Announcement

Collapse
No announcement yet.

Linux 5.15-rc3 Released - Looking "Pretty Normal" Plus Performance Fix

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

  • Linux 5.15-rc3 Released - Looking "Pretty Normal" Plus Performance Fix

    Phoronix: Linux 5.15-rc3 Released - Looking "Pretty Normal" Plus Performance Fix

    Linus Torvalds has now issued the third weekly release candidate of the forthcoming Linux 5.15 kernel...

    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
    Originally posted by Linus Torvalds
    So after a somewhat rocky merge window and second rc, things are now
    actually looking pretty normal for rc3. Knock wood.

    There are fixes all over, and the statistics look fairly regular, with
    drivers dominating as they should (since they are most of the tree).
    And outside of drivers, we have a fairly usual mix of changes -
    architecture fixes, networking, filesystems, and tooling (the latter
    being mostly kvm selftests).

    Shortlog appended, it's not too long and easy to scan through to get a
    flavor for the details if you happen to care.

    Please do give it a whirl,

    Linus
    Also- beautiful people, for those compiling the kernels using the script I shared.. the memcg regression patch was breaking the build for -rc3 since it's now merged in. Script has been updated to no longer apply the patch. (edit: Added support for 5.15-rc3-rt5 that was just released in case anyone needs.)

    Been very happy with the 5.15-rc builds. It's snappy, for sure. Gonna run some fresh pts/osbenchs soon comparing 5.4, 5.10, 5.13, 5.14, 5.15-rc. (for my curiosity, and then to share).

    (btw to the unnecessary haters: I'm not spamming the script, I just want to share for those curious and brave enough to want to build their own Linux kernels with their shiny new 8c/16t processors)

    Lastly, another benefit of running a newer kernel (as opposed to say 5.4 LTS) is that my RX480 Polaris graphics card runs at half the power and has ZeroRPM fan support.

    GPU runs at 13W as opposed to 28-30W on 5.4. Probably worth it to run the newer kernel just for that, and save the 5.4 LTS for my MacBook Pro that has no choice *but* to run 5.4.

    Couldn't be happier right now with my Linux desktop. Was not running this nice since I first built it in January 2014 when it was running vanilla Debian (not knocking Debian, just saying we've come a long way). Happy as hell.

    Code:
    $ watch -d -n 1 'rocm-smi'
    
    ========================ROCm System Management Interface========================
    ================================================== ============================
    GPU Temp AvgPwr SCLK MCLK Fan Perf PwrCap VRAM% GPU%
    0 57.0c 13.192W 300Mhz 300Mhz 0.0% manual 180.0W 6% 0%
    ================================================== ============================
    ==============================End of ROCm SMI Log ==============================
    Last edited by perpetually high; 27 September 2021, 08:21 AM.

    Comment


    • #3
      Quick question, off topic, If someone can chime in:

      With AMD's ZeroRPM, GPU runs relatively much hotter (idles at 57 celcius).

      With 5.4 and no ZeroRPM, fan is not audible (I think 700-900 rpm or so) but runs much cooler. In the low 40s, maybe even high 30s

      What are you guys' thoughts on the effect of the longevity on the card because of this? Will the degree difference matter in your opinion? Insignificant and never to be brought up again?

      Have had the card since Feb 2017 (see join date) and even mined a couple ETH (gambled and lost it all in online poker when ETH was $260. #winning) and have been the sole owner. Would like to keep it running forever, if possible.

      Thanks in advance for any replies and for listening to my TED Talk.

      Comment


      • #4
        Cards have been known to run reliably for years at 85ยฐC, issue seems to be above 90ยฐC and approaching 100ยฐC the lifespan is severely affected. 57, whist not ideal, isn't too bad.
        Just make sure that it's not running too hot when under load.

        Comment


        • #5
          Originally posted by perpetually high View Post
          Quick question, off topic, If someone can chime in:

          With AMD's ZeroRPM, GPU runs relatively much hotter (idles at 57 celcius).

          With 5.4 and no ZeroRPM, fan is not audible (I think 700-900 rpm or so) but runs much cooler. In the low 40s, maybe even high 30s

          What are you guys' thoughts on the effect of the longevity on the card because of this? Will the degree difference matter in your opinion? Insignificant and never to be brought up again?

          Have had the card since Feb 2017 (see join date) and even mined a couple ETH (gambled and lost it all in online poker when ETH was $260. #winning) and have been the sole owner. Would like to keep it running forever, if possible.

          Thanks in advance for any replies and for listening to my TED Talk.
          You will never notice a difference. 57C is well within the limit of the chip.

          Technically (and you can always ignore anything after someone starts a sentence with the word technically) you will extend the life but you will be in the realm of statics to see if you can find or notice a difference. Most likely you will have another component die before the GPU does.

          Comment


          • #6
            Originally posted by vsteel View Post

            You will never notice a difference. 57C is well within the limit of the chip.

            Technically (and you can always ignore anything after someone starts a sentence with the word technically) you will extend the life but you will be in the realm of statics to see if you can find or notice a difference. Most likely you will have another component die before the GPU does.
            ๐Ÿ˜‚ I appreciate the response. Exactly what I was thinking also. Cheers

            Comment


            • #7
              Originally posted by perpetually high View Post

              ๐Ÿ˜‚ I appreciate the response. Exactly what I was thinking also. Cheers
              Worth noting that fans have a lifetime as well. So itโ€™s not guaranteed that having zero rpm enabled actually lowers the overall lifetime of your card.

              Comment


              • #8
                Originally posted by Markopolo View Post

                Worth noting that fans have a lifetime as well. So itโ€™s not guaranteed that having zero rpm enabled actually lowers the overall lifetime of your card.
                Totally did not think of that, lol. Great point. Thanks again to all those that replied.

                Comment


                • #9
                  A huge problem Kernel 5.15-rc3 still has: there still are sleep/resume issues with ZEN3 architectures.
                  There is a large Reddit thread outlining this - maybe some people here in the forum can confirm?

                  Comment

                  Working...
                  X