All, Xeon advantages are faked out:
Recently "Tenchusatu" posted an "analysis"
of ZD bencmarks for "identically"
configured computers from Gateway, G6-450 (PeeII)
and GX450-XL (Xeon):
Message 6920852
where he noticed that the Xeon Winstone99
performance was 7% better than the "equivalent"
P-II. Upon this observation Tenchasatu
started to promote a theory of significant
superiority of full-speed Xeon caches vs.
P-II.
Message 6973712
However, a little more careful
inspection of benchmark data reveals
that the P-II-based machine was crippled
in hard disk part. To see this you need
to look at corresponding Winbench99 from
the same publication:
ftp://ftp.zdnet.com/pcmag/1998/1215/labsdata/m22p1b1b.xls
zdnet.com
Just look at three models: Dell XPS R450 (PeeII),
and the two from GTW, and compare their Disk Winmarks:
Model CPU Drive Disk Winmark
-----------------------------------------------
Dell XPS R450 P-II IBM DTTA-371010 3570
GTW G6-450 P-II IBM DTTA-371440 2460
GTW GX450-XL Xeon IBM DTTA-371440 3660
-----------------------------------------------
All three models are using the same line
of IBM hard drives, so the disk performance
by definition should not differ too much. And
it is easy to see: the Dell with P-II delivers
the same disk score as the super-Xeon model
from GTW. However, for some "reason", the GTW
P-II model, with the same drive, appears to
be 33% slower! Now it is easy to find the
true reason as why the normally configured
Xeon model posted 7% better scores than the
same but misconfigured P-II counterpart,
with crippled disk subsystem. Since the disk
time in Winstone99 constitutes at least 30% of
all run time, it is quite possible that the
properly configured GTW G6-450 could post
even better scores than the Xeon system!
In conclusion, the application performance
advantage of Xeon is highly overstated.
Have a nice whatever, Mr. Tenchusatu and
Mr Phud altogether. |
