Nearly a year after its introduction at Intel Innovation 2023, Intel has finally revealed the full specifications of its new Intel Core Ultra 200S desktop processors, codenamed Arrow Lake. The models are designed for enthusiast desktop and laptop computers; It has an architecture focused on energy efficiency and an NPU for artificial intelligence.
By the way, this is the first processor for desktop computers to have a unit dedicated to artificial intelligence. The developer reports that it managed to reduce the chips’ energy consumption by up to 30% while increasing multi-threading performance by 10%.
Arrow Lake’s “new” architecture
Before going into details about the Arrow Lake architecture, Intel protected the packaging of the CPUs with Foveros 3D vertical stacking technology, reducing the package size by 33% compared to the 14th generation. So, in addition to the Base Tile, which is responsible for assembling all other parts, the chips also have Compute Tile, Filler Tile, I/O Tile, SoC Tile, GPU Tile, and IHS (Integrated Heat Dissipation).
Lion Cove and Skymont kernels on desktops
Compute Tile is the flagship of the new processors, featuring up to 16 Skymont efficiency cores and 8 Lion Cove performance cores. Introduced in Lunar Lake laptop CPUs, these cores are reaching desktops for the first time, but they also bring some improvements and improvements.
Called P-Core, Lion Cove offers greater single-thread optimization and improvements in performance per watt. According to the company, these cores are crafted to perform at maximum performance states and carry up to 36 MB of L3 cache; This is basically the same as what older generations had.
Worked on the Lion Bays for a better PPA (Performance Power Fields) that creates the best energy balance configuration. These cores also feature AI instructions that target the right cores for specific tasks, with up to a 9% increase in IPC (Instructions Per Cycle).
Skymont’s E-Cores, on the other hand, are on a mission to expand coverage when running complex workloads by increasing multi-threading performance. Considering improving its artificial intelligence capabilities, Intel doubled the vectorization and efficiency for VNNI (Vector Neural Network Instructions) with the AVX-512 instruction set.
Among other announcements, Skymont cores improve faster ingestion of instructions as well as deepening the CPU’s parallelism capabilities, allowing more of those instructions to be executed simultaneously.
Increase in IPC and more Cache
An increase in IPC is also included with this type of core, demonstrating 32% more single-thread performance in INT variant and up to 72% in FP compared to previous generation Gracemont cores. In multi-threading there is the same gain in INT, but in FP the performance increase is 55%.
Cache levels haven’t changed much, but they do have a significant difference. Intel will maintain 36 MB for the shared cache at the L3 level, while at the L2 level there will be 3 MB per P-Core and 4 MB per E-Core group.
Announced with the launch of 12th generation Alder Lake processors, Thread Director returns to Arrow Lake. The blue team worked to reduce consumption by 30% and increased system bandwidth based on increasing P-Core telemetry and improving E-Core prediction through a new prediction model.
Improved overclocking
For overclocking enthusiasts, the arrival of the Intel Core Ultra 200S could be a big deal. The company states that there will be a frequency increase every 16.6 MHz in E-Cores and P-Cores. In fact, the company has introduced a dual-base clock system that allows independent customization for all the different tiles available on the processor.
Other features include Tile-2-tile & Fabric overclocking, which allows static ratio to be applied, and DLVR, which is an option to bypass the system’s internal voltage manager. Finally, the company’s official overclocking application, Intel eXtreme Tuning Utility, will receive new automatic functions for users.
Integrated graphics card
One of Intel’s biggest expectations in recent years is about improving the quality of the company’s GPUs. With Arrow Lake, the company will retain the Xe microarchitecture introduced in Meteor Lake, but with twice the resources and greater energy efficiency.
The maximum implementation of the integrated graphics card used in the CPUs consists of 4 Xe cores and 64 Vector Engines, as well as four units for Ray Tracing and 4 MB L2 Cache. The numbers are low, but it’s worth remembering that all that iGPU isn’t designed to run games or complex activities.
Intel’s announcement reports that the new card has twice the power as well as 8 TOPS (Trillion Operations Per Second) for AI-related activities.
Artificial Intelligence and NPU
Speaking of AI, Intel is unveiling the 3rd generation NPU (Neural Processing Unit) in the Core Ultra 200S, making its first appearance in desktop CPUs. The technology has in-depth vectorization and more matrices for calculations, bringing a total of 13 TOPS; this number is significantly lower than the 48 TOPS offered by the Lunar Lake NPU.
Although very similar to the Meteor Lake trim, the unit in Arrow Lake has higher frequencies, further improving performance in real-life use.
In addition to the 8 TOPS available on the GPU, Arrow Lake has 15 TOPS for the AI thanks to the VNNI and AVX instructions of its P-Cores and E-Coreles. These instructions can improve application performance and increase real-time processing of AI-related tasks.
To exemplify the AI capabilities of the Intel Core Ultra 200S, the company says software like Audacity has broad compatibility with any of the three tiles. Users will be able to manually choose which accelerator to use (NPU, GPU or CPU) to perform different tasks.
On the gaming side, Intel presenters are simulating a livestream event with Cyberpunk 2077 using the XSplit application. In software, the entire effects section was switched from the GPU (RTX 4070 SUPER) to the NPU of the Arrow Lake processor via the OpenVino interface.
The result consists of applications such as a centralized webcam, removal of backgrounds and artifacts, and artificial improvements in scene lighting. Intel reported a 15% increase in the game’s frame rate with the switch from GPU to NPU.
SoC Tile
Moving on to the last part, the SoC panel of Arrow Lake processors supports 8K60 decoding with 10-bit HDR and 8K120 encoding with 10-bit HDR. The chips will also be compatible with VP9, AVC, HEVC, SSD and AV1 codecs.
DDR5 support only
Unlike previous generations, Intel Core Ultra 200S processors will only be compatible with DDR5 modules in UDIMM, SODIMM, CUDIMM, CSODIMM formats.
The company supports a maximum frequency of 6,400 MT/S and allows up to 192 GB combined in 48 GB modules. As expected, there is full support for Dual Channel and ECC technologies.
New 800 socket and chipset
Another long-awaited change for the new Arrow Lake generation is the change in the processor socket and motherboards. Now CPUs are available from Asus, Gigabyte, MSI, etc., initially represented by the high-end Z890 model. It uses the LGA-1851 socket with the 800 series motherboards announced by the manufacturers.
The chipset supports up to 24 PCIe 4.0 lanes and 20 PCIe 5 lanes; A total of 48 lanes are offered to the user. There is also compatibility with up to 8 SATA 3.0 ports, up to 32 USB 3.2 and 14 USB 2.0 connections.
Regarding wireless connectivity, the chipset has native support for Wi-Fi 6E networking (2.4 Gb/s) and integrated Bluetooth 5.3, but it is also possible to discreetly use Wi-Fi 7 and Bluetooth 5.4 technologies. This means that manufacturers can improve their systems with these latest connections, although this is not mandatory.
Arrow Lake-HX for laptops
Intel’s presentation provided more details about the arrival of Arrow Lake-HX processors, which use this nomenclature because they are models aimed at high-power laptops.
The company informs that these processors will have a new iGPU with Xe core and improvements in XMX technology for higher performance matrices. Lion Cove and Skymont cores will also receive updates.
Regarding GPUs, company executives say that the graphics card will be completely new and will be developed for artificial intelligence. The model should offer four times more performance in variants such as BF16, FP16, int8 and int4 thanks to its 8
In practice, this should return interesting numbers for the NPU and the entire AI system. Unlike its desktop counterpart, Arrow Lake-HX will have TOPS of up to 99, with 77 from the GPU, 13 from the NPU, and 9 from the CPU. For comparison, the Lunar Lake series offers 120 TOPS overall.
Performance
Finally, the company also introduced some performance graphics related to the Intel Core Ultra 200S, which aims to greatly reduce energy consumption. Internal benchmarks show that the top-of-the-range Intel Core Ultra 9 285K maintains the same frame rate as the Intel Core i9-14900K and reduces consumption by up to 165W.
In terms of games, the manufacturer is promising 15% increases in select titles like Sid Meier’s Civilization VI: Gathering Storm, but it’s 20% behind the Ryzen 9 7950X3D in Cyberpunk 2077. Despite good efficiency figures, the Arrow Lake generation achieves this. it shouldn’t offer players massive generational leaps.
Things are a bit more encouraging when it comes to content creation, as the flagship CPU outperforms the Ryzen 9 9950X by almost 80% in encoding 8K videos to 4K HEVC. However, there are modest improvements of around 6% in applications such as video and image editing.
availability
Intel Core Ultra 200S will be available from October 24 at a minimum price of US$ 294 (approximately R$ 1,600 in direct conversion).
Source: Tec Mundo
I am a passionate and hardworking journalist with an eye for detail. I specialize in the field of news reporting, and have been writing for Gadget Onus, a renowned online news site, since 2019. As the author of their Hot News section, I’m proud to be at the forefront of today’s headlines and current affairs.
