[GRLUG] Pentium 4 and Gentoo Linux?

[Kyle]

On 06/20/2013 10:20 AM, Adam Tauno Williams wrote:
>>> you compile out all kinds of junk you don't want or use by changing 
>>> your USE flags, and that, for me anyway, makes a significant 
>>> difference in speed and power consumption vs. a generic "cram every
>>> feature in, just in case it's ever needed" build.
> Honestly I find even that hard to believe;  unused pages can be swapped
> out.  And once shared libraries are memory-mapped additional references
> are essentially free.  So unless the architecture has a totally crap I/O
> subsystem (so load times are horrible) or a very decrepit MMU - a
> perceptible change is hard to believe.

In theory, you're right.  It shouldn't make a huge difference.

In practice, my laptop has pretty much the same software loaded in both
Gentoo and Ubuntu (it triple-boots those two + Windows 7) and I spent
quite a bit of time optimizing Ubuntu before switching primarily to
Gentoo.  So I have a decent setup for comparing stripped-down vs.
pre-compiled binaries on the same hardware.

Gentoo loads up on a cold boot, complete with preloading libraries and
files into RAM in about 15 seconds.  From a bootchart run, it's largely
I/O bound for the first 75% or so and CPU-bound for the last 25% as X is
starting.  Last time I timed Ubuntu, IIRC it took >60 seconds to boot. 
Don't remember the exact count, but the difference is remarkable.

After booting is done, while Gentoo already has Chromium preloaded into
RAM (one of my most-used programs) during the boot process, Ubuntu takes
probably a good 10-15 additional seconds to open it on first load.

Both of these items are related to the first loading of files, and as
you mentioned, they should be cached in RAM so it doesn't have to do
that again, so your time lost to these is minimal if you keep your
system up for long periods of time.  But this is a laptop - boot speed
does actually make a huge difference for me in my perception of the
responsiveness of the system.  I've never particularly liked standby
mode as an alternative to a full shut down and boot up, so you could
contend that boot speed is completely irrelevant for most people, but it
fits what I'm comfortable with better.

It doesn't stop at boot time, though.  Once it's finished booting,
Ubuntu idles at (and I'm going from memory from a few days ago on this
since I don't want to reboot to check right now) 2-3% CPU on all eight
pseudo-cores average (HyperThreaded quad-core i7).  Gentoo idles at
between 0.25-1% CPU.  Having half the CPU usage at idle makes a
difference for me in my perception of how well the computer runs,
because it allows the CPU to stay in C3 almost all of the time, leaving
me with improved battery life and a cooler-running system due to less
power consumption.  My fan very rarely pops out of its lowest speed on
Gentoo, leaving me with a very quiet system, whereas it goes to its
second speed level quite a bit more often on Ubuntu.  Granted, it's
still not the noise level I get from Windows 7 keeping the thing in a
pretty constant level 3-4 at idle, but it still makes a noticeable
difference in the sound level of my machine.

So, is it due to my perceptions of it?  Sure.  Some of it, I've got
numbers to back up.  Some of it, I don't.  It's a preference.

- Kyle