Intel Architecture Day 2021: Alder Lake CPUs to arrive this fall

Intel's Alder Lake will utilize a hybrid architecture

Intel pulled the curtain back on its upcoming "Alder Lake" chips during a virtual Architecture Day presentation. The company's 12th Generation CPUs will eschew the current framework of providing high-performance cores in favour of a hybrid approach.

Alder Lake will have both low-power and high-power x86 cores. The new performance cores are known as "Golden Cove" are the company's most advanced cores yet. The efficiency cores called "Gracemont" is positioned as the world's most efficient x86 CPU Core.

Intel says that one of its efficiency cores has 40 percent more performance at the same power of a Skylake core (6th Gen), these double when put up against four of these running four threads with Skylake's two running four threads.

These CPUs mirror mobile SoC's where it's common to have a mix of high-performance and low-performance cores. Intel has arranged the cores in logic blocks that differ according to various product segments.

Intel will abbreviate the performance cores with "P" and the efficiency cores will be "E." The new generation of Alder Lake CPUs will also be highly scalable ranging from 9W to 125W to cover most of today's mobile and PC segments.

The desktop version of the CPU will use 8 "E" cores, paired with 8 "P" cores on a new LGA1700 socket. This design will have 30MB of cache memory. These will also support all modern standards of memory (DDR5, DDR4, LPDDR5, LPDDR4X).

To manage the P and E cores for optimal performance, Intel has a new hardware scheduler called Thread Director. It will work in tandem with Microsoft's Windows 11 to assign cores based on real-time calculations.

While Intel recommends Windows 11 for optimal performance, Windows 10 and Linux are still supported.

Besides the new CPUs, Intel also teased its new "Alchemist" GPU which is expected to launch in early 2022. The first generation high-performance GPU will support DirectX 12 Ultimate and hardware Ray Tracing.

It will also feature Intel's version of AI upscaling, similar to AMDs FidelityFX and Nvidia's DLSS. Intel's version will be called "XeSS" or Xe Super Sampling.