AVX-512 returns to Intel’s consumer CPUs with Nova Lake, six years after being dropped

Intel is bringing AVX-512 support back to its consumer CPUs with the upcoming Nova Lake architecture, according to Linux kernel patches and multiple independent sources, ending a six-year hiatus that left AMD with a feature advantage on client platforms.

The AVX-512 instruction set enables 512-bit vector operations, delivering significant performance improvements for workloads in scientific computing, machine learning inference, media encoding, and data processing. Benchmarks on AMD’s Zen 5 architecture show AVX-512 delivering up to 43% higher throughput compared to standard AVX, depending on the number of source operands.

A complicated history. Intel’s relationship with AVX-512 on consumer chips has been fraught. The instruction set was last supported on Tiger Lake (11th Gen, 2020). When Intel introduced its hybrid P-core / E-core architecture with Alder Lake (12th Gen), the P-cores supported AVX-512 but the E-cores did not, creating an asymmetry the company chose to resolve by disabling the feature entirely. Motherboard makers were instructed to block AVX-512 support through BIOS updates, even on chips that physically supported it.

Subsequent generations, Raptor Lake, Meteor Lake, Arrow Lake, and Panther Lake, all shipped without AVX-512, while AMD offered the feature across both its Zen 4 and Zen 5 consumer lineups.

If you believe trustworthy journalism still matters, we'd appreciate your support.

Help us grow

Nova Lake changes that. The new architecture, expected in the Core Ultra 400 series, uses Coyote Cove P-cores and Arctic Wolf E-cores with AVX 10.2 ISA support. The AVX 10.2 specification, which Intel previewed last year, brings AVX-512 capabilities to both core types, resolving the hybrid asymmetry that originally forced the feature’s removal. Linux kernel patches confirm the addition of an AVX-512 optimized path for Nova Lake, and multiple leakers have corroborated the listing.

The return of AVX-512 on Nova Lake, combined with the anticipated return of SMT (simultaneous multithreading) on the Coral Rapids Xeon family, signals a strategic reversal on two architectural decisions that have drawn criticism from power users and enterprise customers.

Source: Tom’s Hardware, Wccftech, VideoCardz

Scroll to Top