32bit vs 64bit

[Bill Davidsen]

Bingo wrote:
> Read this for a comparison in performance for 32 bit, and 64 bit linux: 
> http://bingouv.blogspot.com/2008/08/desktop-linux-performance-comparison32.html 
> .
> 
> To summarize, you will notice the difference between 32 bit and 64 bit 
> for CPU intensive tasks. But it won't change the world. This comparison 
> is based on Fedora-9.
> 
One thing that most people undervalue is that almost none of these tests are 
actually done on 32 bit and 64 bit processors, they are widely done on the same 
processor (to keep thing the same) and the 32 bit test is run on a 64 bit CPU in 
32 bit mode.

This is important because the big difference between 8 and 16 bit, or 16 and 32 
bit was the longer registers and wider path to memory. And so the tests were run 
on different CPUs. Running the test on the same CPU in two modes gives far less 
improvement, and most recent tests measure only the difference between modes.

Since the three advantages of 64 bit operation are lower overhead with large 
memory, ability to run larger user images, and a few more 64 bit registers, it's 
not surprising that the benefits are not huge, and that many systems will just 
not see any measurable difference, due to a lack of large processes or memory. 
The overhead of PAE operation is generally hard to find in the noise.

Obviously there are loads which do show a larger difference, and some of those 
loads are "real loads" rather than benchmarks and test cases, but for most 
configurations and loads it can be "measured but not felt."

Note: this is just to remind people of the factors involved, not to agree or 
disagree with anyone.

-- 
Bill Davidsen <davidsen at tmr.com>
   "We have more to fear from the bungling of the incompetent than from
the machinations of the wicked."  - from Slashdot