On 08/10/2009 05:14 PM, Jens M Andreasen wrote: 
On Mon, 2009-08-10 at 15:22 +1000, Patrick Shirkey wrote:

  
"PAE is required for NX support, and furthermore enables larger
swapspace support for non-overcommit purposes. It has the cost of more
pagetable lookup overhead, and also consumes more pagetable space per
process."

That suggests to me a similar situation to a win modem or auto
resampling. 

    

Why? The sentences above do not specify what is compared and only
handwavingly suggests a cost of "more"  ... Mind you, 64bit pointers
crawling all over the place ain't free either.

Simply because they don't specify the real world cost of using the method.



  

    
- Should this be a recommended method coming from LAD?

    

  

  

Rather than suggesting something that is defunkt by default, yes.

  





I Absolutely agree. I will add that as it's coming from LAD it should
ideally be the most efficient and future proof option with fallback for
situations where it is not possible to implement the complete
recommendation.



I consider this approach part of the fallback.







  

  

    
BTW, it would be interesting to get some real world feedback from
anyone who has this enabled already on a 32 bit platform.

    

  

  

This would most likely be enabled by default for anybody who has
installed a 32bit distro on a machine with 8GB memory.


  





Hmmm, I wonder how many people are in that situation?



IIUC the only way to get 8GB or RAM onto a normal mobo is with a quad
core 64 bit system. Are you suggesting that there are a lot of people
who have one of these beasts who are running 32 bit OS on it?



I know I'm not.



Just out of interest I wonder if Fons would be willing to run 32 bit on
the new system  he is looking at purchasing?



Can we get a show of hands? Is anyone running 32 bit with 8Gb of
memory? How about 64 bit with the same specs?







Cheers.








Patrick Shirkey
Boost Hardware Ltd