slow last 128MB of RAM in a 2GB system?

Bill McGonigle bill at bfccomputing.com
Fri Apr 20 13:34:54 EDT 2007


Just back to this after storms and taxes...

On Apr 13, 2007, at 16:49, Dave Johnson wrote:

> Simply add this to a startup script and you should be all set:
>
> echo "base=0x7c000000 size=0x2000000 type=write-back" >/proc/mtrr

Awesome, this seems to work perfectly.  Some timing tests, using  
'time yum info' as a metric:

kernel line with mem=1920M:
   52.041u 9.184s 3:38.98 27.9%    0+0k 0+0io 78pf+0w

stock kernel line:
   1871.087u 18.762s 59:22.53 53.0%        0+0k 0+0io 68pf+0w

immediately after adding the extra mtrr entry:
   43.896u 7.113s 1:05.10 78.3%    0+0k 0+0io 0pf+0w

So, without this entry the machine runs either 43 or 55 times slower,  
depending on if you're counting CPU time or wallclock time,  
respectively.  Ouch - is that simply a matter of cache impact on  
performance?  I wouldn't have guessed it would be so high.

The speed increase in the last one might be partially due to RAM  
cache from the previous run, but I also see fewer page faults.   
Anyway, I'm glad to have similar performance and get to use a bit  
more RAM (kernel reports 2038696k available).

I'm even more happy to understand the WHY.  Actually I need to go re- 
read your explanation a couple more times before I get it all through  
my thick skull.  I'll see what I can get out of the manufacturer, or  
at least post a warning up at NewEgg.

Many thanks!

-Bill

-----
Bill McGonigle, Owner           Work: 603.448.4440
BFC Computing, LLC              Home: 603.448.1668
bill at bfccomputing.com           Cell: 603.252.2606
http://www.bfccomputing.com/    Page: 603.442.1833
Blog: http://blog.bfccomputing.com/
VCard: http://bfccomputing.com/vcard/bill.vcf

For fastest support contact, please follow:
http://bfccomputing.com/support_contact.html





More information about the gnhlug-discuss mailing list