[rescue] Sun Kit Needed for EE Student Here

["Javi Mahai <lefa at ucsc.edu>" at ucsc.edu]

On Wed, 03 May 2006 07:07:31 -0400
  Patrick Giagnocavo <patrick at zill.net> wrote:

> I wonder whether part of the speedup is due to faster 
>memory bandwidth
> in the Opterons.

The opteron does more per clock than the USIII/USIV, at 
least from our benchmarks. I was surprised to find that 
out, even when compared to the dual G5 sitting under my 
desk, for 99% of the code we have thrown at it... the 
Opteron was as fast/faster than the G5 that was a few MHz 
faster than the Opteron, so AMD has done a fine job with 
their microarchitecture (I was rather impressed).

> 
> Isn't anything over a plain US-II also capable of out of 
>order
> execution?

Nope, SUN never got any out-of-order machine out. The 
UltraSparc V got cancelled because they were out of budget 
power and complexity wise. The only guys with out-of-oder 
in the SPARC arena are Fujitsu, and the research lab I 
work at (we are producing an OOO SPARC chip). The latest 
SPARC64 are the only competitive SPARC parts out there 
when it comes to single thread performance. Unfortunatedly 
a lot of the code in the EDA field is not easily 
threadable or has little parallelism, so SUN machines get 
crushed when working with things like Synopsys et al. It 
is hard to justify buying a new SPARC system when an X86 
system costing far lest is over twice as fast for most 
tasks. Hopefully SUN will soon offer SPARC64 boxes....