Digital BW, The Print

> To get any significant increased Performance of a  Xeon PIII over a
> Coppermine PIII (and that mostly in database type apps) you need more
> than 256k on die L2 cache...

The current Xeons have 512k.

> There is no way in hell you will get more than a 10% performance
> increase in PShop over a similar Coppermine equipped PIII...

That's an interesting statement. Aside from the differences in architecture,
the fastest Xeon is twice as fast as the fastest PIII, 1.4ghz vs 2.8ghz. I
think you'll see much more than 10% difference.

> As you will read there, even SMP won't give more than a max of 15% on
> average in  PShop.. (Filtering operations may IMHO be about 60% faster
> on a dual CPU than a single CPU -- so a dualie only makes sense if like
> I,  you are doing a lot of filtering, and the additional L2 cache of the
> Xeon ain't gonna boost that much..)

I'd agree with that. Dual CPU's only help in some situations, and only when
the application can take advantage of the hardware. In Photoshop, the
performance advantage with dual CPU's varies from 0% to probably 80%.

> Since Pshop still uses a lot of  disk caching (scratch disk), me
> suspects the RAID array makes a big difference in Moreno's numbers...

These were seat-of-the-pants comparisons, and the only RAID was in the
Athlon system. The files were only 100mb, and the cache was cleared before
each run, so everything took place in ram.