Announcement

Collapse
No announcement yet.

Intel's Early Linux 5.10 Graphics Driver Changes Include Tiger Lake HOBL

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

  • Intel's Early Linux 5.10 Graphics Driver Changes Include Tiger Lake HOBL

    Phoronix: Intel's Early Linux 5.10 Graphics Driver Changes Include Tiger Lake HOBL

    While the Linux 5.9 kernel cycle is still young and not seeing its formal release until early October or so, Intel's open-source team has already submitted to DRM-Next their first batch of feature changes desired for Linux 5.10...

    http://www.phoronix.com/scan.php?pag...nux-5.10-First

  • #2
    HTML typo:

    Originally posted by phoronix View Post
    More details on the initial Intel i915 DRM driver changes for Linux 5.10 via the PR
    1. Error: Stray end tag a..

      From line 335, column 173; to line 335, column 177

      ml">the PR</a.></div>
    2. Error: End tag div seen, but there were open elements.

      From line 335, column 178; to line 335, column 183

      he PR</a.></div>
    3. Error: Unclosed element a.

      From line 335, column 84; to line 335, column 166

      5.10 via <a href="https://lists.freedesktop.org/archives/dri-

    Comment


    • #3
      Umm. Patented? Hardly anything new. Embedded stuff has been doing this for ages.
      There are a lot of problems with device specific optimization tables.
      1. It's a hassle.
      2. It creates divergence.
      3. A lot less margin.

      In all these applications I have found badly tuned individuals, resulting in crashes and whatnot.
      In the end, I often resort to disabling all form of individual & factory-tuned power scaling tables.

      This is partly how companies would sell I-temp device series.
      Set lower scaling with temp, set lower voltage point, claim higher temp tolerance, claim longer service life.
      $$$. Same device, different tune.
      Last edited by milkylainen; 08-27-2020, 07:21 AM.

      Comment


      • #4
        Originally posted by milkylainen View Post
        This is partly how companies would sell I-temp device series.
        Set lower scaling with temp, set lower voltage point, claim higher temp tolerance, claim longer service life.
        $$$. Same device, different tune.
        Is that actually true that it has higher temp tolerance and service life or is it a lie? Because as long as that is true it's mostly fine

        Comment


        • #5
          I doubt the patent has anything to do with this optimized vswing table. "HOBL" likely was just the name of some internal umbrella project for improving battery life, with the optimized vswing table being just one small part of it.

          Comment


          • #6
            Originally posted by starshipeleven View Post
            Is that actually true that it has higher temp tolerance and service life or is it a lie? Because as long as that is true it's mostly fine
            It is true that high temperature affects electromigration and expected failure rate.
            High temperature shortens life. No news there.

            What you do when you buy an I-temp product is that you buy the same silicon,
            but binned and with different scaling tables (complex ASIC, GPU, CPU etc).
            At least standard, entry-level I-temp.
            So the calculated service life is higher, because they don't push the silicon as hard.

            F.ex. in some Nvidia (can only speak for a few models I've actually read the datasheet) Tegras,
            when temp gets high, the scaling table throttles the chip harder.

            Here is the ugly part (for complex ASICs)
            So you're _not_ buying a product that can take high temp at full throttle (Itemp < Tjunction)
            You're buying a product that won't break at high temp and is still active at high temp.
            They are usually slower at high temp than their non I-temp counterparts.

            Comment

            Working...
            X