Announcement

Collapse
No announcement yet.

AMD Ryzen, EPYC 5~6% Faster Out-Of-The-Box With Linux 5.11

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

  • intelfx
    replied
    Originally posted by remenic View Post
    I maintain a bunch of EPYC VPSs running Ubuntu 20.04 on kernel 5.4.0-65-generic. Would these benefit from kernel 5.11 too, considering they're virtualized?
    If you employ virtualization, you should not use schedutil at all, because processes inside virtual machines are not visible to the host scheduler. Use performance or aggressively tuned ondemand.

    Leave a comment:


  • remenic
    replied
    I maintain a bunch of EPYC VPSs running Ubuntu 20.04 on kernel 5.4.0-65-generic. Would these benefit from kernel 5.11 too, considering they're virtualized?

    Leave a comment:


  • Michael
    replied
    Originally posted by jospoortvliet View Post
    A cool thing to note about this is that, given Epic also benefits, these improvements are not limited to ZEN 3 but clearly work across the board of at least Zen 2 and Zen 3. Who knows, maybe Zen 1(+) benefit too ;-)

    (I hope so, sporting a 2700X here)
    Aside from any general improvements like I saw in some tests like PostgreSQL, the reliance on ACPI CPPC for frequency invariance makes it only Zen 2 and later.

    Leave a comment:


  • jospoortvliet
    replied
    A cool thing to note about this is that, given Epic also benefits, these improvements are not limited to ZEN 3 but clearly work across the board of at least Zen 2 and Zen 3. Who knows, maybe Zen 1(+) benefit too ;-)

    (I hope so, sporting a 2700X here)

    Leave a comment:


  • mppix
    replied
    Originally posted by Linuxxx View Post
    If you are a workstation user, then why settle on Debian?
    Why not use a purpose-built version of Linux like Ubuntu Studio?
    Oh, I was referring how you can use Debian... Personally, I prefer that but there are many valid choices including Ubuntu.

    Leave a comment:


  • Setif
    replied
    Originally posted by mppix View Post

    Well Debian stable always uses the stable kernel, which is 5.10 for the bullseye release. They would still use it even if 5.11 would be released in time. However, you can always install newer kernels using buster/bullseye-backports once they are released.
    I suspect that some of the 5.11 improvements did not make it into 5.10 because it is the stable release and some of the commits were more on the experimental side, see e.g. the shedutil regression + fix.
    I think Debian is in pretty good shape for bullseye and we get the usual use cases
    -> stable/bullseye: server
    -> stable/bullseye+backports: workstation
    -> testing: laptop/consumer
    You mean LTS.

    Leave a comment:


  • Linuxxx
    replied
    Originally posted by extremesquared View Post
    All these improvements *just* missing Debian 11 cutoffs bode well for the stability of Debian 12.
    I'm not being facetious here. For the patient workstation builder, 2023 will see a (likely) end to shortages, a decent hardware jump in power/efficiency, and an entire release cycle testing kernel code for said hardware. Us debian users get more excited the more boring a release is.
    If you are a workstation user, then why settle on Debian?
    Why not use a purpose-built version of Linux like Ubuntu Studio?

    Leave a comment:


  • mppix
    replied
    Originally posted by torsionbar28 View Post
    AMD CPU's get faster over time (from code enhancements and optimizations), while intel CPU's get slower (Meltdown, Spectre, ITlb multihit, L1tf, Srbds, Tsx async abort, etc). Now that's a value proposition!
    Well Debian stable always uses the stable kernel, which is 5.10 for the bullseye release. They would still use it even if 5.11 would be released in time. However, you can always install newer kernels using buster/bullseye-backports once they are released.
    I suspect that some of the 5.11 improvements did not make it into 5.10 because it is the stable release and some of the commits were more on the experimental side, see e.g. the shedutil regression + fix.
    I think Debian is in pretty good shape for bullseye and we get the usual use cases
    -> stable/bullseye: server
    -> stable/bullseye+backports: workstation
    -> testing: laptop/consumer

    Leave a comment:


  • cytomax55
    replied
    Lol, that kind of logic makes my head hurt.... It's either boring and stable or new and exciting... You can't get excited over something boring... Can you?
    aww jesus my head hurts too much thinking

    Leave a comment:


  • extremesquared
    replied
    All these improvements *just* missing Debian 11 cutoffs bode well for the stability of Debian 12.
    I'm not being facetious here. For the patient workstation builder, 2023 will see a (likely) end to shortages, a decent hardware jump in power/efficiency, and an entire release cycle testing kernel code for said hardware. Us debian users get more excited the more boring a release is.

    Leave a comment:

Working...
X