Announcement

Collapse
No announcement yet.

Ryzen 3000 Series Gain Workaround For AMD P-State Linux Driver

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

  • Ryzen 3000 Series Gain Workaround For AMD P-State Linux Driver

    Phoronix: Ryzen 3000 Series Gain Workaround For AMD P-State Linux Driver

    For those still running an AMD Ryzen 3000 series "Zen 2" desktop it really ought to be time to upgrade soon for better performance and power efficiency given the Zen 5 performance benchmarks thus far, but for those still planning to use the Ryzen 3000 series for some time, a quirk/workaround is on the way for enabling more of those older platforms to work with the AMD P-State Linux driver...

    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
    Isn't it just for one broken BIOS: B450M MORTAR MAX (MS-7B89)​? Anyway, my Ryzen 3000 works with AMD P-State just fine on MAG B550 TOMAHAWK (MS-7C91).

    Comment


    • #3
      Originally posted by j.fikar View Post
      Isn't it just for one broken BIOS: B450M MORTAR MAX (MS-7B89)​?
      And why should this be fixed in the kernel? Maybe a BIOS fix?

      Comment


      • #4
        This patch is nonsense, because it's a fix for a bug in outdated UEFI... This mobo has a newer UEFI versions available, user can just update it.

        Comment


        • #5
          Originally posted by nadro View Post
          This patch is nonsense, because it's a fix for a bug in outdated UEFI... This mobo has a newer UEFI versions available, user can just update it.
          Well, at least I can't update one of my motherboards so it's not always that simple. Updating to next version would require flashing EC and that tool is only available for Windows which I don't have and will not install. (I don't know whose great idea it has been to run EC flashing tool in complex multithreaded OS but well...) UEFI at least has self-updates on that board but it can't flash EC.

          My fault for buying Gigabyte of course...

          Comment


          • #6
            Originally posted by nadro View Post
            This patch is nonsense, because it's a fix for a bug in outdated UEFI... This mobo has a newer UEFI versions available, user can just update it.
            Do you know how many workarounds, patches, and other fixes are in the Linux kernel for broken/out-of-spec hardware, including those that aren't widely used? I can almost guarantee your number is too low in the USB drivers alone, let alone storage, bridge chipsets, processors, etc. One more is just lost in the noise. It's like people forget open source isn't being progressed by volunteers or something... volunteers that fix problems hardware sellers create in the only ways they can. Your own hardware probably only functions because a volunteer contributor submitted a patch for something broken/out-of-spec in your hardware - that's how pervasive the problem is.

            Comment


            • #7
              Originally posted by phoronix View Post
              Phoronix: Ryzen 3000 Series Gain Workaround For AMD P-State Linux Driver

              For those still running an AMD Ryzen 3000 series "Zen 2" desktop it really ought to be time to upgrade soon for better performance and power efficiency given the Zen 5 performance benchmarks thus far

              https://www.phoronix.com/news/AMD-Ry...-P-State-Quirk

              That's rather extreme to advocate the obsolescence of things that were only released in mid-2020, in the case of CPU+IGPU pairings IIRC there wasn't anything non-OEM substantially better released to consumers until AFAICT 2024, and which still work perfectly well being only a couple years old.

              Same DDR4 RAM architecture so let's face it nothing that's memory bound is going to be ANY faster on a DDR4 platform and the performance gains would be limited to the DDR4/DDR5 system throughput gain even for the most modern CPU upgrade. For GPU bound performance it's basically irrelevant to consider desktop performance since anyone who needs more Ryzen 3000 desktop GPU performance would just use a GPU, the same one then could use with Zen 5 processors, vastly overwhelming the IGPU performance in both cases.

              So then you're really just talking about burst performance for non GPU / non DRAM bottlenecked workloads, ok, well that's cutting out the upgrade reason use cases for lots of people so one is left with a basic desktop that might be 50%-70% slower than SOTA levels on compute performance but is still going to be fast enough for 90%+ of people's productivity / general computing needs even with a 6 core Ryzen 3000 or whatever.

              Power savings? Well if you were a data center, ok, sure. But consumer CPUs that one can / would upgrade in the 65-105 W TDP range or whatever the TDP range is probably around the same or higher for the new parts so no huge "peak sustained compute" level savings but there might be some bigger "mostly idle, mixed usage" percentage power savings in average use. But then those would probably be dwarfed by the overall energy cost of ewaste of the old AM4 motherboard, DDR, heatsink/fan, buying the new Zen5 CPU, heatsink / fan, new DDR5, new rather expensive motherboard, manufacturing / transporting / disposing of all that, etc.

              Maybe in another 5-10 years a Ryzen 3600 will be so slow it can't open a modern office document and handle web / email / basic development etc. But I don't think it's time to call it obsolete quite yet just because you CAN spend more and buy something 2x, 3x, 4x, 50x "faster" (but at what use case?) in CPU MIPS. If that was the case then the high end EPYCs would similarly already obsolete all the desktop Zen 5s for the same reason.

              If anything I'd be tempted to look for the optimum time to buy more DDR4 or upgrade to a 3900 from a 3600 or something on clearance sale to preserve the overall performance / future proof an existing system while keeping the motherboard, fan, etc.

              Comment


              • #8
                Maybe this patch will finally enable amd_pstate for my Epyc 7H12 cpu. There have been so many reported here that supposedly will fix support for Rome processors, but so far none have accomplished this yet. Crossing fingers once again . . . .

                Comment


                • #9
                  Originally posted by pong View Post
                  That's rather extreme to advocate the obsolescence of things that were only released in mid-2020, in the case of CPU+IGPU pairings IIRC there wasn't anything non-OEM substantially better released to consumers until AFAICT 2024, and which still work perfectly well being only a couple years old.

                  [..]
                  Power savings? Well if you were a data center, ok, sure. But consumer CPUs that one can / would upgrade in the 65-105 W TDP range or whatever the TDP range is probably around the same or higher for the new parts so no huge "peak sustained compute" level savings but there might be some bigger "mostly idle, mixed usage" percentage power savings in average use. But then those would probably be dwarfed by the overall energy cost of ewaste of the old AM4 motherboard, DDR, heatsink/fan, buying the new Zen5 CPU, heatsink / fan, new DDR5, new rather expensive motherboard, manufacturing / transporting / disposing of all that, etc.

                  Maybe in another 5-10 years a Ryzen 3600 will be so slow it can't open a modern office document and handle web / email / basic development etc. But I don't think it's time to call it obsolete quite yet just because you CAN spend more and buy something 2x, 3x, 4x, 50x "faster" (but at what use case?) in CPU MIPS. If that was the case then the high end EPYCs would similarly already obsolete all the desktop Zen 5s for the same reason.
                  Generally in agreement with this statement. SOME OF US have to live with this little thing called a BUDGET, especially in a recession full of nothing but ghost jobs while the real ones have been replaced with AI bullshit. The fact is that an AM5 system that cost me literally what this AM4 system cost at the time, despite being purchased DURING THE FUCKING PANDEMIC SHORTAGES, would be a 10% performance boost. At most. And the carbon footprint of production and shipping alone would easily dwarf that saved by the more efficient system over the lifespan it'll will have between now and when I replace it. Likewise the energy cost wouldn't "save" me much either over that timespan.

                  No, I'll stick with my "obsolete" Ryzen 3000 system a little longer, thanks. When I upgrade, it'll be a worthwhile jump in performance. Right now what I need is more storage and better backups. This CPU spends far too much of its time sitting idle right now as it is anyway.

                  Comment


                  • #10
                    Originally posted by pong View Post


                    That's rather extreme to advocate the obsolescence of things that were only released in mid-2020, in the case of CPU+IGPU pairings IIRC there wasn't anything non-OEM substantially better released to consumers until AFAICT 2024, and which still work perfectly well being only a couple years old.

                    Same DDR4 RAM architecture so let's face it nothing that's memory bound is going to be ANY faster on a DDR4 platform and the performance gains would be limited to the DDR4/DDR5 system throughput gain even for the most modern CPU upgrade. For GPU bound performance it's basically irrelevant to consider desktop performance since anyone who needs more Ryzen 3000 desktop GPU performance would just use a GPU, the same one then could use with Zen 5 processors, vastly overwhelming the IGPU performance in both cases.

                    So then you're really just talking about burst performance for non GPU / non DRAM bottlenecked workloads, ok, well that's cutting out the upgrade reason use cases for lots of people so one is left with a basic desktop that might be 50%-70% slower than SOTA levels on compute performance but is still going to be fast enough for 90%+ of people's productivity / general computing needs even with a 6 core Ryzen 3000 or whatever.

                    Power savings? Well if you were a data center, ok, sure. But consumer CPUs that one can / would upgrade in the 65-105 W TDP range or whatever the TDP range is probably around the same or higher for the new parts so no huge "peak sustained compute" level savings but there might be some bigger "mostly idle, mixed usage" percentage power savings in average use. But then those would probably be dwarfed by the overall energy cost of ewaste of the old AM4 motherboard, DDR, heatsink/fan, buying the new Zen5 CPU, heatsink / fan, new DDR5, new rather expensive motherboard, manufacturing / transporting / disposing of all that, etc.

                    Maybe in another 5-10 years a Ryzen 3600 will be so slow it can't open a modern office document and handle web / email / basic development etc. But I don't think it's time to call it obsolete quite yet just because you CAN spend more and buy something 2x, 3x, 4x, 50x "faster" (but at what use case?) in CPU MIPS. If that was the case then the high end EPYCs would similarly already obsolete all the desktop Zen 5s for the same reason.

                    If anything I'd be tempted to look for the optimum time to buy more DDR4 or upgrade to a 3900 from a 3600 or something on clearance sale to preserve the overall performance / future proof an existing system while keeping the motherboard, fan, etc.
                    We've got the energy cost covered in 2030 Status | Mother Nature | Apple so it seems we can keep buying new hw model every year.

                    Comment

                    Working...
                    X