[Wizard] New kernel

David English [email protected]
Sun Aug 31 05:12:37 2003


Just worth pointing out that whilst I get 49 bogomips calculated with 
the bf24 (2.4.18) kernel, and 459 bogomips with a stock 2.6.0-test4 
kernel, the bf24 kernel reports almost half as much cpu usage for 
running the same compiles of ogg123 and mpg123, and whilst this cpu 
usage is not necessarily accurate, the speed of other concurrent tasks 
suggests it is pretty well true. The mistaken bogomips seems to have 
little, if anything to do with the actual performance.
The only thing I did find that made a massive difference performance 
wise was setting the ram divisor from 4 to 3 in the bios, cas2 makes 
less of a difference but seems stable enough on the system I'm using atm 
(lasted the kernel compile, etc.)
I'm recompiling 2.6 with 486 optimisations to see if that fixes the 
performance problems, though obviously this is not particularly useful.

- Dave