Announcement

Collapse
No announcement yet.

A New Open-Source Tool To Change Intel GPU Frequencies

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

  • A New Open-Source Tool To Change Intel GPU Frequencies

    Phoronix: A New Open-Source Tool To Change Intel GPU Frequencies

    Ben Widawsky of Intel's Open-Source Technology Center rolled out a new experimental tool aptly called intel_freequency for manipulating the Intel GPU frequency under Linux...

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

  • #2
    intel_freeeeeeeeeeeeeequency

    > intel_freequency

    One too many e's there...

    Comment


    • #3
      Similar tool

      Hi,

      I haven't looked at the tool mentioned in the article but from the summary it appears to be similar to what I wrote a few months ago in order to keep my X121e from shutting down under heavy GPU load.
      Tools like thermald only seem to throttle the CPU which did not help at all in this context.

      The source code can be found here:
      GPU power management for Intel hardware on Linux. Contribute to jmechnich/intel-power-control development by creating an account on GitHub.


      Maybe this can be useful to somebody.

      Cheers.

      Comment


      • #4
        Originally posted by Ben Widawsky
        For many kernel releases now sysfs has supported reading and writing the GPU
        frequency. Therefore, this tool provides no new functionality. What it does
        provide is an easy to package (for distros) tool that handles the most common
        scenarios.
        This just means that distributions now have one more package to take care of, no?

        Still, it does give a nicer output than using sysfs. Guess it can be quite useful after all.

        Comment


        • #5
          Originally posted by Sonadow View Post
          This just means that distributions now have one more package to take care of, no?

          Still, it does give a nicer output than using sysfs. Guess it can be quite useful after all.
          This is just a new addition to the existing intel-gpu-tools package which already has a billion programs.

          Comment


          • #6
            Originally posted by Sonadow View Post
            This just means that distributions now have one more package to take care of, no?
            It will a part of intel-gpu-tools, but it hasn't been merged yet (see git repo, intel_frequency.c is not present).

            Good new, maybe someone will write a GUI, like Radeon-Profile.

            Comment


            • #7
              I just manually change my intel graphics speed via a shell script. I set up the script so it alters the speed based on whether the AC adapter is plugged in. I have it run every time I boot up the computer. I could do it during cron but I don't unplug my laptop that often while it is already running.


              For my laptop, I actually HAVE to slow down my GPU by about 300MHz because if I don't and I put the GPU under full load, the whole computer reboots. Pretty annoying but whatever, I don't use the computer for gaming and even when underclocked the GPU is plenty powerful.

              Comment


              • #8
                Originally posted by jmech View Post
                Hi,

                I haven't looked at the tool mentioned in the article but from the summary it appears to be similar to what I wrote a few months ago in order to keep my X121e from shutting down under heavy GPU load.
                Intel's magic uncore doesn't handle the GPU? That seems like quite an oversight and something they should fix. The GPU absolutely adds to the power draw and heat output, and affects how far turbo can go on the CPU side.

                Always assuming that someone didn't override the CPU's default thermal/power/speed envelope. It would not be the first time I've seen someone have weird problems after they forced overclock and disabled thermal throttle. "Because it GOES FASTER" they say, as their computer burns behind them.

                Comment


                • #9
                  Excellent I was looking for something like this a few weeks ago. I took my laptop (i7 Dell ultrabook) with me on Holidays over xmas and kept rebooting due to overheating issues (I blame the hot Australian Summer). I tried everything including flashing bios couldn't figure out how to force it to run in a low power state. In the end I went to an office supply store and bought a $20 cooling pad to sit it on top of out of desperation and it stopped all the reboots.

                  Comment


                  • #10
                    Originally posted by Zan Lynx View Post
                    Intel's magic uncore doesn't handle the GPU? That seems like quite an oversight and something they should fix. The GPU absolutely adds to the power draw and heat output, and affects how far turbo can go on the CPU side.

                    Always assuming that someone didn't override the CPU's default thermal/power/speed envelope. It would not be the first time I've seen someone have weird problems after they forced overclock and disabled thermal throttle. "Because it GOES FASTER" they say, as their computer burns behind them.
                    If I had wanted a fast machine, I wouldn't have gone for the X121e in the first place...
                    No, but seriously: the problem occurred on a more or less stock Debian system (CrunchBang) without using any additional tools for fan throttling or similar. However, I also partly blame a particular game (Shadowrun Returns) for it as it seemed to produce disproportionately high GPU loads, even when there wasn't that much happening on-screen (it's a turn-based strategy game). Still this shouldn't have lead to the system's overheating.

                    I completely agree with you on that I would have expected as well that some automatic GPU throttling would happen before the temperature reaches 100 degrees Celsius and the machine goes into emergency shutdown. The tool I mentioned in my previous post completely solved the issue for me, it will just sit in the systray, watch the temperature of e.g. the x86 package and reduce the GPU clock if the temperature rises above some limit (and vice versa if the temperature goes down).

                    Comment

                    Working...
                    X