kernel-packages team mailing list archive

["Christopher M. Penalver" <christopher.m.penalver@xxxxxxxxx>]

Sam Davies, thank you for reporting this and helping make Ubuntu better. In Trusty, the timer for amd64 is 250Hz. I'm aware that the Ubuntu Kernel Team strongly prefers remaining as synched with upstream on default configurations as possible. Despite this, if this new default wouldn't work out for you, would you please execute the following in a terminal:
apport-collect 474977

As well, would you be able to provide some benchmarks for your hardware
that demonstrate the improvements, and discussion on any negatives from
your hardware?

Thank you for your understanding.

** Changed in: linux (Ubuntu)
   Importance: Undecided => Wishlist

** Changed in: linux (Ubuntu)
       Status: Confirmed => Incomplete

-- 
You received this bug notification because you are a member of Kernel
Packages, which is subscribed to linux in Ubuntu.
https://bugs.launchpad.net/bugs/474977

Title:
  Stock ubuntu desktop kernel should be using 1000hz timer resolution

Status in “linux” package in Ubuntu:
  Incomplete

Bug description:
  The standard ubuntu kernel uses 100hz timer resolution (for amd64,
  250hz for x86). I don't know why this is, I'm assuming it's just used
  because it's the kernel default or something?

  Either way it's not ideal for typical desktop usage, 100hz especially
  really does introduce noticeable lag, can cause skips in audio and
  just generally makes the whole desktop tend to microstutter a bit.
  According to Con Kolivas (the guy who used to maintain the -ck patches
  and who wrote the original completely fair [staircase] scheduler for
  linux) the kernel devs ran some tests and for most desktop use cases,
  using 1000hz over 100hz timers caused pretty much no performance
  impact, whilst in some corner cases it caused up to a 1% performance
  drop. On the other hand, the desktop interactivity gains were very
  noticeable by the users.

  DISCLAIMER: Of course it's always hard to justify these sorts of
  claims because desktop responsiveness is a notoriously woolly and hard
  to measure thing, but without at least suggesting some solutions I
  don't see how we can improve it, hence this bug report.

  Personally I think that a 1% performance drop is totally acceptable on
  a desktop in order to obtain a much smoother running system. In my
  opinion (one which I'd sincerely hope most people here share) we
  should be aiming at making Ubuntu Desktop the best desktop O/S you can
  have, which means providing the best user experience. With lower
  resolution timers, you can occasionally have audio skipping, your
  mouse inputs aren't always responded to quite as fast, your compiz is
  a bit more jerky, your games can sometimes lag for fractions of a
  second etc etc etc. Just general responsiveness suffers in the case of
  low res timers.

  Now I know these sort of claims are often made and because it's
  desktop interactivity, it's very hard to prove anything because it's
  subjective, but I encourage anybody reading this to compile a kernel
  with both 100hz timers and 1000hz timers and have a play around, just
  use the desktop like you normally would, typing text, moving windows,
  playing music etc. I'm sure you will notice quite a difference. I
  would politely request that those in charge of setting these kernel
  configuration options try out the different timer resolutions and have
  a play around and see if it actually does make an improvement or not.

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/474977/+subscriptions