XANSYS Message: 17032 [Go back to message list]
[bookmark on del.icio.us]
No rating yet
Rate item:

Subject: Re: latest NT graphics cards for ANSYS.... ? inter im result.....
Author: Sheldon Imaoka
Date: 2000-10-24 22:59:00

/DEMO is very cool for graphics performance, but what I've found is that if
using 3D graphics, the cursor should not be on the graphics window;
otherwise, much slower (2x) speed times will be reported. I never bothered
to figure out if this was due to my crappy graphics cards or with ANSYS, but
this was exhibited on a few different PCs using 5.5.x and 5.6.x. (I can't
recall the specific versions since it was a while ago, but I just checked,
and the same thing occurs at 5.6.2 on my laptop -- then again, the ATI Rage
Mobility-P on a laptop isn't one which screams, anyway) You can try this --
issue something like /DEMO,100,1 and move your cursor over the Graphics
window, and the plots will be about twice as slow. Hence, when doing
benchmarks/comparisons, please keep this in mind, if this affects you.

Another thing I've never figured out is why, when comparing /GRAPH,POWER and
/GRAPH,FULL, that speed times don't vary dramatically for 3D graphics.
Attached is a small PDF file showing some differences between win32 and 3D
graphics. (Remember that this is done just now on my notebook, so it's not
with a good graphics card) It's a model of a meshed cube. With 3D
graphics, the differences between Full and PowerGraphics seem minimal. I
see a similar trend with many (but not all) other types of ANSYS models. I
haven't pinpointed why this is so. Maybe because the object is stored as a
"3D" object, rotation speed is negligible for either a "full" model or
powergraphics object, as long as both fit in memory since OpenGL stores
stuff as a 3D array, or something like that, rather than a 2D graphics
algorithm which does a projection into 2D at every rotation (like
z-buffering algorithm, etc.)... I forgot the details of this stuff. Also,
with win32 graphics, contour plots, element plots, etc. seem faster with
/POWER,GRAPH, but actions like multiplots (GPLOT) aren't as dramatically
faster.

Anyways, I was never much of a graphics performance person, but I like using
/DEMO as Paul mentioned below, with different configurations. I use it
mainly for 3D OpenGL drivers since, for x11 or win32, I don't think the
graphics card is used much (or directly) by that driver. However, it's
interesting to use win32 since many of us probably still use it, although,
if I recall correctly, the win32 results tend to be more related to CPU
rather than graphics card (as noted above)...

Like I mentioned above, I used to think I knew this graphics stuff, but
there are so many permutations, I soon gave up. If it plots fast and
correctly, I'm golden -- anything else, I whine loud and persistently to my
boss (in response to a very hilarious posting of Mark Rodamaker's).

Laters,
Sheldon
CSI

----- Original Message -----
To:
Sent: Tuesday, October 24, 2000 12:55 PM

result.....

> Mike,

> Try /DEMO,100,1 to compare graphics performance.

> This will rotate the current display 100 times and print out the data to
the
> output window.

> Paul Tallon
> ANSYS Inc.

> > -----Original Message-----
> > From: Michael W. Hiller [SMTP:mhiller@i...]
> > Sent: Tuesday, October 24, 2000 3:22 PM
> > To: Xansys
> > Subject: [xansys] Re: latest NT graphics cards for ANSYS.... ?
> > interim result.....

> > thanks for all the suggestions last week on video cards.....

> > I went with the NVIDIA chipset... less risk than the "newer" chipset we
> > discussed.
> > rec'd and installed a GeFORCE 32 meg DDR card card card today....
> > (check out UBID.COM - - they seem to be selling a lot of VisionTeks and
> > the
> > prices are lower now than last week - - winning prices now around $179
vs
> > $189 last week)

> > I wrote a do loop macro to compare my old card (Diamond GL1000Pro) to
this
> > new one... this macro rotates a large model (70,000 brick45, 2500
volumes)
> > and alternates EPLOT and VPLOT. The macro then calculates the DELTACPU
> > and
> > DELTAWALL (from the benchmark macros! thanks PADT!).

> > The deltas are about 25% quicker with the new card.... probably CPU or
> > AGPx1
> > limited now. The interactive spinning is incredibly faster...
virtually
> > no
> > delay compared to my old card....

> > results were even about 2% better with new drivers downloaded from
> > Austin////

> > picked up another 1% using ANSYS 5.7 beta2.... will be trying it on
beta3
> > later in week.....

> > well this was the best $200 upgrade I ever saw..... maybe do the Quadro
> > hack in a couple weeks....

> > well thanks for all your suggestions.....

> > I did not mean to turn XANSYS into a hardware discussion, but with all
the
> > consideration being given recently to the batch CPU and Harddrive
> > benchmarks, I thought some "interactive" data might be appropriate....
> > after all, I think most of us spend the majority of our time in front of
> > the
> > tube doing preprocessing anyway......
> > enuf said....

> > take care and regards,

> > Michael W. Hiller
> > In2 Solutions, Inc.
> > MCAE Consultants specializing in Multi-Physics ANSYS FEA and Pro/E
> > Modeling
> > and Design Automation
> > "Offering consulting, training, implementation, and other individualized
> > solutions for your product design and analysis challenges"
> > 229 Strathburgh Lane Suite 201
> > Cary, NC 27511
> > Tele: 919/851-0373
> > Fax: same number, Tele first to arrange
> > or E-FAX: 253/276-9344
> > Email: mhiller@i...
> > Visit our newly expanded WEB site: www.in2solutions.com

Attachment
graph.pdf
Type: application/pdf
Size: 8k Download

Posts possibly associated with message #17032AuthorDateScore
17028Re: latest NT graphics cards for ANSYS.... ? inter im result.....Paul Tallon2000/10/24 
17032Re: latest NT graphics cards for ANSYS.... ? inter im result.....Sheldon Imaoka2000/10/24 
17039Re: latest NT graphics cards for ANSYS.... ? inter im result.....John Sims2000/10/25 
17059Re: latest NT graphics cards for ANSYS.... ? inter im result.....Chris Hawkins2000/10/25 
17065Re: latest NT graphics cards for ANSYS.... ? inter im result.....Sheldon Imaoka2000/10/25