Announcement

Collapse
No announcement yet.

Intel Speed Select Driver Issue Was Hurting Performance In Some HPC Benchmarks

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Intel Speed Select Driver Issue Was Hurting Performance In Some HPC Benchmarks

    Phoronix: Intel Speed Select Driver Issue Was Hurting Performance In Some HPC Benchmarks

    Intel's Speed Select Technology introduced since Cascade Lake for providing more granular power/performance controls was done in the name of performance but it turns out an ISST Linux driver inefficiency could lead to a 10%+ performance hit for some HPC benchmarks...

    https://www.phoronix.com/scan.php?pa...elect-Hurt-HPC

  • #2
    It also is just another recent example of the increasing complexity of CPU power management features/controls having the potential for significant unintended consequences.
    It's hard to make changes in complex systems that don't have unintended consequences. That's often why senior devs don't like junior or beginning devs (or worse, management dictating a change) to start changing things willy-nilly because the style or language de jour dictates some other way of doing things. There may have been a damned good reason certain code reads the way it does.

    Comment


    • #3
      Originally posted by stormcrow View Post

      It's hard to make changes in complex systems that don't have unintended consequences. That's often why senior devs don't like junior or beginning devs (or worse, management dictating a change) to start changing things willy-nilly because the style or language de jour dictates some other way of doing things. There may have been a damned good reason certain code reads the way it does.
      Look, my setup works for me. Just add an option to reenable spacebar heating.

      Comment


      • #4
        FFS, stop trying to be clever. The performance advantages when everything is working correctly are extremely minimal, and the potential for regressions and bugs are greatly increased. This isn't theoretical, this has been a recurring theme. Speedshift, Speedstep, Speedselect. Every single time, stuff breaks, and the solution has been to go back to ACPI, Which works, works extremely well, and so long as the device vendor is halfway competent, almost never introduces regressions to any meaningful level (worst I've seen on ~10 machines was 3% performance delta)
        Scale back your "features" for ME, stop with the broken "Speed*" and "security" extensions that inevitably break, and focus on things that actually contribute to completing a task faster so the system can get back to idle quicker, preferably by doing something other than disregarding security.

        Comment

        Working...
        X