Can recompiling your kernel lead to speed improvements

[Kurt Wall]

On Wed, Dec 06, 2006 at 10:21:39PM +1100, James McDonald wrote:
> I have an AMD Athlon(tm) Processor CPU and am running Ubuntu  2.6.17-10-386.
> 
> Can I expect performance gains if I recompile the kernel specifically 
> for the AMD processor... I just read a Ubuntu thread that had a bunch of 

Yes, but the question is how much. I wouldn't expect more than a modest
gain (WAG, < 10%) over a stock kernel provided by modern distributions.

> people that say it makes *no* difference. 
> http://ubuntuforums.org/showthread.php?t=24853

Balderdash. The only way to be sure is to test it empirically. The
problem with general statements is that the way in which you use your
system affects its percieved and actual performance. In some cases,
simply using the run-time tunables might be all you need. In other
cases, tuning compile-time options might yield a small improvement.
In all cases, it depends on your workload. If your apps are bound
by graphics operations, a faster kernel won't make as much difference
as optimized graphics drivers.

> My thinking is if you are compiling on you own machine and have the 
> correct optimization shouldn't that give slight improvement?

I concur. Disabling features you don't/won't/can't use makes the overall
kernel smaller.

> And while I'm at it
> <rant>
> I'm gaining frustration with some of the more adminish tasks I could do 
> with ease under FC* and am now finding difficult under Ubuntu...

Before Collins jumps in to defend *buntu, I'll just suggest ditching
what you don't like for something you do.

> I am so tempted to begin the task of moving absolutely everything off 
> this box so I can blow it away and put a mature distribution on it...
> </rant>

[nod]

Kurt
-- 
The older I grow, the more I distrust the familiar doctrine that age
brings wisdom.
		-- H.L. Mencken