-20
submitted 1 day ago* (last edited 1 day ago) by madthumbs@lemmy.world to c/linuxsucks@lemmy.world

BSD has better security, load handling, network stack, documentation, cohesion, power efficiency, etc. -Try to explain why BSD isn't dominant on servers without mentioning the GNU/GPL cult.

Wasting 30-45% power isn't just about about laptops wasting power anymore, it's about datacenter‑scale inefficiency caused by architectural gaps in Linux’s power management stack.

CPU idle state handling is worse than Windows Server Linux often fails to enter the deepest C‑states under load patterns that Windows handles correctly.

  • Windows Server uses vendor‑tuned ACPI tables, DPTF, and firmware hooks
  • Linux relies on generic ACPI, often missing vendor‑specific power logic
  • Result: CPUs sit in C2/C3 instead of C6/C7, wasting 20–40% power

This is especially bad on:

  • Intel Xeon Scalable (Ice Lake, Sapphire Rapids)
  • AMD EPYC (Naples → Milan → Genoa)

Linux scheduler inefficiency The Completely Fair Scheduler (CFS) is great for fairness, terrible for power.

  • It spreads tasks across cores making more cores wake and using more power
  • Windows Server aggressively core‑packs workloads with fewer active cores and lower power draw

This alone can cause 15–25% power waste under mixed workloads.

PCIe ASPM is frequently disabled On servers, Linux often ships with:

  • ASPM off
  • L1 substates off
  • PCIe power gating off

Why?
Because vendors disable these features in firmware when Linux drivers don’t reliably support them.

That’s 5–10% wasted power across NICs, NVMe, HBAs, GPUs, accelerators.

Driver power gating is incomplete Windows drivers are vendor‑authored.
Linux drivers are:

  • reverse‑engineered
  • community‑maintained
  • missing proprietary power hooks

This affects:

  • Mellanox/ConnectX NICs
  • Intel i40e/ixgbe
  • AMDGPU/ROCm
  • NVIDIA (even with proprietary drivers)
  • NVMe controllers

Power gating gaps can add 10–20% inefficiency.

Kernel regressions are common and persistent A single kernel update can break:

  • idle states
  • turbo scaling
  • NUMA balancing
  • scheduler heuristics
  • cpufreq governors

These regressions often sit unfixed for months, especially in LTS kernels.

top 1 comments
sorted by: hot top new old
[-] PressureBig3940@lemmy.world 2 points 1 day ago

Sharing an anecdote: I lost a 1U server because Linux didn't spin up the fans while the CPU was under 100% load on all cores.

This never happened in Windows.

this post was submitted on 30 Aug 2026
-20 points (4.5% liked)

Linuxsucks

405 readers
124 users here now

For knowledge and awareness about what using Linux is really like and pointing at its cultish toxic community. We also cover FOSS /FLOSS failures, and issues with GPL since it relates to Linux. Moderation is heavy handed to appeal to our target users.

Rules:

  1. FOSS (especially GPL) advocates and Linux (specifically GNU/Linux) evangelists aren't welcome (GNU Hurd will count as Linux). -We ask that you block us and we will perma-ban for violations of this rule.
  2. Try to stay on topic (that is LINUXSUCKS!). No bashing other OSs or each other. We aim for a cohesive non-toxic community with the interest of cautioning people / would be victims from Linux. At our discretion, we will remove what we consider pollution.
  3. Moderators run the sub, not the users. It's not democratic, and we don't care how you think it should be run. -Feel free to make and run your own community.
  4. "Not your Billboard" -Over-represented companies will have positive posts removed and appropriate bans may be issued. -Those companies included so far will be Valve/Steam/Proton, Google, and Brave (browser).

⚠️Collection of Articles

⚠️Linux/FOSS can damage hardware or firmware

🪤Linux Community Advice Breaks Windows

🫃Mixing Apps Bloats LiGNUx

😎Linux running servers isn't a brag

👾Is Linux Running Games Near Windows Performance Impressive?

🚮Wasted Ram on Different Toolkits and Distro-Agnostic Packages in Linux

🛰️Critical ISS Systems do NOT run Linux

⚠️Abandoned Software is Dangerous (and common on Linux)

🧻FOSS Devs Quit and Sellout on Unappreciative Userbase

⚠️Firmware Flashing is Riskier on Linux

💊Linux Community Toxicity Ties Directly into Inferiority Complex Psychology

👍The Positives of Telemetry

👾Failures of Proton

⚠️How Linux Stores Browser Passwords is a Security Issue

💊The Psychology of Conspiracy Thinking

🤬Rabid LiGNUxers Ignorantly Get Angry at Devs

😳Does Linux Dominate SuperComputers? -Yeah, but So What!

😎Linux Myths Compilation

⛪The Linux Cult - Religious Parallels

😎Before Wayland: “Linux is secure, Windows is insecure.” -(Selective Dishonesty)

⏳Why Wayland is Taking so Long

😎Major Desktop Applications Missing on Linux

⚠️Sudoedit vs Edit as Root

😎Linux is NOT Great for Developers

😎Kernel Level Anti-Cheat is fine!

⚠️The Privacy / Security Paradox

⚠️Linux Storing Passwords as Plain Text

😎How to Continue Using Unsupported Windows Online on Old Hardware

😎Don't Trust Desktop Marketshare

💊Piracy is Less Harmful than GPL

😎The Most Influential Formerly‑Proprietary Projects that Became Important to Linux

🤬GNU Holds Linux Back (Directly)

💊It is Linux Fault -Why OEMs, Apps Don't Support Linux

🔐 Secure Boot + TPM 2 vs. Linux Alternatives

🛠️Checking for Linux Hardware Compatibility is Bullshit!

🤪Distro-Hopping Madness!

⚠️The Myth of Linux Security

👾Linux Gaming: The Roast It Has Earned

💊 Linux is Better for Old Computers?

💊Linux Sucks at its Core

💊Lies told about BSD to retain support of cult

💊FOSS Apps Run Better on Win/Mac

🤪Isolation is Dangerous!

💊Linux Sucks -Even the Kernel

😎Haiku for Older Hardware

💊Linux Wastes Power - Even on Servers!

founded 2 years ago
MODERATORS