How do I change from a regular kernel to a PAE kernel ?

[Steven W. Orr]

On 09/30/09 01:27, quoth Linuxguy123:
> On Wed, 2009-09-30 at 00:43 -0400, Steven W. Orr wrote:
>> On 09/30/09 00:21, quoth Linuxguy123:
>>
>>>>> What difference in speed will I notice in doing this ?
>>> Evolution seems to be faster, but 'free -m' isn't showing that I am
>>> running out of memory or anything.   I generally have ~1600 MB free now.
>>> I gained 1 GB, so before it would have been ~400 MB free. 
>>>
>>> Now I should increase the size of my swap file partition from 2GB to
>>> 4GB...  
>>>
>> I just did this last week. Works fine. But I just have one question:
>>
>> You say your situation is that you're now running at around 1.6G free mem. So
>> why bump the swap area? You can do it, but the whole point was to use more
>> memory. If you didn't run out of swap before then you're even less likely to
>> run out now, unless you now plan on doing things that you didn't do before.
> 
> I'm running 1.6 GB free when not running any of my heavy duty
> applications like ufraw (batch), digikam, eclipse (sometimes 2 or more
> instances), gimp, a couple browsers, evolution and a couple Open Office
> apps. 
> 
> Did I mention that I want to start editing HD video ?
> 
> I need 8 GB of RAM ! 
> 
> BTW: The PAE kernel seems to run faster. 
> 

I seriously doubt that the PAE kernel is running faster. If anything, I would
suspect it of running slower because of the added overhead of a page table
algorithm that now handles a larger address space than it did before. I could
very well be wrong since I have not looked at what the PAE kernel actually does.

My point is that people who think something is faster by using the
non-scientific touchy-feely method are the same people who get brain cancer
from cell phones. If you didn't take measurements before adding the extra RAM
 with the old non-PAE kernel, then you have no basis for commenting on
performance improvement. Change one thing and measure each time. That's the
only way. In your case you added RAM and then changed the kernel to PAE to be
able to access all of the new RAM. Since register to register access is about
~1ns, and register to memory access is about ~10ns, and a disk access is
around ~10ms, then any extra ram you throw at a problem may end with a speed
increase of up to ~>700 times. You didn't actually take *any* measurements so
you really have no basis for saying that a PAE kernel is faster than a non-PAE
kernel. If you really want to tell if it is then you need to boot with a
non-PAE kernel and try to do something in less than 3.2G of ram that doesn't
access the disk and then run the same test with a PAE kernel with no added ram.

And we're all above average in bed too. ;-)

-- 
Time flies like the wind. Fruit flies like a banana. Stranger things have  .0.
happened but none stranger than this. Does your driver's license say Organ ..0
Donor?Black holes are where God divided by zero. Listen to me! We are all- 000
individuals! What if this weren't a hypothetical question?
steveo at syslang.net

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 261 bytes
Desc: OpenPGP digital signature
Url : http://lists.fedoraproject.org/pipermail/users/attachments/20090930/a168351a/attachment-0001.bin