Re: Emulator, Inc

[Sheldon Simms <sheldonsimms@yahoo.com>]

On Mon, 13 Oct 2003 00:19:26 -0700, Eric Smith wrote:

> "Bryan Parkoff" <nospam@nospam.com> writes:
>> I know that Pentium 4 has some problems, but I stay on Xeon Pentium III
>> until Pentium 4 will be fixed.
> 
> I wrote:
>> What problems are these?
> 
> Sheldon Simms <sheldonsimms@yahoo.com> writes:
>> All that, however, it hardly matters since the Pentium 4 is capable
>> of being run at much higher core frequencies, which makes up for the
>> cycle for cycle disadvantage.
> 
> Yes, but that's a deliberate design tradeoff that Intel made.  It's not
> like it was a screwup.

I wasn't trying to argue that it was a screwup, just expounding on
the "problem" that Bryan was talking about.

However, at one time one *could* argue that it was a screwup. The
original (Willamette) Pentium 4 was slower than the competition,
even with a core frequency almost twice as high.

The main thing that has saved the Pentium 4 is Intel's ability
to push the process technology and include a lot of on-chip cache.

Lastly, Bryan has been reading the Emulator, Inc site. There's
a huge anti-Pentium 4 rant there (dating from the Willamette time)
which is largely based on the fact that the Pentium 4 lacks
barrel shifters. I guess the author of the rant was disappointed
to see the performance of his Pentium III-optimized assembly code
go into the toilet when all of those rotates and shifts hit the
Pentium 4 barrel-shifterless ALUs.

Personally, I find the AMD architecture nicer than the Pentium 4,
but I expect the Pentium 5 to be a big improvement.