Ade <ade@here.invalid> wrote in message

> It is interesting to see you say about your dissapointment with XVR-1000
> performance. I got to play with an XVR-500 not that long ago and was
> surprised at just how spritely it was relative to the XVR-1000.
> Considering the amount of money it cost us when it was launched, the
> performance has been very dissapointing. The Sun document referenced by
> the other poster is interesting in that it shows a significant
> performance deficit from what i would have expected previously.

Don't get me wrong: I like the XVR-1000 in my 900Mhz SB2K (I got the
workstation -new- for very little off a bankrupt company in '02 so I
can't complain) and it drives the SUN 24" LCD just fine but it's true
that its performance for general use (I.E: not just Pro/E or CATIA) is
a bit disappointing. I started wondering about that issue when I
played with SPECviewperf 6.1.2 on my sb2k and got about 1/10th the
numbers of a basic PIV with an NVidia Quadro4 550 card (I reproduced
the SUN SPEC numbers for my hardware very closely so it's not just my
mistake.. ).

The XVR-1000 is a great card with lots of RAM and features but it
tends to lack behind a bit in performance. SUN apparently agrees with
that issue as some OpenGL extensions are unavailable by design on the
XVR-1000 as they would be too slow on it (GL_ARB_texture_cube_map for
instance, was available on PGX64 in OpenGL 1.3 but a software fallback
wasn't even provided on the XVR-1000). Anyway the XVR-1000 will be
EOL'ed soon so it's not so much an issue anymore...

It's great to see SUN move toward industry-standard graphics chipset
(I.E: 3DLabs WildCat IV) but I hoped they wouldn't just provide us
with 33/66Mhz PCI graphics cards after all the APG8x vs. UPA2 hype..
At last, PCI-X is coming so perhaps we'll see a SUN version of the
3Dlabs VPro 990 (2GFlops) supported in a current line of workstations
some time in the future..

I had almost given up on finishing the OpenGL client port of the
Torque engine to Solaris/Sparc because of the (relative) lack of fps
on my XVR-1000 (not to mention the other cards I tried) when I had the
chance to test my port on an XVR-1200 and was very happily surprised
by its performance there. In fact, the difference in end-user
performance was much more (for my app) than what the numbers suggested
(XVR-1200 vs. XVR-1000) so I started to wonder if OpenGL features'
hardware acceleration wasn't directly related. I hope the XVR-600 will
be able to reproduce this performance (I'm concerned about its lack of
GL_ARB_multitexture, unlike the XVR-1200) but I sure cannot buy an
XVR-1200 (anyone want to donate one? ).

Vincent S. Cojot, coyote@step.polymtl.ca