This week Intel held its annual Architecture Day and even for an event that often previews major upcoming changes to the company’s silicon, this year Intel is hailing its new advancements as some of the “biggest shifts in a generation” across its CPUs, GPUs, and more.
When it comes to consumer devices, the big new product lines to pay attention to are Alder Lake—Intel’s next-gen CPU architecture—and Alchemist—which is the codename for Intel’s upcoming consumer graphics card and the first product in Intel’s newly branded Arc line of high-performance gaming (HPG) products.
However, before we talk about new CPUs, we have to talk about the design of Intel’s new processors themselves, as the big changes to Intel’s core (no pun intended) architecture is what’s really paving the way for much of Intel’s performance and power efficiency gains.
Instead of simply packing more cores onto a smaller die, for the first time, Intel is switching to a new hybrid architecture comprised of faster performance cores augmented by energy-sipping efficiency cores, very similar to what many ARM chip makers have been doing for years on mobile devices.
Previously codenamed Golden Cove, Intel says its new Performance-core microarchitecture features more decoders (up from 4 to 6), more execution ports (up from 10 to 12), bigger register files, and improved branch prediction. All told, compared to current 11th-gen Cypress Cove processors running at the same frequency, Intel says its new Golden Cove performance cores offer around 19% better general performance. Additionally, thanks to the addition of Intel’s new Advanced Matrix Extensions, those cores are also getting a big boost in AI processing.
And as you’d expect from a hot new CPU architecture, Alder Lake is also designed to better support new component standards including DDR5 RAM, PCIe Gen 5, Thunderbolt 4, Wi-Fi 6e, and more.
Meanwhile, complementing its performance cores, Intel’s new Efficient-core microarchitecture (previously codenamed Gracemont) is said to support enhanced multitasking thanks to improved multithreaded performance with what Intel claims is its most efficient x86 microarchitecture yet. In short, Intel says its new efficiency cores provide 40% more single-threaded performance at the same power levels or delivers similar performance while using 40% less power. And when compared to Intel’s Skylake chips for overall throughput, four of Intel’s new efficiency cores are said to provide 80% more performance compared to two Skylake cores running four threads.
Also, to better manage workloads between its performance and efficiency cores, Intel is introducing its new Thread Director scheduler, which provides hardware-based support for dynamically assigning processes to the proper core for the job.
As for Alder Lake itself, new chips will be built on the Intel 7 process, with three different major designs intended to cover desktops, high-performance laptops, and a range of thin, low-power devices. New Alder Lake chips will be available with a varying number of cores and threads with the most powerful CPUs topping out at 16 cores, 24 threads, and 30MB of LL cache.
When it comes to Alchemist, Intel didn’t dive quite as deep into the architecture behind its upcoming GPU but it did reveal a number of important features and design points. One of the biggest revelations is that instead of fabricating Alchemist’s chip itself, Intel is relying on TSMC’s N6 node to produce the silicon for its upcoming graphics card, signaling a big change for a company that has previously relied heavily on its own in-house fab.
Furthermore, instead of focusing on the number of execution units, Alchemist’s architecture is based on what Intel is calling XE cores, which will be comprised of 16 vector engines and 16 matrix engines, with additional ray-tracing units that support both DirectX Raytracing and Vulkan Ray Tracing.
Thanks to updated graphics driver components and memory managers/compilers, Intel says Alchemist can deliver 15% (or up to 80% better in some cases) throughput for CPU-bound games, while also improving load times by as much as 25%.
Alchemist already supports Unreal Engine 5 and DirectX 12 Ultimate and Intel also showed off its own super sampling tech called XE SS, which is designed to improve graphics quality or graphics performance by up to 2x in ideal conditions. Intel even says the SDK for XE SS will be available to developers starting this month.
When moving beyond traditional consumer computing products, Intel is also making some big waves with the introduction of its new IPU (infrastructure processing unit) and Mount Evans being Intel’s first ASIC IPU. And for enterprise graphics, Intel also detailed its new Ponte Vecchio GPU, which Intel claims offers industry-record AI performance and up to 45 TFLOPS of FP32 throughput, with the chip based on TSMC’s most advanced N5 process.
With Intel’s CPUs now on the back foot when compared to AMD’s current chips when it comes to both performance and value (assuming you can find them in stock), the company needs Alder Lake and Alchemist to be a hit if it’s going to reclaim silicon superiority.