64-bit - is it worth it?

Brian Hawkins brianhks at activeclickweb.com
Mon Apr 21 08:34:11 MDT 2008


I tried going 64 with FC6.  There were a lot of little problems with various
applications that went away when I went back to 32bit.  I was able to run 32
bit apps just find under 64, it just meant that I had to basically have two
linux installs, I had all the 64 bit libraries as well as the 32 bit
libraries.

As for the 4gig of memory thing this is the scoop.  I had 4gig when I was 64
and I still have 4 gig at 32 but this is how it works.  The 32bit kernel
uses 64 bit addressing so it can recognize way more then 4 gig.  User space
however can only see 4 gig.  Under 32 bit the 4th gig's addresses are
reserved for all of your hardware (ie usb port, video, network etc..).

So in order to get more than 3 gig of memory to work with 32 bit you have to
enable a memory hole in the bios.  What it does is places the 4th gig in the
5gig address space.  The effect is that your user space apps still see 3 gig
but your kernel sees all 4 with a gap for the hardware addresses. (if
anything this will probably clear up what that memory hole option is that
you've always seen in your bios settings but never new what it did :))

How do you take advantages of that 4th gig you say?  Well on my box I run
VMWare which runs in the kernel and when it loads it uses up that 4th gig
first.  It is kind of cool to start up a vm and not see a hit on my user
memory.

I'm very happy with 32 bit and 4gig of memory, but I'm still running fc6.
The issues I faced may be better with later software.

Brian



More information about the PLUG mailing list