Linus Torvalds, the creator of Linux, has once again voiced strong criticism of x86 microarchitecture levels, dismissing them as fundamentally flawed and counterproductive.
The remarks surfaced during discussions around recent updates to the Linux KConfig build system, where Torvalds objected to how functional capability levels are defined for the AMD64 (x86_64) architecture.
“Let’s not get into x86-64 microarchitecture levels. It’s a silly place. The whole ‘v2’, ‘v3’, ‘v4’ etc naming seems to be some crazy glibc artifact that is very silly and should die. I don’t know who came up with the ‘microarchitecture levels’ garbage, but as far as I’m concerned it’s completely unofficial and a totally broken model.”
The x86-64 microarchitecture levels were originally proposed in 2020 by Florian Weimer of Red Hat. The initial proposal introduced levels A, B, and C, later evolving into the now-familiar v2, v3, and v4 model widely referenced in toolchains and distributions.
Torvalds argues that these levels introduce unnecessary abstraction rather than clarity. According to him, attempting to linearize CPU feature evolution ignores real-world hardware behavior, where instruction sets appear, evolve, and sometimes disappear across generations.
A notable example is AVX-512, which debuted in Intel’s 11th Gen Rocket Lake processors but was later removed from multiple subsequent consumer CPU generations—highlighting the fragility of rigid capability tiers.
🏭 Intel’s Strategic Uncertainty Amid Leadership Transition #
While architectural debates continue in the open-source community, Intel faces far more immediate internal challenges.
At the UBS Global Technology Conference, Intel interim co-CEO David Zinsner emphasized that the company’s overarching strategy remains intact despite executive turnover.
“The board is very clear that the core strategy remains unchanged.”
Zinsner reaffirmed Intel’s commitment to Intel Foundry Services (IFS) alongside its internal chip design efforts, stressing that both must succeed in tandem.
“We still want to be a world-class foundry and a Western supplier of cutting-edge chips… but for the Foundry to succeed, its number one customer—Intel’s own products—must succeed.”
Intel currently trails competitors such as TSMC and NVIDIA, particularly in AI-focused silicon. Zinsner acknowledged execution gaps in chip design and cross-organization optimization, noting that some systemic issues remain unresolved.
He also indicated that the next CEO will need deep expertise across both foundry operations and product development to restore Intel’s competitive standing.
👔 Lip-Bu Tan Emerges as a Potential CEO Candidate #
Industry attention has turned to Lip-Bu Tan (Chen Liwu) as a possible successor to lead Intel.
Tan previously resigned from Intel’s board following disagreements with former CEO Pat Gelsinger, reportedly over the company’s turnaround strategy, organizational bloat, bureaucratic inertia, and a lagging AI roadmap.
A seasoned semiconductor executive, Tan is widely regarded as a technically grounded leader capable of reshaping Intel’s culture and strategic execution.
If appointed, Intel would join NVIDIA, TSMC, Broadcom, and AMD in being led by executives of ethnic Chinese background—an increasingly notable trend in the global semiconductor industry.
🧠 Conclusion #
Torvalds’ criticism underscores a broader tension between software abstractions and hardware reality, especially as CPU architectures grow more fragmented. At the same time, Intel’s leadership uncertainty and competitive pressure highlight how deeply execution and architectural clarity matter—not just in kernels, but across the entire semiconductor ecosystem.