3DS video card choices
13 messages in this thread
You know, everybody says that their board is n% faster than the other guy's
board because they use the xxx chip. However, they are talking about WINDOWS
speed and that is misleading, especially here as this is the 3D Studio /
Animator forum and they (as well as Acad r12) are DOS based programs. How
about some discussion that is _meaningful_ to us 3D Studio and Acad users.
All this talk about accelerator chips just clouds the issue. For instance,
the difference in speed from one card to another when running AutoCAD for
DOS seems to have nothing to do with whether the card has a fancy Windows
chip in it, or whether it's on a local bus. It has to do with the DRIVERS.
If there is a display list processor supplied with the card, or in the card's
ROM, it will run Acad fast. If it doesn't, a Windows chip ain't gonna help
any!
As far as 3D Studio goes, I can't seem to get any straight information about
which card will work best for rendering, or flic playback, or redraw speed
in the editor. For that matter, I can't even get any information about
whether the drivers shipped with these cards are any better than the stock
VESA drivers shipped with the program!
Since most of the card manufacturers ship their cards with drivers made by
either Panacea or Vibrant, I can't really see where one card is going to
have much of an advantage over any other anyway.
So if you guys are going to get on the Adesk forums and talk about how great
your hardware is, please address the issues that concern us DOS guys instead
of confusing us with all of this Windows accelerator stuff.
Gregory:
Actually, I only responded to questions from folks on this forum. I didn't
just jump in and start a new thread to hype my Windows speed. I actually
know only a little about 3D Studio — which is why I'm following this forum
— and only a little more about AutoCAD.
On our Graphite cards, our drivers are produced by Vibrant. Perhaps you
can ask Jack to give you an idea of the relative performance of Graphite in
the various Autodesk applications.
Previous messages on this thread did cover a number of issues regarding how
different cards perform in 3D Studio. Some of the responses came from
people who have obviously tested a LOT of cards — perhaps if you posted
some specific questions, they could help answer them for you. I will be
happy, of course, to answer any questions you might have about Hercules'
products.
Jim Carlton
Hercules Computer Technology
Jim-
Sorry, I didn't really mean to jump all over you. I've just gotten
really sick of all the hype and bulls**t going on these days about
accelerator chips, local busses, etc. Your messages were a good springboard
for me to jump from. My comments were meant to be a bit more global and
weren't specifically aimed at you or any one company in particular.
Greg LeVasseur
Greg:
No offense taken. I understand the frustration that users feel, believe
me. Imagine what its like to work a very large deal, only to have it
collapse because some company who isn't even shipping a board announces
that they have something that will do a billion-million ToastAndJam Marks
under Windows. This happens EVERY SINGLE DAY.
Sometimes I think I'd rather drive a long haul semi and see the country.
I agree with the hype message though. If I ever get to "enthusiastic"
about our products, just let me know. I always try to give the facts
without too much advertising. It's obligatory to add some of the company
line since they pay for this account!
Let me know if I can answer any questions for you, either on our products
or on our competitors.
Jim Carlton
Hercules Computer Technology
Good point. Jack Bayt of Vibrant seems to have been talking about new
drivers from his company that address many of the potentially slow areas of
the 3ds interface.
>> … what's all this talk about accel windows !!? (my Emily Latella
impression)
Greg,
Actually, as I understand it (and those who _REALLY_ know what they're
talking about, please jump right in) that the only thing that a driver will
get you is effective and easy access to the available modes of the card.
Thus programs such as the one developed by Vibrant which come with the
Graphite, ATI cards and others allow you to set-up different resolutions and
bit-depths for the editing display and the rendering display (in 3DS). Thus
with cards like the Graphite and the ATI Ultra (+/Pro) you could set your
editing screen to 1024×768 and set the RCPADI to 640x480x24bit…etc.
Since 3DS is caculating each bit and sending this to the cards buffer,
intellegent accelerating chips etc.. to nothing to increase the 3DS
performance (or atleast not much). Any old "dumb" t4000 based system will
perform just as fast in 3DS. But, since the information is being translated
to the card across the bus, the bus throughput and thus VL-bus based cards
are of interest for increasing FLC playbacks (but not for rendering speed).
Since it doesn't really make a "hill of beans" difference to 3DS which
processor your VGA is based, the difference then becomes the performance of
the card on other frontiers, such as windows. Many (or most) of the users
of 3DS also use Windows and ACAD programs for image editing, scanning,
etc. Once you have a selection of cards, all equally capable of providing the
same resolutions and bit depth, then the desciding factor is it's
performance in other programs where specialized chips can make a difference.
Ang
Gregory –
Actual speed in 3DS is based on a number of items. First and foremost is
CPU. Nothing, I mean nothing, will make 3DS perform faster than $ spend on
the 486-66. I bet when the 586 starts shipping, the first customers will
be 3DS users.
Now beside rendering, all other interaction with 3DS is via either the
Materials Editor or the various display modules. There is basically
nothing that can speed up the materials editor much beyond an basic VGA and
a GOOD driver, since the materials editor is not video intensive. A well
designed driver will not slow you down at higher resolutions (beside longer
material rendering times<g>). To the best of my knowledge, all drivers
that implement high-resolution Material Editors actaully make the materials
editor HARDER to use, because everything gets visually smaller on the
screen. Of course, there is one driver that is an exception<g>.
In the display modules is where an accelerated board will help. The
supplied VESA driver with 3DS only works on 256 color VGA modes, and it's
performance implementation is very poor compared to what can be achieved
with the right code. If you are going to stay with a VGA, a 16 color mode
is the only way to go to get "snappy" performance. The key to accelerated
boards are that they provide higher resolution (1024 and up) without
noticable performance loss.
[More]
[Continued]
In addition, many of the boards support higher refresh rates than VGA's for
a more visually appealing screen.
One area that REALLY benefits from accelerated boards is in the keyframer
when moving from frame to frame. Accelerated boards will enable very
smooth motion, while VGA's will tend to be sluggish, especially the VESA
VGA driver!
What many users have discovered is that some external drivers are
dramatically slower in the display modules than the built in VGA. There
are two reasons for this. 1) 3DS is wraped around an interface that
REALLY slows it down, ADI. The best way to think of it is like taking a
Porsche (3DS) and putting in a Honda Civic engine (ADI) into it. What a
waste! 2) The main ADI calls that 3DS sends to an ADI driver to draw items
with is generally one of the SLOWEST adi calls! Most driver developers did
not pay attention to these calls and 3DS suffers greatly from it.
What Vibrant has just developed is a technique where much of the overhead
that ADI imposes on the system has been removed, thus improving user
interface performance from anywhere from 20% to 60%+. This, combined with
a good accelerated board, really make the 3DS display modules scream!
(Just ask Greg Pyros<g>).
When choosing a board make sure that it does support the VESA VGA bios,
because the internal built-int flic player in 3DS does not use any external
driver. Thus any board is NOT VESA VGA compatible (and even some that are)
will have a bunch of problems viewing flics while in 3DS. We are currently
working on solving that problem…keep your fingers crossed<g>. Jack Bayt –
Vibrant
Jack-
Thanks for your very complete response. The work you're doing at Vibrant
on 3DS sounds really promising. I'm looking forward to hearing more about
it.
Incidentally, did you happen to catch Steve Gibson's series of articles
in InfoWorld on the pros and cons of local bus? I'd be very interested to
hear what you think.
I uploaded a message about this on the ACAD forum but I think it bears
repeating over here so here it is:
I have just read a _fascinating_ series of articles in InfoWorld by Steve
Gibson. Steve Gibson, in case the name isn't familiar, is the creator and
developer of Spin-Rite, the president of Gibson Research Corp., and he writes
the weekly "Tech Talk" column in InfoWorld.
In a series of five articles from 12/21/92 to 1/25/93 Steve gets right
down in the mud and crunches the numbers. The conclusion he comes up with
is startling (and disappointing). His "bottom line" is the following:
"I have been unable to measure any performance gain from using a
state-of-the-art local bus adapter over the same technology adapter on the
ISA bus."
[More]
[Continued]
"In fact, manufacturers that make identical ISA and local bus adapters
have priced their local bus versions the same as their ISA products. Why?
Because they aren't any faster. The bottleneck is not in the bus — it's in
the CPU on one side and the graphics coprocessor on the other. Neither can
move the data faster then the ISA bus can move it"
If you know anyone with a subscription to InfoWorld, or if your library
has it, I suggest you check this out. Gibson is a master at presenting this
tech stuff in a clear, readable, and understandable manner and his logic is
irrefutable.
Greg LeVasseur
P.S. He _does_ mention one instance where local bus did show a performance
improvement and that is in "nonmainstream applications involving
nothing but colorful static image movement to the screen." That sounds
a lot like playing a flic file so maybe local bus is useful in this one
area.
Yes, I read the artical and I happen to agree with it. In general,
VL-BUS and EISA boards will not provide enhancement for 3DS over the
ISA versions. Now FLIC playing is a whole other ballgame. Anything
that can reduce the overhead of transfering large blocks of pixel
data to the video board will show increased performance in flic
playback (like the DELL DGX!) I haven't seen the local bus version of
the IIT AXGA chip yet, it may also benefit from local-bus on animator
as will the Weitek P9000.
Jack-
Yes, it's me again! A couple of times in your message you mention
accelerated boards. I assume you are not really talking about Windows
accelerators like the S3 chip et al. So what boards are you referring to?
Also, do you have anything yet in the way of drivers for 3DS that are better
than the VESA drivers? I'd be interested to know what you've come up with so
far.
Greg
Actaully when I'm refering to accelerated boards, I DO mean boards
based on S3, ATI MACH 8/32, WD 90C31, Weitek P9000, IIT AXGA,
34010/20, 8514, XGA, DELL DGX…. you get the idea. Any of these
technologies that provided some form of hardware graphics operations
will help performance in the display modules IF the driver implements
the extendeded capabilities.
The slowest card for display modules is a 256 color VGA, due to the
amount of data fired over the bus just to do things like viewport
clearing.