| Re: survey: accelerated 3D volumetric rendering [message #23934] |
Wed, 28 February 2001 09:47 |
Paul Woodford
Messages: 43
Registered: June 2000
|
Member |
|
|
In article <97j5ma$d01$1@canopus.cc.umanitoba.ca>,
"Richard Tyc" <richt@sbrc.umanitoba.ca> wrote:
> One interesting thing to note: on our dual Pentium III 733Mhz Dell, the real
> time rendering of volume objects is not that much faster than my single CPU
> office PC (only a Pentium II 350). I would have thought it was at least 2x
> faster running the same app, but not so with my simple subjective test.
I once did a IDLgrVolume rendering speed test with a dual-processor
Windows NT PC, and found almost a 2x speedup when I set the hints
property to use multiple processors. I think the volume was something
like 128 x 128 x 32.
--
Paul Woodford, Ph.D.
Essex Corporation
For faster email response, replace us dot net with essexcorp dot com
|
|
|
|
| Re: survey: accelerated 3D volumetric rendering [message #23939 is a reply to message #23934] |
Wed, 28 February 2001 07:31  |
Richard Tyc
Messages: 69
Registered: June 1999
|
Member |
|
|
David Fanning <davidf@dfanning.com> wrote in message
news:MPG.1505e00a9af63e39989d6e@news.frii.com...
> Rick Towler (rtowler@u.washington.edu) writes:
>
>> What are peoples experiences with accelerated 3d volumetric rendering.
I am
>> sure very high end unix viz workstations have the ability to accelerate
>> volumetric rendering but what about lower end hardware. Is this the
domain
>> of high end video adapters only?
>>
>> In my case we have a Sun Ultra60 with the Creator3d framebuffer and PC's
>> running consumer versions of nvidia's Geforce line. The Creator3d is
>> painfully slow rendering anything. The Geforce cards are quite fast
with
>> your standard polygon rendering but volumetric rendering isn't supported
in
>> hardware.
>>
>> Does anyone have any experience with this using nvidia's Quadro line or
with
>> 3dLabs cards? What about other platforms?
>>
>> fwiw, Sun just released the Expert3d lite which does support accelerated
>> volumetric rendering and when bundled runs for $2000. I guess that is
low
>> end....
>
> My experience with volume rendering with several different
> "inexpensive" graphics cards for PCs is that software
> rending is *always* faster than hardware rendering. (Not
> to mention prone to far fewer rendering errors.)
>
It is my understanding that for volume rendering (using IDLgrVolume) you
have no choice but software rendering ( its a software based ray tracing
scheme) regardless of the hardware/software rendering switch. IDL does not
yet include OpenGL support for volume rendering so advanced graphics cards
would not help anyway. Multiple CPU's will help since you can set the HINTS
property to use multiple CPU's. Until a standard OpenGL volume rendering
scheme becomes available on all platforms IDL supports, they may not jump
into it. I hope they do.
I got into this a while back when I was looking at a new PC workstation for
our lab. I was considering buying the high end Wildcat 4110 card which is
blistering fast, especially on our CAD/CAM machines. I opted for a dual
Pentium Dell workstation with a G400 dual head card to speed up real time
viewing/rendering of our volume objects.
One interesting thing to note: on our dual Pentium III 733Mhz Dell, the real
time rendering of volume objects is not that much faster than my single CPU
office PC (only a Pentium II 350). I would have thought it was at least 2x
faster running the same app, but not so with my simple subjective test. My
crude measurement is based on spinning a 3D object in our medical app (using
Trackball) and visually observing how fast the image is rendered. The worst
thing is , my Dell has 512Mb RAM (Win NT4), office PC has 64Mb RAM (Win 98)
!!
Rich
|
|
|
|
| Re: survey: accelerated 3D volumetric rendering [message #23950 is a reply to message #23939] |
Tue, 27 February 2001 21:30 |
Randall Frank
Messages: 17
Registered: October 1999
|
Junior Member |
|
|
Rick,
Ok, I'll bite. I don't want to get into the latter part
of this thread, but I see major speed increases in 3D performance
with GeForce and Quadro cards in HW over SW with a couple of notable
exceptions.
So, you want to volume render in HW well, you can do 3D
texturing or 2D texturing. With 2D texturing, you can use axis
aligned slices or shear-warp. Actually, many of the 2D texturing
methods can be done in IDL right now (a long time ago, I wrote
a HW volume rendering object in .pro for a demo). Volume
rendering is really a matter of fill rates, so standard PC gaming
HW does a very nice job. See any of the work by Ertl from
IEEE vis (code is available for download). The combination
of 2D texturing and multi-texturing works very well.
Rick Towler wrote:
>
> What are peoples experiences with accelerated 3d volumetric rendering. I am
> sure very high end unix viz workstations have the ability to accelerate
> volumetric rendering but what about lower end hardware. Is this the domain
> of high end video adapters only?
>
> In my case we have a Sun Ultra60 with the Creator3d framebuffer and PC's
> running consumer versions of nvidia's Geforce line. The Creator3d is
> painfully slow rendering anything. The Geforce cards are quite fast with
> your standard polygon rendering but volumetric rendering isn't supported in
> hardware.
The (early) Creator3D was a rasterization only engine. All triangle
transforms and lighting are actually in SW. Thus, the card can be
poor with high polygon counts, but actually works pretty well for
volume rendering applications.
The GeForce 3 (announced this week) does have 3D textures if that is
what you mean by "volumetric rendering". Up to 512^3 in size. However,
the effective framebuffer fill rate of that card has not increased over
the GeForce 2. Thus, it will not really be much faster for volume
rendering until someone writes a renderer using the pixel pipeline to
move "chunks" of the blends into the pixel combiner paths (which
could be done). In the current class of cards, the HP, 3D labs and
ATI cards all have 3D texturing (actually, the nVidia drivers support
it, but it is in SW until the GF3) and there are demos on their sites.
If you are looking at the low end, the ATI card is less than $300.
You can also use the multi-texturing stuff on the GeForce to get a
good approximation of 3D textures (again, see the Ertl papers and
web site) for around $400. Or, wait for the GF3 for around $600.
For many applications, axis aligned planes with 2D texturing will
work. The key to volume performance is fill rate, so it is hard
to beat a VooDoo 5500 or GeForce 2 Ultra today (if used correctly).
Several people around here use the HP fx10 for volume rendering
and like it as well. Finally, the 3D Labs GVX420 and Wildcat
cards work pretty well for volumes.
Bottom line, I would suggest you not get too caught up in
equivocating 3D texturing with volume rendering. This will get
very murky soon with DX8 and cards like the GeForce 3. You can
definitely get good volume rendering in low end PC cards today.
3D textures are available for several PC cards today as well.
Now, how this relates to IDL, short of my initial comments, I am
not sure... and this is all IHMO.
Ref:
http://wwwvis.informatik.uni-stuttgart.de/eng/research/pub/p ub2000/wgh00-rezk.pdf
Hope it helps.
>
> Does anyone have any experience with this using nvidia's Quadro line or with
> 3dLabs cards? What about other platforms?
>
> fwiw, Sun just released the Expert3d lite which does support accelerated
> volumetric rendering and when bundled runs for $2000. I guess that is low
> end....
>
> -Rick Towler
--
rjf.
Randy Frank | ASCI Visualization
Lawrence Livermore National Laboratory | rjfrank@llnl.gov
B451 Room 2039 L-561 | Voice: (925) 423-9399
Livermore, CA 94550 | Fax: (925) 423-8704
|
|
|
|
| Re: survey: accelerated 3D volumetric rendering [message #23956 is a reply to message #23950] |
Tue, 27 February 2001 17:42 |
davidf
Messages: 2866
Registered: September 1996
|
Senior Member |
|
|
Rick Towler (rtowler@u.washington.edu) writes:
> David, you are always pushing that software rendering..... :) I can't wait
> for the day the software renderer is retired.
Well, if nothing else, it's useful for comparing with
hardware rendering. As you well know, if you have done
any object graphics programming at all, you don't get
much feedback sometimes. I've spent hours and hours
trying to figure out what I was doing wrong programmatically,
only to discover there is a bug in my hardware OpenGL
implementation. (Number 9 VooDoo 3D, a hot card in its
day. But of course the company is now in Chapter 11).
I'd buy an NVidia card if I thought I could get it
installed without causing my computer to go bonkers for
two weeks. :-(
> Software rendering is useless for real time rendering of 3d object graphics
> scenes (at least the scenes I work with).
You and I are obviously rendering different 3D scenes. :-)
I don't find it useless at all. But I can imagine scenarios where
that could be the case.
> Most people probably shy away from volumes but I thought at least a few med
> imaging people would pipe up.
I think those guys have real jobs. Probably don't
have much time to debate on the newsgroups. :-)
Cheers,
David
--
David Fanning, Ph.D.
Fanning Software Consulting
Phone: 970-221-0438 E-Mail: davidf@dfanning.com
Coyote's Guide to IDL Programming: http://www.dfanning.com/
Toll-Free IDL Book Orders: 1-888-461-0155
|
|
|
|
| Re: survey: accelerated 3D volumetric rendering [message #23958 is a reply to message #23956] |
Tue, 27 February 2001 17:08 |
Rick Towler
Messages: 821
Registered: August 1998
|
Senior Member |
|
|
David, you are always pushing that software rendering..... :) I can't wait
for the day the software renderer is retired.
Software rendering is useless for real time rendering of 3d object graphics
scenes (at least the scenes I work with). It is true that hardware
rendering comes with quirks but at least on the PC the latest generation of
nvidia products are quite good. It all comes down to the drivers and in the
consumer market nvidia is the only company that I know of shipping decent
OpenGL drivers. The professional market is another question.
Most people probably shy away from volumes but I thought at least a few med
imaging people would pipe up.
-Rick Towler
"David Fanning" <davidf@dfanning.com> wrote in message
news:MPG.1505e00a9af63e39989d6e@news.frii.com...
> Rick Towler (rtowler@u.washington.edu) writes:
>
>> What are peoples experiences with accelerated 3d volumetric rendering.
I am
>> sure very high end unix viz workstations have the ability to accelerate
>> volumetric rendering but what about lower end hardware. Is this the
domain
>> of high end video adapters only?
>>
>> In my case we have a Sun Ultra60 with the Creator3d framebuffer and PC's
>> running consumer versions of nvidia's Geforce line. The Creator3d is
>> painfully slow rendering anything. The Geforce cards are quite fast
with
>> your standard polygon rendering but volumetric rendering isn't supported
in
>> hardware.
>>
>> Does anyone have any experience with this using nvidia's Quadro line or
with
>> 3dLabs cards? What about other platforms?
>>
>> fwiw, Sun just released the Expert3d lite which does support accelerated
>> volumetric rendering and when bundled runs for $2000. I guess that is
low
>> end....
>
> My experience with volume rendering with several different
> "inexpensive" graphics cards for PCs is that software
> rending is *always* faster than hardware rendering. (Not
> to mention prone to far fewer rendering errors.)
>
> I pretty much always have software rendering on as the
> default, and I make *sure* I have it on for any object
> graphics programs I distribute that have anything unusual
> going on in them.
>
> Cheers,
>
> David
> --
> David Fanning, Ph.D.
> Fanning Software Consulting
> Phone: 970-221-0438 E-Mail: davidf@dfanning.com
> Coyote's Guide to IDL Programming: http://www.dfanning.com/
> Toll-Free IDL Book Orders: 1-888-461-0155
|
|
|
|
| Re: survey: accelerated 3D volumetric rendering [message #23964 is a reply to message #23958] |
Tue, 27 February 2001 15:03 |
davidf
Messages: 2866
Registered: September 1996
|
Senior Member |
|
|
Rick Towler (rtowler@u.washington.edu) writes:
> What are peoples experiences with accelerated 3d volumetric rendering. I am
> sure very high end unix viz workstations have the ability to accelerate
> volumetric rendering but what about lower end hardware. Is this the domain
> of high end video adapters only?
>
> In my case we have a Sun Ultra60 with the Creator3d framebuffer and PC's
> running consumer versions of nvidia's Geforce line. The Creator3d is
> painfully slow rendering anything. The Geforce cards are quite fast with
> your standard polygon rendering but volumetric rendering isn't supported in
> hardware.
>
> Does anyone have any experience with this using nvidia's Quadro line or with
> 3dLabs cards? What about other platforms?
>
> fwiw, Sun just released the Expert3d lite which does support accelerated
> volumetric rendering and when bundled runs for $2000. I guess that is low
> end....
My experience with volume rendering with several different
"inexpensive" graphics cards for PCs is that software
rending is *always* faster than hardware rendering. (Not
to mention prone to far fewer rendering errors.)
I pretty much always have software rendering on as the
default, and I make *sure* I have it on for any object
graphics programs I distribute that have anything unusual
going on in them.
Cheers,
David
--
David Fanning, Ph.D.
Fanning Software Consulting
Phone: 970-221-0438 E-Mail: davidf@dfanning.com
Coyote's Guide to IDL Programming: http://www.dfanning.com/
Toll-Free IDL Book Orders: 1-888-461-0155
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
