Hell Freezes Over as AMD and Intel Come Together for x86

Original Article ↗ Hacker News Discussion ↗

Hardware trust and management engines

  • Some want the ability to truly disable Intel ME and AMD PSP, viewing them as potential backdoors due to closed code with deep system access.
  • Skeptics argue even “disabled” firmware can’t be verified, echoing “trusting trust” concerns: if you didn’t build the stack yourself, you can’t be sure.
  • Others say you must either trust the platform or design systems so that untrusted platforms are isolated (e.g., behind a trusted firewall), but note firewalls rarely stop outbound “phone home” behavior.
  • A minority points to open hardware / open ISAs as the only real way to reduce this trust gap.

x86 vs ARM vs RISC‑V trajectory

  • One side thinks AMD–Intel cooperation signals shared defense of a weakening x86 against ARM’s rise in mobile, laptops, servers, and even supercomputers.
  • Others push back: x86 “death” has been predicted for decades; its backwards compatibility and software ecosystem should keep it relevant for at least another decade.
  • ARM is seen as already dominant in mobile/embedded and growing in servers and PCs; RISC‑V is viewed as more open but currently far behind in performance, mainly suitable for embedded use.
  • Several expect long‑term ISA plurality rather than a single winner.

Performance and efficiency debates

  • Heated debate over whether ARM can match “highest‑end” x86:
    • Some cite Apple M‑series and Snapdragon X as matching or beating top x86 in single‑core and efficiency.
    • Others note that multi‑socket x86 workstations and servers (Threadripper, EPYC) still vastly exceed any current ARM desktop SoC in total throughput.
  • Cloud ARM chips (Graviton, Ampere) are argued to be similar or slightly behind in raw perf/power, but meaningfully cheaper, making them attractive on a price‑normalized basis.

Cloud, servers, and supercomputers

  • Many note strong ARM inroads: AWS Graviton (cost advantage), ARM‑based supercomputers like Fugaku and Astra, and NVIDIA Grace/Grace‑Hopper systems.
  • Some question how much TOP500 rankings say about per‑chip merit versus overall cluster scale.

Platform openness and firmware standards

  • Concern that an ARM “win” could lead to Android‑style, board‑specific Linux that depends on device trees and vendor kernels.
  • Others counter that ARM can and does use UEFI/ACPI in servers; the fragmentation is about vendor choices, not the ISA.
  • Debate over UEFI/ACPI vs u‑boot + devicetree:
    • Pro‑ACPI/UEFI camp values a single standardized boot and enumeration path.
    • Pro‑devicetree camp argues UEFI/ACPI are opaque blobs that enable persistent firmware rootkits; DT keeps policy and code out of the OS runtime.

Industry structure and future moves

  • Speculation about an AMD–Intel merger or Qualcomm acquiring Intel is met with skepticism due to antitrust and licensing issues.
  • Some think x86 vendors should also back RISC‑V to hedge ARM and keep competition alive.