management-all without consuming unreasonable amounts of power. That's the nature of an efficiency core after all. The GPU tile on Arrow Lake is fabricated on the TSMC N5P Process, which is an improved version of the 5nm node already known to bring better performance and efficiency compared to the standard N5 process. Based on an improved version of its Xe architecture, the GPU tile includes four Xe-cores, each with ray tracing units and enhanced vector engines. It uses TSMC's N5P manufacturing process for better power efficiency at higher clock speeds, thus allowing it to give the GPU up to 2x the graphics performance compared to the previous 14th Generation Processors. With 64 vector engines with 16 per Xe-core and full support for AI workloads through DP4a instructions, the GPU tile looks architecturally solid on paper. Besides this is the addition of XeSS, Xe Super Sampling which should allow for even greater improvements in graphical output via AI image upscaled actions, especially compared to previous generations of Intel's chips. In short, the balance between power efficiency and performance does make the N5P-based GPU tile stand out from other desktop chips, but don't expect to play AAA titles at maximum settings; integrated graphics is still far away from that yet, no matter what the marketing wants you to believe. Designed on the TSMC N6 process node, the SoC and I/O tile is responsible for system connectivity, memory, and data flow control in general within Intel's Arrow Lake architecture. There is high-bandwidth DDR5-6400 memory, ensuring that system resources become available with minimum latency. Simultaneously, it also controls up to 24 PCIe 5.0 lanes for fast data transfer to and from GPUs or NVMe SSDs. Also, it provides the latest kinds of connectivity like Thunderbolt 4, Wi-Fi 7, and Bluetooth 5.4 to provide the latest in communication standards for Arrow Lake platforms. Built on TSMC's N6 process node, it strikes a balance between performance density and energy efficiency since the manufacturing process keeps power consumption very low while offering high throughput. While not quite as vital to the performance of its compute or GPU siblings, this SoC-I/O tile really serves a critical function in enabling all of that efficient data flow across the platform from CPU to GPU to memory, all the way down to peripheral devices. It's also worth noting that Intel has increased the memory speeds with Arrow Lake up to DDR5-6400 at JEDEC settings, which is up from DDR5-5600 as saw on their previous 14th Gen Raptor Lake Refresh chips. This means each Arrow Lake CPU regardless of the silicon lottery is guaranteed to handle the uplift in memory speeds, but above this through the implementation of X.M.P memory profiles, with some kits of DDR5 hitting up to 8000 MT/s technically voids Intel's warranty. In terms of motherboard support, Intel is introducing a new socket, namely the LGA 1851 socket, which means a whole load of new motherboards are coming out to support the launch of Arrow Lake. The chipset for Arrow Lake, or at least the premium chipset launching alongside the chips is named Z890 and brings a load of possible I/O options and configurations. The Z890 chipset features up to 24 x PCIe 4.0 lanes, 10 x USB 3.2, with up to 14 x USB 2.0 and 8 x SATA ports. How these features are enabled and implemented are primarily down to motherboard vendors, with many premium models offering both USB4, Thunderbolt 4 and the latest in consumer networking such as the new Wi-Fi 7 CNVis, and even some with 10G Ethernet for users who require it. As it stands, Intel is launching five new Core Ultra 200-series processors with prices starting at $294 for the Intel Core Ultra 5 245KF (6P+8E/14T), with the flagship Core Ultra 9 285K (8P+16E/24T) coming in at $589. The top chip, the Intel Core Ultra 9 285K, has 8 P-cores and 16 E-cores, and can boost up to 5.6 GHz on the P-Cores and it also comes with 36 MB of L3 cache, with a base TDP of 125 W; it has a turbo TDP of up to 250 W. While TDP is something of an oxymoron in the world of client desktop chips, we typically see motherboard vendors through their interpretation of multi-core enhancement (MCE) skirt these limits anyway to allow it to stay ahead of the competition. Since multi-core enhancement tends to often push CPUs beyond their official limits, for users, gains may come at the cost of thermal efficiency and potentially stability. This can pose instability issues, especially as motherboard vendors on their preset profiles typically putting too much CPU VCore voltage through to accommodate the many (chips) and not what each piece of silicon is capable of. The Intel Core Ultra 200-series Arrow Lake processors are officially expected to go on sale on October 24. Also, during its Arrow Lake press briefing, Intel did mention that its VPro SKUs based on Arrow Lake for commercial systems and SMEs will definitely be coming out, but would not be drawn on when this might be. In another twist, Intel also announced that its Core Ultra H and HX series for premium gaming laptops are expected to arrive in Q1 2025; this is provided all things go to plan for Intel as things haven't been so great for the chip giant lately. Intel could really do with a trouble free launch for Arrow Lake as it continues the process of laying off some 16,000 workers following a disastrous second quarter where it reported $1.6 billion in losses. ®

Intel's upcoming Arrow Lake H laptop chips will offer beefier GPUs for AI workloads

They're meant for high performance notebooks, and notably they have slower NPUs than Lunar Lake. Alongside its new family of Arrow Lake desktop hardware, Intel today also gave us a few tidbits around its upcoming Arrow Lake H mobile chips for high performance laptops. First off, they're not expected to arrive until the first quarter of 2025 -- but the slight wait might be worth it, as Intel says they will offer powerful new Xe GPUs with XMX. Thanks to that upgrade, the GPU alone will offer four times better AI workload processing than its previous chips, alongside double the ray tracing performance and twice as much cache (8MB L2). Notably, though, these new chips will still lag behind the company's less powerful Lunar Lake processors when it comes to NPU and overall AI TOPS (tera operations per second) figures. Arrow Lake H's NPU will hit 13 TOPS, the new GPU will reach 77 and the CPU will offer 9 TOPS. Taken altogether, it'll offer up to 99 TOPS of performance. Lunar Lake, meanwhile, sports a 48 TOPS NPU and up to 120 TOPS of system-wide AI performance. The difference makes sense when you consider what these chips are meant for. Lunar Lake is mostly geared towards ultraportables and slim workstations, while Arrow Lake H chips are targeted at demanding notebooks with desktop-like performance. While they can technically be called AI PCs, Arrow Lake H's low NPU performance doesn't meet the bar for Microsoft's Copilot+ badge (those require at least 40 TOPS NPUs). You'll be able to run basic AI features, like Windows Studio Effects in video chats, but not more complicated tasks like Recall. Intel didn't have many other details to share about Arrow Lake H, but we'll likely hear more at CES 2025.

Intel Announcement Preview: Intel Core Ultra 200 Arrow Lake CPUs

Intel Arrow Lake Core Ultra 200 Series: Technical Specifications and Release Schedule Today, Intel unveils its latest desktop processor lineup, the Core Ultra 200 series, codenamed Arrow Lake. This series features a maximum total of 24 cores, comprising 8 high-performance Lion Cove P-cores and 16 energy-efficient Skymont E-cores, engineered to manage intensive computational tasks and gaming effectively. The processors are equipped with a substantial 76 MB cache, divided into 36 MB of L3 and up-to 40 MB of L2, which facilitates rapid data retrieval, thereby reducing latency and boosting overall system performance. These specifications make the Core Ultra 200 series well-suited for applications that demand quick access to large datasets, including artificial intelligence and machine learning workloads. Additionally, the processors boast base clock speeds of 3.7 GHz for P-cores and 3.2 GHz for E-cores, with boost capabilities reaching up to 5.7 GHz and 4.6 GHz respectively. This level of clock speed ensures strong single-threaded performance, which is essential for high-performance applications such as gaming and content creation. Model Cores & Threads Architecture Cache Base Clocks Boost Clocks TDP (PL1 / PL2) Core Ultra 9 285K 24 cores (8 P-cores + 16 E-cores), 24 threads Lion Cove (P-cores), Skymont (E-cores) 76 MB (36 MB L3 + 40 MB L2) 3.7 GHz (P-cores) / 3.2 GHz (E-cores) 5.7 GHz (P-cores) / 4.6 GHz (E-cores) 125 W / 250 W Core Ultra 7 265K / 265KF 20 cores (8 P-cores + 12 E-cores), 20 threads 66 MB (30 MB L3 + 36 MB L2) 3.9 GHz (P-cores) / 3.3 GHz (E-cores) 5.5 GHz (P-cores) / 4.6 GHz (E-cores) 125 W / 250 W Core Ultra 5 245K / 245KF 14 cores (6 P-cores + 8 E-cores), 14 threads 50 MB (24 MB L3 + 26 MB L2) 4.2 GHz (P-cores) / 3.6 GHz (E-cores) 5.2 GHz (P-cores) / 4.6 GHz (E-cores) 125 W / 159 W The KF models lack an activated integrated GPU, thus a dedicated graphics card will be mandatory. The Arrow Lake processors exclusively support DDR5 memory, with supported JEDEC speeds of DDR5-6400. Thermal management has been addressed, with the CPUs operating at a Thermal Design Power (TDP) of 125 W (PL1) and 250 W (PL2). Furthermore, the Core Ultra 200 series utilizes the new LGA 1851 socket and is compatible with Intel's upcoming 800-series motherboards announced today. This new socket design is intended to support future generations of Intel CPUs, providing users with a clear and straightforward upgrade path for their desktop systems. Core Architecture With Arrow Lake intel moved towards two new cores, one for performance, the other being energy efficient, Lion Cove and Skymont. These architectures are designed to meet diverse computing requirements by offering improved performance and energy efficiency. Lion Cove serves as the foundation for Intel's high-performance cores, focusing on delivering fast single-threaded performance. It achieves this through enhancements in instructions per cycle (IPC) and incorporates an AI-driven self-tuning controller. This controller dynamically adjusts to real-time operating conditions, optimizing both performance and thermal management. Such features make Lion Cove well-suited for intensive applications, including gaming, content creation, and complex computational tasks. Lion Cove's architecture includes broader execution units and sophisticated branch prediction mechanisms, which collectively increase the number of instructions processed per cycle and minimize processing delays. Additionally, larger cache sizes are implemented to reduce latency and accelerate data access, essential for maintaining high performance in data-heavy applications. Intel's Arrow Lake-S desktop processors will include 3 MB of L2 cache for each performance core. This is a substantial increase compared to previous generations. The current Raptor Lake family of processors have 2 MB of L2 cache per performance core. The 2021 Alder Lake CPUs feature 1.25 MB of L2 cache per core, and Rocket Lake processors offer 512 KB of L2 cache per core. Accumulated with Arrow Lake CPUs. The total L2 cache capacity for performance cores is expected to increase to 24 MB, a significant expansion. While specific details regarding the L3 cache enhancements for Arrow Lake remain undisclosed, the total shared smart cache runs to 36MB. The E cores run a shared cluster core of 4MB L2. Conversely, the Skymont architecture-based cores focus on efficiency and scalability. As the third generation of E-core designs, Skymont prioritizes high performance per watt, targeting energy-efficient computing environments. It features a 9-wide decode stage and enhanced execution units, which improve its ability to manage parallel tasks effectively. Furthermore, Skymont incorporates advanced power management technologies that allow the cores to sustain high performance while reducing power consumption. These features are particularly beneficial for mobile and embedded devices, where battery life and energy efficiency are critical factors. IGP: Intel Xe-LPG Graphics Architecture Intel's Xe-LPG architecture is the foundation of their latest integrated graphics solutions. At the heart of the Xe-LPG architecture are eight Xe graphics cores, each with 128 vector engines, totalling 1,024 unified shaders. This setup delivers significant computational power, allowing the architecture to handle demanding tasks across different applications effectively. Additionally, Xe-LPG includes eight samplers and four-pixel backends, which improve its ability to manage complex rendering processes and maintain high performance. A notable feature of Xe-LPG is its support for real-time ray tracing, enabled by eight dedicated Ray Tracing Units. The architecture includes XMX units for AI that offer 67 tera operations per second (TOPS) of artificial intelligence performance, making it suitable for AI tasks like image recognition and machine learning. Xe-LPG fully supports DirectX 12 Ultimate ensuring compatibility with advanced graphics features such as variable rate shading and mesh shaders. Power efficiency is another important aspect of Xe-LPG, as it delivers twice the performance per watt compared to previous integrated graphics solutions. The improvement allows high(er)-performance graphics without significantly reducing battery life, making Xe-LPG an excellent choice for laptops and other mobile devices where power consumption is vital. New Socket, New Motherboards - Wifi7 - PCIe Gen 5 Intel's Z890 motherboard chipset, part of the 800-series, is designed to support the latest Arrow Lake processors, delivering improved performance and connectivity for desktop (gaming) PCs. A major feature of the Z890 chipset is its support for up to 48 PCIe lanes. This allows for flexible setups that can include multiple high-speed devices such as graphics cards, NVMe SSDs, and various peripherals. With PCIe 5.0 support significantly boosts data transfer rates and overall system performance. The Z890 chipset will offer a wide range of connectivity options. It includes (optional per motherboard/brand/SKU) two USB4/Thunderbolt 4 ports, providing and high-speed connections for devices like external storage and high-resolution monitors. Additionally, the chipset features a variety of USB ports to support different peripherals. Specifically, it offers 14 USB 2.0 ports, 5 USB 3.2 Gen 2x2 (20 Gbps) ports, 10 USB 3.2 Gen 2 (10 Gbps) ports, and 10 USB 3.2 Gen 1 (5 Gbps) ports. This extensive selection ensures that users can connect multiple devices simultaneously without sacrificing performance. On the memory side, the Z890 chipset supports DDR5-6400 memory modules, making it suitable for applications that require high memory bandwidth, such as gaming, content creation, and data-intensive tasks. It can accommodate memory modules up to 48 GB, allowing users to expand their system's memory capacity as needed. For storage, the Z890 chipset includes 8 SATA III (6 Gbps) lanes and expanded M.2 support using PCIe 5.0 for primary M.2 slots. The chipset also features 60 High-Speed Input/Output (HSIO) channels, with 26 channels from the CPU and 34 from the Platform Controller Hub (PCH). This large number of HSIO channels allows a wide range of devices to be connected without compromising performance or bandwidth. Available in October 24th Intel has officially announced the release timeline for the Core Ultra 200 series. The formal introduction of the new processors is today October 10th, coinciding with the unveiling of Intel's new desktop platform featuring the LGA-1851 socket, which replaces the previous LGA-1700 series after three years of use. The market launch of the Core Ultra 200 series is planned for October 24th, thus two weeks after today, aligning with the release of media reviews, this website included.

Intel's first Arrow Lake CPUs are here, and they're all about power efficiency

Intel has finally unveiled the first round of Arrow Lake CPUs to hit the market, and they promise better performance than their predecessors for less power. This is potentially a big deal because in addition to their improved efficiency, these are the first desktop CPUs Intel has released with an onboard Neural Processing Unit (NPU). So while these first Arrow Lake chips aren't delivering huge performance gains over their Intel Raptor Lake predecessors, they do promise to deliver similar performance for less power and heat, with the additional advantage of onboard NPUs. NPUs have been all the rage in laptop chips over the last year or two because they're optimized to chew through the kinds of workloads that most AI apps demand, so having a CPU with a powerful (and power-efficient) NPU has become important if you care about taking advantage of AI-centric features like those available on Copilot+ PCs. Of course, Microsoft only makes those Copilot+ features available on Windows 11 machines with NPUs capable of achieving 45+ TOPS (trillion operations per second), and since these new Arrow Lake CPUs all deliver just 13 TOPS they likely won't be able to support those advanced features. But most AI applications these days are capable of utilizing all of the processing power in your PC to run, not just the NPU, so pairing a beefy graphics card (like the Nvidia GeForce RTX 4090) with one of these new Arrow Lake chips will ensure you'll have more than enough computing muscle to chew through AI apps running locally on your PC. Here's everything you need to know about the new Intel Arrow Lake chips dropping October 24. Intel is releasing five Arrow Lake desktop CPUs to start, and you'll be able to order them from participating retailers starting October 24 at a starting MSRP of $294 for the Intel Core Ultra 5 245KF. But that's the entry-level Arrow Lake CPU, which has the lowest clock speed and fewest cores available. To purchase the beefy Intel Core Ultra 9 285K, the flagship top-of-the-line Arrow Lake CPU, you can expect to pay a cool $589. Intel is releasing an inaugural batch of five Arrow Lake CPUs to start, with core counts ranging from 14-24 and speeds from 5.2-5.7 GHz. They all have onboard NPUs capable of 13 TOPS, and most have 4-core onboard GPUs. As usual, the chips with -F in the name don't come with onboard GPUs, so you'll need to pair them with a discrete AMD, Intel or Nvidia graphics card in a PC. Intel isn't promising that its new Arrow Lake CPUs will deliver major gains in performance over previous chips. Instead, it's trying to pitch these chips as the start of a new generation of NPU-equipped Intel desktop CPUs that are more efficient than prior models, with a lot of room to grow in terms of performance. That's a nice way of saying this first round of Arrow Lake chips don't appear to be much more powerful than the top-of-the-line Raptor Lake chips Intel released late in 2023. However, they do promise to draw less power and run cooler while delivering comparable power. As you can see from the charts above and below (which were provided by Intel), the company claims its new flagship Arrow Lake desktop CPU is on par with the top-of-the-line Intel Core i9-14900K in terms of performance, yet demands significantly less power to do the same work. This may not be the most exciting marketing campaign for performance freaks who just want to build the fastest, most capable gaming PC they can, but it's good news for efficiency nerds like me who are getting a little tired of watching our power bills spike during marathon gaming sessions. Intel also provided some of its own internal benchmarks of gaming performance on the new flagship Intel Core Ultra 9 285K ($589) vs. the top-of-the-line AMD Ryzen 9 9950X ($649), and the results are mixed. As you can see from the chart of benchmark results embedded above, Intel isn't promising its new top-tier CPU can reliably beat AMD's best. Instead, it claims that (at least at 1080p on high settings) the results are mixed, with AMD's top-tier chip delivering better performance in games like Cyberpunk 2077 and Red Dead Redemption 2 while the new Arrow Lake flagship delivers more frames per second in games like Total War: Warhammer 3 and Sid Meier's Civilization VI: Gathering Storm. Of course, we'll have to wait until these Arrow Lake chips get out into the market and enthusiasts start benchmarking them under real-world conditions to really know how well they perform. I have to say, I find Intel's frank and efficiency-first marketing of these first Arrow Lake chips to be kind of refreshing in a market dominated by flashy marketing campaigns and big promises. While gaming enthusiasts may understandably be underwhelmed by the promised performance gains of these first Arrow Lake CPUs vs. their Raptor Lake predecessors, I believe it when Intel says they represent the floor of a new CPU family with lots of room to grow. Intel has been on the back foot for a few years now as it tries to compete with Apple, AMD and now Qualcomm in the laptop chip business, which is all about delivering great performance with minimal heat and power draw. And while Intel continues to lag behind the competition in a number of key metrics, we're seeing the fruits of its labor show up in cooler, quieter desktop CPUs. The problem is, I don't know how much of a difference that will make in high-end PCs optimized for gaming and content creation. Sure, it's great if your new high-end CPU draws 50-100 watts less power (on average) than your old one, but how much of a difference will that makes in a beefy gaming rig with an Nvidia GeForce RTX 4090 sucking down 450-800 watts of power in an intense gaming session? However, these new Arrow Lake chips could be a huge boon for mini PCs, especially mini gaming PCs, because small cases are where quiet, cool high-performance CPUs really shine. Of course, we'll have to wait until the end of October to see how much of an impact they really make.

Intel Launches Arrow Lake Core Ultra 200S -- big gains in productivity and power efficiency, but not in gaming

Intel announced the first five models of its Arrow Lake desktop processors, otherwise known as the Core Ultra 200S series, with prices ranging from the $294 14-core Core Ultra 5 245KF to the flagship $589 24-core Core Ultra 9 285K. The chips will come to market on October 24, 2024. Intel says Arrow Lake brings up to 15% more multithreaded performance, 5% more single-threaded performance, drastically reduced power consumption, and the first dedicated AI accelerator (NPU) in a mainstream desktop chip. Surprisingly, things aren't as positive on the gaming side of the equation. Intel says its new flagship Core Ultra 9 285K matches AMD's flagship Ryen 9 9950X in gaming. However, according to our testing, Intel's own current-gen 14900K flagship is 8 to 10% faster than the Ryzen 9, meaning Intel's new Core Ultra 9 285K flagship could be slower in gaming than its own current-gen halo part. Additionally, Intel says that AMD's fastest gaming-optimized chip, the Ryzen 7 7800X3D, is 5 to 7 percent faster than the Core Ultra 9 285K, but based on our 7800X3D tests and the relative positioning, the gap could be much larger. Of course, our testing will tell the final tale, and the mid-range processors could offer a stronger value proposition. After AMD's lackluster gen-on-gen gaming improvements with its mainstream Zen 5 chips and Intel's lower predictions for the new Ultra 9, Intel's new lineup makes it look like we'll have to wait for the looming AMD Ryzen 9000X3D processors for any (potential) gains in top-tier gaming performance this year. That leaves Arrow Lake's higher performance in productivity applications and drastic power reductions. Intel claims up to a 165W decline in total system power in gaming, resulting in 10C lower operating temperatures on average as the key Core Ultra 200S key selling points. The Core Ultra 200S series has plenty of other innovations, too. Arrow Lake has new architectural design elements for desktop PCs, including a five-chiplet design that leverages cutting-edge Foveros 3D packaging in tandem with a radically redesigned compute tile that intersperses Skymont E-core clusters among the Lion Cove P-cores. Intel also discarded Hyper-Threading and support for both DDR4 and DDR5 memory like it had with the prior-gen Raptor Lake processors, instead going with only DDR5 while adding support for the new CUDIMM DDR5 modules that boosts easily attainable memory overclocking speeds to DDR5-8000 (and beyond). Intel also has new overclocking knobs for enthusiasts and a new 800-series chipset. Besides these desktop parts, Intel announced that its Core Ultra H and HX series for enthusiast-class laptops are coming to market in the first quarter of 2025. There's a lot of ground to cover. We'll start with the SKUs and move directly to the performance claims, then cover the new architectural components. Intel has switched to the same 'Core Ultra' branding it uses for the mobile market but uses the 'S' suffix to differentiate the desktop models. Intel begins the series at '200S' instead of '100S,' which a representative said "makes sense" without elaborating. Considering the Core Ultra 100-series was Meteor Lake, we assume it was a reference to Arrow Lake having more in common with the Lunar Lake 200-series parts. The five Arrow Lake SKUs slot into the Ultra 9, 7, and 5 families with 24, 20, and 14 cores, matching their prior-gen counterparts. However, the P-cores no longer support Hyper-Threading so the total thread counts are lower. Intel says it has actually increased performance in multi-threaded workloads despite the removal of Hyper-Threading. Intel has the standard overclockable K-series models along with two KF-series chips that come without the integrated graphics engine, so you can save some cash on the two lower-end models if you plan to use a discrete GPU. Intel doesn't have a KF option for the Ultra 9 285K, but a representative said that could be an option in the future. Intel's Ultra 9 weighs in at $589, matching the launch day pricing of its previous-gen counterpart. The $394 Ultra 7 265K debuts for $15 less than the previous-gen Core i7-14700K it replaces, while the $309 Core 5 245K is $10 less than the prior-gen Core i5-14600K. These chips line up against the same competing AMD Ryzen 9000 processors as their prior-gen counterparts did, as outlined in the above table (detailed specs are also in the prior album). Intel has dropped peak clock speeds across the board, with the Ultra 9 dropping weighing in at 5.7 GHz, 300 MHz less than the prior gen, while the Ultra 7 and 5 get 100 MHz reductions in boost speeds. However, Intel has adjusted P-core base clocks upwards by 500 to 700 MHz. The E-cores have also seen 200 to 600 MHz improvements in boost clocks, and 600 MHz to 1 GHz improvement in base clock speeds. Intel supports up to 192GB of DDR5-6400 with DDR5 CUDIMMs, a new type of DIMM that comes with an integrated clock redriver that boosts easily-attainable stable clock frequencies by stabilizing the data eye. Intel also points to much higher overclocking headroom with CUDIMMs and says DDR5-8000 appears to be the sweet spot. We're awaiting more detailed memory specs, such as the base memory frequency with standard DDR5 DIMMs and clock speeds based on the number of DIMMs per channel (as before, the processors support dual channel). Arrow Lake does support ECC memory, but it won't be supported on consumer platforms -- instead, that is reserved for enterprise systems. Despite Arrow Lake's claimed lower operating power consumption, the chips still come with similar maximum TDP ratings of 250W for Ultra 9 and 7 (3W less), and 159W (22W less) for the Ultra 5. Intel says the lower power consumption occurs during normal workloads with an up to 40% reduction in package power consumption, which we'll expand on below. Intel also increased the maximum CPU temperature (TJMax) to 105C for Arrow Lake, a 5C higher limit. All three K models come with 24 lanes of PCIe 5.0, with an additional 20 lanes of PCIe 4.0 provided by the chipset. The Ultra 9, 7 and 5 all come with the same Xe-LPG graphics engine with 4 Xe cores as the Meteor Lake chips, not the newer Battlemage engine found in Lunar Lake, with the Ultra 9 and 7 sporting a 2.0 GHz boost, while the Ultra 5 drops to 1.9 GHz. Intel says the iGPU offers twice the performance of the graphics on the 14th-gen processors. The chips also feature the same NPU engine for AI acceleration as Meteor Lake, not Lunar Lake. This engine provides up to 13 TOPS of INT8 throughput, which doesn't meet Microsoft's CoPilot+ requirement of 40+ TOPS. Intel says it decided to go with the smaller engine to balance the die area requirements -- a larger engine would chew into the space available for other additives, like cores and cache. The chips drop into the LGA 1851 socket, so they are not compatible with existing motherboards. Existing coolers should be compatible with the requisite mounting hardware, but the need for a kit could vary by vendor. Unlike AMD, Intel won't commit to using the LGA 1851 socket for future processor generations, simply saying that it doesn't comment on potential future products. Here we can see Intel's gaming performance claims against AMD's Ryzen 9000, but as always, view these vendor-provided benchmarks with caution. Intel used its Application Optimizer (APO) software for these tests, so results will be lower without that software enabled. We provided the test notes in an album at the end of the article. We'll also explain the cause of the lower-than-expected gaming benchmarks in the architecture section. Intel says its new flagship Core Ultra 9 285K matches AMD's flagship Ryzen 9 9950X in gaming, but AMD's chip is about 10% slower than Intel's current-gen Core i9-14900K, so the new flagship Core Ultra 200S series should be slower in gaming than Intel's prior-gen 14900K flagship. Making matters even more challenging for Intel, AMD's fastest specialized gaming chip, the Ryzen 7 7800X3D, is 15 to 20% percent faster in gaming than the 9950X in our tests. However, Intel says the 285K will only lag the 7800X3D by '5 to 7%,' which doesn't add up based upon real-world testing of the 7800X3D and its relative positioning to the 9950X. As you can see, the 285K trails the 9950X by up to 13% in Cyberpunk 2077, and also trails in four other titles. Intel's benchmarks show the 285K matching 9950X in 6 titles, while beating it in four additional titles. However, the large +28% advantage in Total War: Warhammer III is in the Mirrors of Madness benchmark that stresses the CPU cores in an almost synthetic type of workload. The size of this single outlier also influences the average performance heavily, so we could see the 285K come in slightly lower overall than the 9950X in various reviewer test suites. Intel also provided a very narrow selection of five gaming benchmarks against the 7950X3D, showing that it significantly lags in two of the benchmarks while having a 15% advantage in one title. We expect the deltas will be in the 7950X3D's favor in a larger range of titles. In the same slide, Intel points to its strong performance advantage over the 9950X3D in a range of heavily-threaded productivity workloads, like Blender and Cinebench, as a proof point of its balanced level of performance. Intel also teased the Ultra 7 265K as offering roughly 5% less performance than the 14900K, but at up to 188W lower power consumption and 15C lower temperatures (geomean). Intel didn't provide any benchmarks to substantiate those claims. Intel is also keen to point out its lower power consumption metrics in gaming, which we'll cover below, but only included comparisons to its own prior-gen chips, not AMD's Ryzen 9000. Overall, Intel claims the Core Ultra 9 285K lowers the CPU package power consumption by up to 58% in lightly-threaded workloads over the 14th-gen models. Intel used several productivity workloads to illustrate the lower package power consumption, but some of those benefits will also be applicable to some games as well. Intel says it pairs the lower power consumption with up to 19% more performance in multi-threaded workloads, and can deliver the same level of performance at half the power consumption of the 14900K. For gaming, Intel presented a range of games with the Ultra 9 285K compared to the 14900K, showing similar performance yet with drastically lower power consumption values when measured as a complete system. For the system power measurements, Intel used values from an entire system, comparing like-for-like with a prior-gen 14900K with a Z780 motherboard against Core Ultra 9 285K system with an 800-series motherboard, citing up to 165W less power consumption in Warhammer Space Marines 2. Again, Intel's performance claims of performance parity against the 9950X don't entirely line up here, as the 14900K is roughly 10% faster than the 9950X, yet Intel shows the 285K matching the 14900K in these benchmarks. Across seven titles, Intel says the 285K delivers a geomean of 73W less power consumption than the 14900K. This correlates to up to 15C lower CPU package temperatures during gaming, which will then ease cooling requirements. The company also claims the Ultra 9 285K delivers the same gaming performance (across seven titles) at 125W as it does at 250W, but presented all of its other benchmarks with the 250W power limit. Perhaps most importantly, Intel avoided gaming power consumption comparisons to AMD's Zen 5 Ryzen 9000 lineup, which AMD also positions as being far more power efficient than the company's Zen 4 products. For Intel to make power efficiency a convincing selling point it will have to show a tangible lead over the Ryzen processors. Intel did point to a power efficiency advantage over the Ryzen 9 9950X in multi-threaded productivity applications, but we'll have to wait for independent benchmarks to see how the chips compare in gaming. Intel was more forthcoming with productivity application benchmarks. As you can see across a wide range of benchmarks above, Intel highlights the 285K has faster performance in multi-threaded workloads than the 9950X, despite the fact that it removed Hyper-Threading from the Core Ultra series. If true, that means Intel provides more performance with 24 threads than Ryzen does with 32 threads. Overall, Intel claims a 15% gen-on-gen improvement in threaded work, and a 13% advantage over the 9950X. Intel also says the processor delivers up to 36 TOPS of total INT8 performance in AI workloads, with 8 TOPS from the GPU, 13 TOPS from the NPU, and 15 TOPS from the CPU. Notably, Intel chose the Xe-LPG graphics engine instead of the newer Battlemage architecture, so the iGPU supports DP4a and doesn't come with the more performant XMX engines. Intel used a very similar package design to its five-tile previous-gen Meteor Lake laptop processors instead of the newer Lunar Lake design. However, Intel roped in the newer Lion Cove P-core and Skymont E-core microarchitectures for the compute tile instead of the Redwood Cove and Crestmont cores used in Meteor Lake. The Arrow Lake design employs a compute tile (chiplet) fabbed on the TSMC N3B process node, a GPU tile with the N5P node, while the SoC and I/O tiles use TSMC's N6 process. Intel uses Foveros 3D packaging to mount those tiles to an underlying base tile fabbed on the Intel 1227.1 process node. There's also a 'dummy' filler tile to provide mechanical rigidity. Intel's decision to split the memory controller and PHY into their own tile (I/O tile) was to improve yields, but this creates memory latency issues that contribute to the lower gaming performance. Intel does offer the option to overclock the tile-to-tile interface, but we'll have to see how that plays out in real-world testing. Intel also redesigned its core layout, which now has quad-core E-core clusters interspersed among the P-cores. Intel used to place all E-cores in their own dedicated block, but says it spread the cores out to reduce hotspots for this design. Intel also connected the E-cores to the 36MB L3 cache, so they now share L3 with the P-cores for the first time. The P-cores and E-cores still have their own dedicated L2 caches, with 3MB for the P-cores (a .5MB increase over the prior gen), and 4MB of L2 shared among each E-core cluster. Intel says the totality of the new design yielded a 33% reduction in package size and said the design allowed it to quickly port innovations from Lunar Lake to Arrow, thus resulting in launching the two chips a mere month apart. We've included the slides detailing the advances of the P-core and E-core architectures, but we've already covered these microarchitectures in-depth in our Luner Lake deep dive, which you can read here. Overall, we're looking at a 9% increase in IPC over Raptor Lake (lower than the 14% cited with Lunar Lake because Intel did that comparison to Meteor Lake), and a 32% increase in integer IPC and 72% increase in floating point IPC for the E-cores. Finally, we have details on the 800-series chipsets and overclocking. The tile-based architecture brings several new overclocking features, such as the ability to overclock the tile-to-tile interface, which should help at least partially address memory latency issues that stem from the remote I/O tile. Intel increased clock granularity and now allows increasing or decreasing the clock rate in 16.6 MHz increments. Arrow Lake also now has two adjustable base clocks due to the tiled architecture. The compute tile and SoC tile each have their own adjustable base clock. Intel employs a DLVR power delivery subsystem for the processor to provide granular per-core control of voltage, but overclockers can disable this feature to improve overclockability. On the topic of core overclocking, Intel says that the P-cores will not have much frequency headroom -- the chip is designed to expose peak performance with out of the box settings. However, the E-cores should have more headroom for tuners. Intel also has a one-click AI-driven overclocking feature in XTU, just as it has had in the past. The 800-series which is largely a refresh from the 700-series chipset. The chipset provides 24 PCIe 40 lanes, a robust complement of USB interfaces, Thunderbolt 4, 1GbE, along with built-in Wi-Fi 6E and Bluetooth 5.3. Support for Thunderbolt 5, Wi-Fi 7, 2.5 GbE and Bluetooth 5.4 will come from discrete chips on selected motherboards. Intel's gaming benchmarks are somewhat questionable, especially in regard to whether or not the Arrow Lake processors are faster than Intel's own current-gen models. Given those disparities, its also hard to determine the relative positioning compared to AMD's Ryzen 9000 series -- it is possible the Ryzen 9 model is faster than intel's flagship, if only by a small margin. That could breathe new life into AMD's Zen 5 lineup, but it will also means that Intel's current-gen Raptor Lake Refresh chips could be the chips to buy for gamers for quite some time. Intel is making a case that drastically improved power consumption in gaming will be a big draw for enthusiasts, but we'll have to see how that compares to Ryzen 9000. Intel avoided those comparisons, but we'll be sure to provide the insight with our review. Intel does appear to have made inroads in performance in productivity applications, and that will definitely attract the more productivity-minded among us. As always, the proof will be in the shipping silicon. If tradition holds, reviews will go live on October 24.

Intel's Arrow Lake CPUs keep AI simple to make gamers happy

Gamers don't see the need for AI. So Intel went with an older NPU for Arrow Lake. Intel's new Arrow Lake architecture, aka the Core Ultra 200S series, brings AI capabilities onto Intel desktops. But the chip doesn't use the Copilot+ capabilities of Intel's mobile Lunar Lake chip -- its designers used the older NPU found on Meteor Lake instead. For now, this means that if you buy an Arrow Lake chip, you won't be able to use it with some of the new AI enhancements found within Windows 11's 2024 Update, like generative AI and the controversial Recall function. Arrow Lake's optimized NPU 3 only provides 13 TOPS, while Microsoft set 40 TOPS as the bar for Copilot+ status. Intel still plans to ship more than 40 million AI PCs in 2024, using the vaguer "AI PC" definition that accompanied the launch of Meteor Lake. Arrow Lake is the not the first desktop PC architecture to include an NPU for AI. Last January, AMD announced the Ryzen 8000 series of desktop processors with an NPU capable of 39 TOPS. However, the blink-or-you'll-miss-it Ryzen 8000 was quickly replaced by the Ryzen 9000 processor six months later in June -- without an NPU, but ushering in Zen 5 with a potent 16 percent performance improvement over the prior generation. Neither the Ryzen 8000 nor Arrow Lake meets the threshold to be called a Copilot+ PC. Arrow Lake, like Meteor Lake, is a disaggregated architecture -- a fancy name for a modular design. Theoretically, couldn't Intel have just made more of the 45 TOPS NPU 4 found within Lunar Lake, and added it to the Arrow Lake package? Yes...and no, Intel executives said. First off, Arrow Lake's NPU is the same NPU 3 as Meteor Lake, Robert Hallock, an Intel vice president and general manager of client AI and technical marketing at Intel, told reporters. "So we've had a lot of time to learn it and optimize it, and it made sense that we would fit this in," he said. (Meteor Lake's NPU generated 11.5 TOPS, while Arrow Lake's optimized version produces 13 TOPS.) However, Intel's customer base of enthusiasts also indicated that they didn't want to give up certain features, like a potent GPU, to check the box of AI. And Intel, which faced pressure to deliver Arrow Lake on time, found it easier to just use an established design. "We actually had a long, a really long chat internally about how to allocate the transistor budget on this part," Hallock continued. "To be clear, yes, it was fully possible to put a 50 TOPS, 40 TOPS NPU on this product, but to do so would also require shrinking the core count, changing the GPU core count. You start making sacrifices in sort of fundamental performance dimensions that enthusiasts really care about -- that didn't feel like the right mix. And we also talked at length about sort of the enthusiast market's disposition on AI as a whole. And I think it's fair to say it's somewhat reluctant." Instead, Intel believes that software developers aren't always using AI hardware effectively, that a combination of components (CPU, GPU, and NPU) is better than an NPU alone, and various AI models are being effectively compressed small enough that they don't need a massive NPU. All told, Arrow Lake-S (the Core Ultra 200S series) has a total of 36 platform TOPS. Although Arrow Lake has an integrated GPU, many customers will pair the chip with a discrete GPU, which provides far more AI horsepower than just the NPU alone. "We've proven that Meteor Lake's 13-TOPS NPU is more than enough," Hallock said. "We've got enthusiast users who are somewhat skeptical, and so we sized an NPU that kind of fits all of those constraints while preserving and protecting the CPU performance that people care about most and still gives enough AI to embrace workloads that are coming down the pipe." Interestingly, Intel will debut two mobile chips in the Arrow Lake family during the first quarter of 2024. One, code-named Arrow Lake-HX, will essentially be a mobile copycat of the Core Ultra 200S chips that Intel will begin shipping in a few weeks. A second family of Arrow Lake-H chips for notebooks provides many more TOPS overall than Arrow Lake-S or -HX: 99 platform TOPS in all. However, those will come from a beefed-up GPU that uses XMX extensions and more Xe cores. The NPU will still provide 13 TOPS. Roger Chandler, vice president and general manager for enthusiast PC and workstation product marketing at Intel, reiterated what Hallock said -- that software developers aren't using the full capabilities of the existing NPUs effectively. Intel's goal, he said, was to offer a "balanced platform." "When I look at AI right now, we're like, 10 seconds into a 20-hour movie," Chandler said.