> hm, the P55C has it's (tiny) performance gain from doubled L1 caches, 
> IIRC, and maybe another organization of associativity of it's 
> cachelines (just a thought).
>
yes the double L1 cache is important, and the associativity was 
increased to 4way. i think a important part is the addional MMX-unit, 
which push apps with alot of fpu instructions. But we leaving the focus 
of this mailinglist ;-)

to put some more results, just for some kind of `destoying legends` 
like: Sparc is twice as fast as a pentium.
I love all my Pizzabox SparcStations, but I hate those performance legends.
If somebody is interessted I can also add gcc options, but i have to 
look into my database.

more oggenc results done with NetBSD 3.0:
1x    AMD    K5 Model 0        75    17000.381sec                     
ASUS TXP4  (Intel 430TX), 512kB L2, 256MB RAM, Seagate ST310211A
1x    Intel    Pentium P54C    75    17735.011sec                    
ASUS TXP4  (Intel 430TX), 512kB L2, 256MB RAM, Seagate ST310211A
1x    AMD    Am5x86 X5       133    23462.422sec                    
Aopen AP43 (SiS 85C496), 512kB L2, 128MB RAM, Seagate ST310211A
1x    Sun    SuperSparc          36    29942.42sec                     
SparcStation 10, -, 512MB RAM, Seagate ST51080N
1x    Ross    7C601 \w 7C605 & 7C602    40    38548.668sec  SparcStation 
20, -, 512MB RAM, Seagate ST32550WC
1x    Motorola    68040           25    96937.706sec                   
Centris 660AV, -, 68MB RAM, IBM DCAS-34330

because hp700 isn't very usefull at the moment i used OpenBSD 3.8,
but i think this values are wrong, because with other benchs got a 
better result
1x    HP    PA-7100                 50    63222.390sec         
           715/50, 64kB L1, 256MB RAM, Seagate ST31230N