Announcement

Collapse
No announcement yet.

RADV Vulkan Driver Now Has Working Support For Radeon RX Vega GPUs

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

  • #21
    The issue is, that we don't have an easy toolset for overclocking under Linux. Having a stability stress test running in a window while adjusting voltages and clocks, tweaking the rocky edges with a big stress test? I don't know such a thing under Linux. Save the final profile(s) and have it active on any system configuration you want... thats BIOS flashing.

    Comment


    • #22
      Originally posted by Shevchen View Post
      The issue is, that we don't have an easy toolset for overclocking under Linux. Having a stability stress test running in a window while adjusting voltages and clocks, tweaking the rocky edges with a big stress test? I don't know such a thing under Linux. Save the final profile(s) and have it active on any system configuration you want... thats BIOS flashing.
      Like a graphical control panel type thing? Yeah. we've had those conversations already. This forum has some hairy conversation on that very topic.

      Comment


      • #23
        Originally posted by bridgman View Post

        I hear you. What those scummy AMD people really need to do is start supporting open source kernel drivers, so changes could be made at driver level rather than having to reflash the VBIOS.

        Seriously, why is anyone talking about flashing VBIOS these days other than the fact someone figured out the details for VBIOS clock/voltage manipulation years ago and nobody has apparently done the same with the drivers ?
        I never really understood this. AFAIU the driver parses the powerplay tables in the vbios and then configures the SMC based on the table. I don't see how modifying that would be harder than messing with the vbios, especially since we even have the struct definitions for the table.

        That said, power management is not one of those things I like to play with, since I'm never quite sure what the risk would be of bricking the GPU due to e.g. bad voltage settings.

        Comment


        • #24
          Originally posted by bridgman View Post
          Seriously, why is anyone talking about flashing VBIOS these days other than the fact someone figured out the details for VBIOS clock/voltage manipulation years ago and nobody has apparently done the same with the drivers ?
          Probably because Windows guys are more interested in overclocking and since they don't have a FOSS kernel driver this was their only option.
          ## VGA ##
          AMD: X1950XTX, HD3870, HD5870
          Intel: GMA45, HD3000 (Core i5 2500K)

          Comment


          • #25
            Originally posted by BNieuwenhuizen View Post

            I never really understood this. AFAIU the driver parses the powerplay tables in the vbios and then configures the SMC based on the table. I don't see how modifying that would be harder than messing with the vbios, especially since we even have the struct definitions for the table.
            Because flashing the VBIOS is a one-time only thing. You do it and are done with everything forever. No driver profiles, no re-configuration if you switch kernels, no nothing. Also works between different operating-systems and if the driver is wonky, your VBIOS isn't.

            Originally posted by BNieuwenhuizen View Post
            That said, power management is not one of those things I like to play with, since I'm never quite sure what the risk would be of bricking the GPU due to e.g. bad voltage settings.
            Well, overclocking has always been a personal thing at own risk. Just because one group of people doesn't want to do it is not a reason to not give the option for the group that wants to do it. (at own risk of course)
            Last edited by Shevchen; 07 September 2017, 05:39 PM.

            Comment


            • #26
              Originally posted by bridgman View Post

              I hear you. What those scummy AMD people really need to do is start supporting open source kernel drivers, so changes could be made at driver level rather than having to reflash the VBIOS.

              Seriously, why is anyone talking about flashing VBIOS these days other than the fact someone figured out the details for VBIOS clock/voltage manipulation years ago and nobody has apparently done the same with the drivers ?
              Maybe because VBIOS doesn't accept commands from the driver beyond a certain level? Also since you hear, please tell your friends inside AMD that we want a Ryzen linux version without a security processor and not the opposite (adding security circuits even in our pants). So laser cut a small 2% portion or make a disabling jumper around a cpu supply pin, I need to know that this thing will not start again on it's own. Also you should now that here in EU a small portion of signatures it's enough to start something against you and some others.

              Comment

              Working...
              X