On Fri, Apr 3, 2009 at 2:54 PM, Matthew W. Ross
<[email protected]> wrote:
> Isn't PAE a non-trivial memory performance hit? Or perhaps it used to be 
> non-trivial,
> but newer memory technologies are fast enough to make it trivial today?

  I don't know the details, but I believe so.  PAE has two major
effects that I'm aware of:

* The physical address word increases from 32 bits to 36 bits.
* A third level of page table indirection is introduced, to allow the
MMU to map the additional memory with minimal architectural change.

  I can't see how additional page table indirection could *not* slow
things down.  And it wouldn't surprise me to find that the large
physical address word means slower operations need to be used for
memory management.

  I don't know much about the memory architecture of long mode
(64-bit), but it may be that the page table layout is less cumbersome,
being designed for larger memories from day one.  There may be a
performance gain from running Win64 for that reason, even if one only
has a smaller memory system and Win32 software.  Anyone have
benchmarks?

  (Obviously a Win64 system will be a win if one has larger memories,
but I'm curious if it would pay off on, e.g., a 1 GB RAM system.)

-- Ben

~ Finally, powerful endpoint security that ISN'T a resource hog! ~
~ <http://www.sunbeltsoftware.com/Business/VIPRE-Enterprise/>  ~

Reply via email to