Turbostat Utility Lands New Features With Linux 6.13

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts
  • phoronix
    Administrator
    • Jan 2007
    • 67157

    Turbostat Utility Lands New Features With Linux 6.13

    Phoronix: Turbostat Utility Lands New Features With Linux 6.13

    The turbostat utility that lives within the Linux kernel source tree for reporting CPU frequency/idle statistics and other metrics is gaining some new capabilities as part of the Linux 6.13 cycle...

    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
  • coder
    Senior Member
    • Nov 2014
    • 8859

    #2
    It really bugs me that ATX 3.0 didn't add a standard I2C interface for querying stats from the PSU. They didn't have to mandate that all PSUs implement it, but at least having a standard interface would save us from having to rely on entirely proprietary mechanisms, like Corsair's USB-based iCue interface.

    Comment

    • caligula
      Senior Member
      • Jan 2014
      • 3314

      #3
      Originally posted by coder View Post
      It really bugs me that ATX 3.0 didn't add a standard I2C interface for querying stats from the PSU. They didn't have to mandate that all PSUs implement it, but at least having a standard interface would save us from having to rely on entirely proprietary mechanisms, like Corsair's USB-based iCue interface.
      What's wrong with iCUE? It only requires 1,2 GB of space. https://forum.corsair.com/forums/top...ue-take-12-gb/

      Comment

      • coder
        Senior Member
        • Nov 2014
        • 8859

        #4
        Originally posted by caligula View Post
        What's wrong with iCUE? It only requires 1,2 GB of space. https://forum.corsair.com/forums/top...ue-take-12-gb/
        The issue I'm concerned about is the lack of a single standard that all software can support. If we rely on vendor-specific extensions, then it becomes incumbent on every software program to implement support for every vendor-specific interface. Obviously, that's not going to happen. Plus, the more interfaces need to be supported by more programs, the more opportunities there are for bugs to creep in.

        It's surprising to have to explain why standards are a good thing, in this day and age.

        Comment

        • yump
          Senior Member
          • Aug 2021
          • 505

          #5
          Originally posted by coder View Post
          It really bugs me that ATX 3.0 didn't add a standard I2C interface for querying stats from the PSU. They didn't have to mandate that all PSUs implement it, but at least having a standard interface would save us from having to rely on entirely proprietary mechanisms, like Corsair's USB-based iCue interface.
          ATX12VO has an analog signal for % of PSU rated output, and that's what we all *should* be using. That the DIY market has continued to cling to the obsolete and inefficient multi-rail standard is... a choice.

          Comment

          • caligula
            Senior Member
            • Jan 2014
            • 3314

            #6
            Originally posted by coder View Post
            The issue I'm concerned about is the lack of a single standard that all software can support. If we rely on vendor-specific extensions, then it becomes incumbent on every software program to implement support for every vendor-specific interface. Obviously, that's not going to happen. Plus, the more interfaces need to be supported by more programs, the more opportunities there are for bugs to creep in.

            It's surprising to have to explain why standards are a good thing, in this day and age.
            Sure, that was sarcasm. The iCUE software is worst piece of shit I can imagine. Imagine installing 6+ of those for each device from different vendors. 10 GB of space used for power/temp sensors. I also remember some have told they disable iCUE because it drops the frame rates too much in games.

            Comment

            • the-burrito-triangle
              Phoronix Member
              • Jul 2024
              • 79

              #7
              Originally posted by yump View Post

              ATX12VO has an analog signal for % of PSU rated output, and that's what we all *should* be using. That the DIY market has continued to cling to the obsolete and inefficient multi-rail standard is... a choice.
              An analog signal is NOT the best way to report load usage. EMI, and a myriad of other common issues can add so much noise that it becomes meaningless. A digital interface that can be queried is much better than passing an analog signal to the mobo where an ADC would need to be present to read it. The PSU designers can find the best place to put the multiplexed ADC and ancillary circuits for lowest noise and measurement of ALL the voltages and possibly the AC input current and power factor.

              And your "single rail" isn't true--ever. What is true is that modern PSUs use a single AC-to-DC SMPS for the primary 12V output and then _multiple_ DC-to-DC SMPS for the remaining supply voltages (3.3V, 5V, etc.). No matter how you slice the pie, there are multiple "power rails". It's just an argument of topology. Turns out a single AC-DC SMPS for the primary voltage and secondary DC-DC SMPS for the others is more efficient overall.

              Comment

              • yump
                Senior Member
                • Aug 2021
                • 505

                #8
                Originally posted by the-burrito-triangle View Post

                An analog signal is NOT the best way to report load usage. EMI, and a myriad of other common issues can add so much noise that it becomes meaningless. A digital interface that can be queried is much better than passing an analog signal to the mobo where an ADC would need to be present to read it. The PSU designers can find the best place to put the multiplexed ADC and ancillary circuits for lowest noise and measurement of ALL the voltages and possibly the AC input current and power factor.
                Digital would be better, but I assume the reason they went with analog is that it's cheap to buffer and pin-out a signal that PSU platforms already have for over-power protection circuits anyway. Most PSUs get away without having microprocessors at all. Also I_PSU% is a current signal, so noise immunity should pretty good.

                I think the ADC is on the CPU, but don't quote me on that. That'd be the most obvious place, given the intended purpose of allowing the CPU to throttle to avoid overpower trips.

                Power factor is very close to 1 at all reasonable loads for anything made since like, 2010, because it's legally required approximately everywhere.

                And your "single rail" isn't true--ever. What is true is that modern PSUs use a single AC-to-DC SMPS for the primary 12V output and then _multiple_ DC-to-DC SMPS for the remaining supply voltages (3.3V, 5V, etc.). No matter how you slice the pie, there are multiple "power rails". It's just an argument of topology. Turns out a single AC-DC SMPS for the primary voltage and secondary DC-DC SMPS for the others is more efficient overall.
                Outside of DIYPC, other DC-DC converters are not part of the PSU, and are right-sized for their loads (which they are right next to for less I^2R loss), instead of ludicrously overspecced just-in-case someone uses the PSU on a really old machine that heavily loads the 5V/3V3 rails. A modern PSU has two supplies: 12V, and 12V standby.

                Check out any integrated graphics OEM desktop from the last 10 years. They all have single-voltage PSUs, and most idle under 10 W.

                Comment

                Working...
                X