[Future Technology Research Index]
[SGI Tech/Advice Index]
[Nintendo64 Tech Info Index]
[WhatsNew]
[P.I.]
[Indigo]
[Indy]
[O2]
[Indigo2]
[Crimson]
[Challenge]
[Onyx]
[Octane]
[Origin]
[Onyx2]
(check my current auctions!)
Using Viewperf to Compare Graphics Performance
I could say quite alot on this subject, but for the moment here's
something to bare in mind when reading Viewperf results:
- When looking at the performance difference between two different graphics
options for the same machine, make sure the main CPU is the same in
both cases.
Basically, if the main CPU is not the same then the benchmark figures
are hard to compare and statistically unreliable. A good example is
comparing the following two systems (I bet someone, somewhere, has
made this particular mistake in the past):
- System A: Indigo2/XZ with R4400 150MHz
- System B: Indigo2/Extreme with R4400 200MHz
Suppose you currently own system A and have just received a data
sheet for system B. You're browsing through the polygon/sec,
XLines/sec (etc.) numbers... how do you know these differences are
down to the new graphics system (Extreme) and not the fact that
System B was tested with a faster CPU? Often, the CPU used in the
test is listed in small print somewhere and it's almost always the
top of the range available. Like as not, you'll be considering a
graphics upgrade, not a graphics + CPU upgrade.
Viewperf is a nightmare when it comes to this aspect of graphics
performance. If you check the SPEC web site, then for a particular
system (eg. Indigo2 or AlphaStation) you will find:
- The same graphics option shown in use with different CPUs,
- The same CPU in use with different graphics options.
Naturally, PR people can use this to great effect. Digital did this
on one of their AlphaStation/Indigo2 comparison WWW pages: they
mixed performance results from Viewperf data that came from
different Indigo2 configurations which had been submitted to
SPEC by SGI. For each metric mentioned, they took the worst
aspect from a particular configuration and used that in the table
shown. The end result? The table implies that the data is from the
Indigo2 system which they said was tested, when in fact much of the
data came from Indigo2 systems with slower processors,
less memory, older graphics options and outdated
performance figures.
All in all, the particular page in question had over 10 factual
inaccuracies and more than 15 logical combinational errors.
Don't get me wrong though: I've no doubt that every workstation
vendor gets up to this kind of PR nonsense, but if you're going to
use Viewperf and other published data when making a purchasing
decision then you must make sure you know what it is you're actually
looking at, especially if the data comes from a company web
site.
(check my current auctions!)
[WhatsNew]
[P.I.]
[Indigo]
[Indy]
[O2]
[Indigo2]
[Crimson]
[Challenge]
[Onyx]
[Octane]
[Origin]
[Onyx2]
[Future Technology Research Index]
[SGI Tech/Advice Index]
[Nintendo64 Tech Info Index]