Intel’s Arrow Lake processors, which will soon be physically released to end users, represent a significant evolution for Intel and are intended as a response to AMD’s Zen 5 chips. These processors, which belong to the 15th generation of Core processors, will be available in desktop (Arrow Lake-S) and laptop (Arrow Lake-H) variants and offer significant performance improvements compared to their predecessors. Arrow Lake is based on a new combination of performance and efficiency cores, known as “Lion Cove” for the performance cores (P-cores) and “Skymont” for the efficiency cores (E-cores).
Overview and specifications
The desktop models of the Arrow Lake-S series will run on the new LGA-1851 socket and will (initially) be combined with Z890 motherboards, while the smaller chipsets will certainly follow later. There will be different power levels, including 125 W, 65 W and 35 W TDP variants. The Core Ultra 9 285K, the top model, reaches a clock speed of up to 5.7 GHz, while lower models such as the Core Ultra 5 245KF reach 5.2 GHz. The processors are equipped with large caches: The top model has 40 MB of L2 and 36 MB of L3 cache to maximize memory bandwidth. Here is an overview of the first 4 CPUs, with more to follow to complete the portfolio.
More efficient coexistence of P and E cores and their upgrading
With the Arrow Lake generation, Intel is increasingly focusing on improving efficiency instead of just maximizing performance. This is made clear by the increased use of efficiency cores (e-cores). While previous generations of Intel’s Core processors, such as Raptor Lake, focused on high clock frequencies and performance cores, Arrow Lake uses a more balanced distribution of performance (P-cores) and efficiency cores. In the top configuration, for example, the Core Ultra 9 offers 8 P-cores and 16 E-cores. The aim is to increase performance in multi-threaded applications without increasing energy consumption excessively.
Another example of this shift towards efficiency is the use of new “Skymont” E-cores, which are designed to offer higher performance per watt. Intel is focusing on improving multi-core performance through more efficient cores, rather than simply increasing the number of P-cores or maximizing clock speeds. This is intended to achieve a better balance between high single-thread performance and energy-efficient multitasking.
In addition, the Arrow Lake processors exclusively support DDR5 memory, which leads to higher memory bandwidth and efficiency, while at the same time the old DDR4 standard is being phased out. This is intended to reduce energy consumption throughout the system while still achieving high performance levels. In general, Intel is now pursuing an approach with Arrow Lake that focuses on both efficiency and performance in order to achieve the best balance, particularly in applications such as gaming and productive workloads, and that is a good thing. Another highlight is the support of Thunderbolt 4, but Arrow Lake remains cautious when it comes to AI features: the integrated Neural Processing Unit (NPU) only offers 13 TOPS of performance, which is less impressive compared to the competition. More on this in a moment.
Graphics unit with Xe LPG
For the integrated graphics, Intel relies on an improved version of the Arc Alchemist GPUs, with double the number of shader units compared to previous generations. Nevertheless, the integrated graphics are not expected to be able to compete with dedicated GPUs. It is worth noting that the CPU tiles will be manufactured by TSMC using the 3nm process (N3B), while some models could also use Intel’s own 20A production.
Intel’s new Xe-LPG architecture, which is used in the Arrow Lake processors, is therefore not a new development, but a further development of the integrated graphics solutions that were previously introduced in the Alchemist GPUs of the Arc series. Xe-LPG is based on a scaled-down version of the dedicated Arc GPUs and offers a compact architecture for integrated graphics designed for efficiency and moderate performance. Xe-LPG in Arrow Lake S is said to double the number of shader cores compared to the previous generation, which should significantly increase graphics performance. These improvements also include optimized rendering techniques and more efficient use of memory bandwidth, which has a positive effect on the execution of multimedia applications.
A special feature of the Xe-LPG architecture is the support of modern graphics APIs such as DirectX 12 and Vulkan, which optimize the use of the hardware for current games and software. The integration of graphics accelerators and advanced multimedia functions into the architecture makes it particularly interesting for systems that require an energy-efficient graphics solution without relying on dedicated GPUs. This should also improve the overall performance of data-intensive workloads by reducing the CPU load. Despite these improvements, the architecture is not designed to compete with dedicated graphics cards, but rather aims to provide solid performance for light graphics tasks and everyday applications, which should go beyond a pure “tick-box” solution and provide real added value for multimedia and light gaming applications.
Foverors Packaging as a link between the different manufacturing technologies
For the Arrow Lake processors, Intel generally relies on Foveros Packaging, a 3D chip stacking technology that integrates several semiconductor die components in a single package and enables the vertical arrangement of different chips. This technology makes it possible to combine different manufacturing processes within one processor, for example by using TSMC’s 3nm technology for some tiles and Intel’s 20A technology for others. Foveros technology reduces the communication paths between the cores and other functional units, which reduces latency and increases the data transfer rate, while thermal efficiency is increased due to the proximity of the components. This enables higher clock speeds and more stable performance. In Arrow Lake, Intel thus combines the CPU cores, the graphics engine and possibly AI accelerators on a single chip to optimize performance and energy efficiency. The technology allows for flexible manufacturing and provides a balance between performance, power consumption and cost, which is particularly beneficial for multitasking and data-intensive applications.
21 Antworten
Kommentar
Lade neue Kommentare
Mitglied
Veteran
Mitglied
1
Urgestein
Urgestein
Veteran
Urgestein
Urgestein
Mitglied
Mitglied
Veteran
Veteran
Mitglied
Urgestein
Urgestein
Mitglied
Mitglied
Mitglied
Alle Kommentare lesen unter igor´sLAB Community →