Quadro FX Vs GeForce FX 5900 Ultra

[california0071]

Can anyone tell me what the advantages are to having a Quadro FX over the new GeForce FX 5900 Ultra???

I am interested in buying a new card and am keen to get the card to suit my needs. I am interested in pursuing my interest in 3d Cg computer animation and so am keen to get the best card for the job. Naturally, I also don’t mind a bit of gaming now and then.

Is the GeForce FX 5900 a more powerful card given it is 256 bit, twice that of a Quadro FX? Does this make the GeForce FX 5900 a more suitable card for 3d animation with 256-bit technology?

Is the Quadro FX worth it given the GeForce FX 5900?

Thanks in advance
California0071

[fantomas]

quadro's have some serious advantage's, such as hardware anti-aliasing, also they have specialized drivers for all the 3d apps. But you will be fine with a gaming card for most of what you want to do. :)

[Behemoth]

if you are using 3D to pay your bills, if every minute saved is dollars saved, then quadro FX is worth its price. if you are student or 3D hobbiest, fx5900 or other nvidia gaming cards(dont try other brands :p) is enough.

here is a list of a professional card/gaming card difference taken from a PNY presentation:

The customers expectations
The term professional workstation usually implies high driver and hardware quality, excellent reliability, responsive support, and high performance.
The expectations of workstation users exist because they have a different point of view when they’re looking at their computers. It’s a tool, not a toy, they have to work with it.
Their goal is to design, create scenes or demonstrate their work. Time is money, so they don’t accept crashes, baffling bugs or avoidable delays in workflow.
They don’t change their adapter twice a year, so they expect driver support for at least 2 years.
Because in most companies the purchased hardware has to be qualified, it should be available for a long period of time, no matter if it’s up to date


Public Opinion
People often ask for the differences between Quadro and GeForce adapters, because their technique appears to be very similar and their price is differntiating.
Even in CAD-relevant newsgroups there are many estimations of users telling that GeForce adapters work fine in 3dsmax, Inventor, AutoCAD and so on.
What people mostly forget, is that especially this similarity dropped the prices for professional adapters. NVidia’s concept worked for both, the quality and versatility of consumer adapters and the pricing for professional products.
But there are lots of differences, and NVidia spends ambitious efforts to create the perfect solution for both parties.


1.1 Hardware antialiased lines
A unique feature of the Quadro GPUs is to support antialiased lines in hardware. This has nothing in common with the GeForce full-scene antialiasing.
It works for lines, not for shaded polygons, without sacrificing performance or taking extra video memory for oversampling. Most professional applications support this feature because it is standardized by OpenGL.


1.2 Logical Operations
Another unique feature of the Quadro GPUs is supporting OpenGL Logical Operations. Those can be implemented as a last step in the rendering pipeline, before contents is written to the framebuffer. Workstation applications use this functionality to draw on top of a 3D scene, for example to mark a selection by a simple XOR function.

Having this function done in hardware prevents from significant performance losses like a GeForce Adapter would show.


2. The OpenGL Differences
On consumer and workstation adapters, OpenGL is used for different purposes:
Most common applications for GeForce adapters are full-screen OpenGL games.
CAD applications are working with OpenGL windows, in combination with 2D-elements
The unified driver architecture provides an optimized implementation for both, professional and consumer demands. While consumer applications have a quite simple request, bug-free functionality and performance over all, there are several optimizations for professional applications that would not improve anything in full-screen OpenGL. They only make sense for window-based OpenGL and are therefore working on Quadro-based hardware only.
They will be explained next?


2.1 Clip Regions
A typical workstation application contains 3D and 2D elements. While viewports display window-based OpenGL, menus, rollups and frames are still 2D elements. They oftenly overlap. Depending on how they are handled by the graphics hardware, overlapping windows may noticeably affect visual quality and graphics performance.
When a window has no overlapping windows, the entire contents of the color buffer can be transferred to the frame buffer in a single, continuous rectangular region. However, if other windows overlap the window, the transfer of data from the color buffer to the frame buffer must be broken into a series of smaller, discontinuous rectangular regions. These rectangular regions are referred to as “clip regions.

GeForce Hardware supports just one clip region, mostly sufficient for displaying menus in OpenGL. Quadro GPUs support up to 8 clip regions in hardware, keeping up the performance in normal workflow using CAD/DCC applications.


2.2 Hardware-Accelerated Clip Planes
Clip planes allow sections of 3D-objects to be cut away so the user can look inside solid objects. Looking inside objects is particularly useful for visualizing assemblies. For this reason, many professional CAD /DCC applications provide clip planes.
The Quadro family of GPUs supports clip-plane acceleration in hardware—a significant performance improvement when they are used in professional applications. Tests 6 and 10 of the SPECopc Viewperf MedMCAD-01 Test define a clip plane, and are useful for quantifying the performance benefits of clip-plane support on the Quadro2 family.


2.3 Quadro Memory Management optimization
Another feature offered by the Quadro family of GPUs is Quadro memory management optimization, which efficiently allocates and shares memory resources between concurrent graphics windows and applications. In many situations, this feature directly affects application performance and so offers demonstrable benefits over the consumer-oriented GeForce GPU family.
The graphics memory is used for the frame buffer, textures, caching and data. NVIDIA’s unified memory architecture dynamically allocates the memory resources instead of keeping a fixed size for the frame buffer. Instead of the remaining frame buffer memory being wasted because it is unused, UMA allows it to be used for other buffers and textures.
Especially when applications require much memory, using quad-buffered stereo or full scene antialiasing, it becomes more important to manage the resources efficient. The table below contains different constellations on GeForce and Quadro hardware showing the advantage of this effectiveness, both have 32MB on board


2.4 Common problems solved with Quadro drivers:
Two-sided Lightning
Quadro hardware supports two-sided lighting. Objects which are not created as solids may show triangles from their “back-side? viewing the objects from the inside.
Two sided lightning prevents the diffuse and specular lightning components from dropping to zero when the surface normal points away from the light source. As a result, these “backward-facing?triangles remain visible through all viewing angles.


2.5 Hardware Overlay Planes
The user interfaces of many professional applications often requires elements to be interactively drawn on top of a 3D model or scene. The cursor, pop-up menus or dialogs appear on top of a 3D-viewport. These elements can damage the contents of the covered windows or affect their performance and interactivity.
To avoid this, most professional applications use overlay planes. Overlay planes let items be drawn on top of the main graphics window without damaging the contents of the windows beneath. Windows drawn in the overlay plane can contain text, graphics, and so on—the same as any normal window.
The planes support a transparency bit, which when set, allows pixels underneath the overlayed window to show through. They are created as two separate layers, nothing has to be blended together. This prevents damage to the main graphics window and improves performance. Likewise, clearing an overlayed window to the transparency bit and then drawing graphics within it allows user-interface items to be drawn over the main graphics window.
Clearing and redrawing only the overlayed window is significantly faster than redrawing the main graphics window. This is how animated user-interface components can be drawn over 3D models or scenes.


2.6 Quad buffered stereo
The Quadro GPU family supports quad-buffered stereo; the GeForce GPU family does not. Quad-buffered stereo is an OpenGL functionality, not depending on special stereo hardware to show the effect. Two pictures are generated, both double-buffered, one per eye. Displaying is done alternately or interlaced, depending on the output device.
Many professional applications like 3ds max, SolidWorks or StudioTools let users view models or scenes in three dimensions, using a stereoscopic display. It can be done by a plugin, like in Solidworks, by a application driver like MAXtreme in 3ds max, an external viewer like QuadroView for autocad-based products or by the application itself. The use of stereoscopic display is to have an overview in complex wireframe constructions, making walkthroughs much more realistic and impressive or simply to improve the displaying of proportions in large 3D-scenes.
Stereo support on the Quadro GPU family significantly benefits professional applications that demand stereo viewing capabilities.

[bkswaney]

Why no hardware AA support on the gaming cards?
It seems it would really bring up the performance. :)

[Behemoth]

Originally posted by bkswaney
Why no hardware AA support on the gaming cards?
It seems it would really bring up the performance. :)
coz games dont use it i guess :D
by the way, its AA lines, it only smooths the lines.

[bkswaney]

Originally posted by Behemoth
coz games dont use it i guess :D
by the way, its AA lines, it only smooths the lines.

I'm not talking about the Quadro.

Why not use HW AA for gaming? Or do they?

[Behemoth]

Originally posted by bkswaney
I'm not talking about the Quadro.

Why not use HW AA for gaming? Or do they?
oh, i think the FSAA is already hardware-ed, otherwise FX AA should perform as bad as geforce4?

[Lezmaka]

Originally posted by bkswaney
I'm not talking about the Quadro.

Why not use HW AA for gaming? Or do they?

1.1 Hardware antialiased lines
A unique feature of the Quadro GPUs is to support antialiased lines in hardware. This has nothing in common with the GeForce full-scene antialiasing.
It works for lines, not for shaded polygons, without sacrificing performance or taking extra video memory for oversampling.


Basically, it'd only be worth it if you ran games in wireframe mode...

[greggman]

I'm curious if this is still important

> 1.1 Hardware antialiased lines

Do I really need these? When I render it's not going to use the hardware is it therefore Hardware antialiased lines are hardly a benefit

> 1.2 Logical Operations

I didn't know older cards didn't support these. Maybe I'm mistaking them for something else. I though all consumer cards supported the various logical operations since like Voodoo days. Anyway, even if I was wrong about the past now with pixel shaders all operations are supported on consumer hardware.

> 2.2 Hardware-Accelerated Clip Planes

Seems like these could easily be handled in shaders as well

> 2.1 Clip Regions
> 2.5 Hardware Overlay Planes

OS X on the Mac has shown that putting all your windows in textures simulates these features as well. I suspect the difference in speed would not be noticable. All they would have to do for consumer cards is provide the same solution.

> 2.4 Common problems solved with Quadro drivers:

I assume this is a driver issue, not a hardware issue. Shame on them for making you pay double the money for 1 line of code. If it was hardware then it's now possible to fix with shaders

> 2.6 Quad buffered stereo

I don't know any pros even using this feature.

Isn't it about time the facade be dropped or am I still missing something?

[mezkal]

I'm curious if this is still important

> 1.1 Hardware antialiased lines

Do I really need these? When I render it's not going to use the hardware is it therefore Hardware antialiased lines are hardly a benefit

> 1.2 Logical Operations

I didn't know older cards didn't support these. Maybe I'm mistaking them for something else. I though all consumer cards supported the various logical operations since like Voodoo days. Anyway, even if I was wrong about the past now with pixel shaders all operations are supported on consumer hardware.

> 2.2 Hardware-Accelerated Clip Planes

Seems like these could easily be handled in shaders as well

> 2.1 Clip Regions
> 2.5 Hardware Overlay Planes

OS X on the Mac has shown that putting all your windows in textures simulates these features as well. I suspect the difference in speed would not be noticable. All they would have to do for consumer cards is provide the same solution.

> 2.4 Common problems solved with Quadro drivers:

I assume this is a driver issue, not a hardware issue. Shame on them for making you pay double the money for 1 line of code. If it was hardware then it's now possible to fix with shaders

> 2.6 Quad buffered stereo

I don't know any pros even using this feature.

Isn't it about time the facade be dropped or am I still missing something?

Kids.

Hardware Anti-Aliased Lines : Well considering you're going to working in wireframe the vast majority of your time (unless you're a texture artist), the ability of the card to able top supply you with smooth lines is a boon. Imagine how much more your eyes will like you (and consequentially help you be more productive) if they don't have to squint to make out the flow of the normals on a plane that you can't quite remember was convex or concave when you got up to make a coffee.

Logical Operations : You'll note that technology, especially the 3D graphics field, is ever developing. The number of logical operations, say simulating a procedural texture over a mesh, expands with each GPU generation. You need this extension. Just because a 3D/fx had some texture management (ie Logical Operation) back in the day, don't assume that what was useful then (hardware calced phong shading and 1 level mipmaps) is OK now.

Hardware-Accelerated Clip Planes : Your comment that the same end could be achieved with shaders is laughable. Shaders are predefined procedural maps with an ability to modify the alpha between planes and they way textures interact. How, pray tell, would that help you clip a plane away from a model so you can look inside it? We're talking model management here, not a game. Don't think like a game player. Think like a 3D artist.

Hardware Overlay Planes: Stupid arguement and with poor innocent littel OSX thrown in as a red herring in the bargain too. Woo.
The Nvidia driver already does hardware overlay planes for consumer gear. You'll notice it when you select "window transparency when moved" in Nview. But for consumer cards do this only for that function. Now consider that as a function of your fave 3D app and further consider if your 3D app could have control of these planes and use them for cache, procedural mapping and other cool things. Once again, stop thinking like a gamer.

Quad Buffered Stereo : You obviously know no professional 3D artists, medical imaging companies, metallurgical researchers etc. I do. I've seen this technology in use. A lot. Visualisation tools evolve too you know?

Look I realise that you're a kid and that you think Nvidia owe something too you. You think that thier millions of R & D dollars should be spent in bringing you a consumer priced card that will give professional performance.

IT. AINT. GONNA. HAPPEN.

You reckon a shlub like you keeps Nvidia in business? So then shut up. I'm sick to my back teeth of hearing penny pinching dickheads and gormless students whine that the Quadros offer nothing over thier consumer versions. They do. What's more they offer said features to those who PAY. That's how business works. Noones deceiving you. Infact, when a recent bug in a Quadro driver set allowed the Geforce 4 -> Quadro Patch script in Rivatuner to work, thousands of people literally assaulted poor Unwinder with a myriad of questions most of which sounded like "OMFG How can I get my FX card to be Quadro OMFG OMFG OMFG". Now even when the features were unlocked only SOME worked. Wanna know why? Coz the Quadro and FX cards ARE DIFFERENT.

So there. I'm done.

[greggman]

You reckon a shlub like you keeps Nvidia in business? So then shut up. I'm sick to my back teeth of hearing penny pinching dickheads and gormless students whine ...

Hmmm, people that resort to personal attacks generally don't have a clue and try to hide it by being assholes. Grow up.

Hardware Anti-Aliased Lines:

I don't need them. I've spent the last 21 years of my life developing video games full time. I don't wear glasses or contacts and my eyes don't hurt. I don't have a problem figuring out which way a line points because it's not anti-aliased. My point is this single feature hardly seems worth the extra cost

Logical Operations:

I don't know what the point of your reply was. My point was that all current cards that support at least pixel shader 2.0 will handle any operation logical operation that GL supports. Remember GL is like a 15 yr old standard and the features that high end 3D software uses are still written for that 15 yr old standard

Hardware-Accelerated Clip Planes :

Are easily achieved with shaders. Transform the clipping plane into screen space, check each pixel in a pixel shader to see if it's on one side or the other side of the plane. If on the clipped side, don't draw it. There are examples all over the net. Here's one from Microsoft

http://www.mvps.org/directx/articles/clipcube.htm

Hardware Overlay Planes:

My point for bringing up OS-X is it's proof that current cards can simulate hardware overlay planes through creative programming. I'm not trying to make any OS-X vs Windows comparison. Hardware Overlay Planes could make programs like Maya much more responsive since pulling up the marking menus or other menus would not require redrawing the 3D view when the menus are removed. Having to redraw those views, especially for a complex scene could be slow. But, the same feature of not having to draw the views can be solved in other ways. One using the method OS-X uses to draw its windows, namely drawing each window to a texture. While it wouldn't be as fast as Hardware Overlay Planes I doubt the difference would be noticable and it would certainly be faster than redrawing.

Quad Buffered Stereo :

> You obviously know no professional 3D artists

Hmmm, well, let's see. I've been programming games professionally for 21 years now and am currently working at Sony Japan on PS3 under the ICO producer. I have a feeling I know quite a few artists. None of them use Quad Buffered Stereo. My point is not that that feature is useless, my point is very few people use it. Most 3D artists are in games or movies, neither of those industries need this feature. ** If ** it's the only truely special feature of these cards then at least in these industries a GL card is not needed.

As an example. An ATI X800 Pro is about $500. An ATI FireGL V7100 is about $1000, twice the price. The specs are the same according to ATI's own site:

http://www.ati.com/products/fireglvseries/ATI_FireGL_Family_Brochure.pdf
http://apps.ati.com/ATIcompare/

Both have 256 meg of memory.
Both have an 8 Gigapixel/sec fillrate
Both appear to draw about the same number of polys per sec

What I'd really like to see is an actual comparision of consumer cards AND FireGL cards. So far I have not been able to find a single one. If GL cards were noticibly faster fine, but so far I've seen no actual benchmarks to show this.

[mezkal]

Hmmm, people that resort to personal attacks generally don't have a clue and try to hide it by being assholes. Grow up.

Hardware Anti-Aliased Lines:

I don't need them. I've spent the last 21 years of my life developing video games full time. I don't wear glasses or contacts and my eyes don't hurt. I don't have a problem figuring out which way a line points because it's not anti-aliased. My point is this single feature hardly seems worth the extra cost

Logical Operations:

I don't know what the point of your reply was. My point was that all current cards that support at least pixel shader 2.0 will handle any operation logical operation that GL supports. Remember GL is like a 15 yr old standard and the features that high end 3D software uses are still written for that 15 yr old standard

Hardware-Accelerated Clip Planes :

Are easily achieved with shaders. Transform the clipping plane into screen space, check each pixel in a pixel shader to see if it's on one side or the other side of the plane. If on the clipped side, don't draw it. There are examples all over the net. Here's one from Microsoft

http://www.mvps.org/directx/articles/clipcube.htm

Hardware Overlay Planes:

My point for bringing up OS-X is it's proof that current cards can simulate hardware overlay planes through creative programming. I'm not trying to make any OS-X vs Windows comparison. Hardware Overlay Planes could make programs like Maya much more responsive since pulling up the marking menus or other menus would not require redrawing the 3D view when the menus are removed. Having to redraw those views, especially for a complex scene could be slow. But, the same feature of not having to draw the views can be solved in other ways. One using the method OS-X uses to draw its windows, namely drawing each window to a texture. While it wouldn't be as fast as Hardware Overlay Planes I doubt the difference would be noticable and it would certainly be faster than redrawing.

Quad Buffered Stereo :

> You obviously know no professional 3D artists

Hmmm, well, let's see. I've been programming games professionally for 21 years now and am currently working at Sony Japan on PS3 under the ICO producer. I have a feeling I know quite a few artists. None of them use Quad Buffered Stereo. My point is not that that feature is useless, my point is very few people use it. Most 3D artists are in games or movies, neither of those industries need this feature. ** If ** it's the only truely special feature of these cards then at least in these industries a GL card is not needed.

As an example. An ATI X800 Pro is about $500. An ATI FireGL V7100 is about $1000, twice the price. The specs are the same according to ATI's own site:

http://www.ati.com/products/fireglvseries/ATI_FireGL_Family_Brochure.pdf
http://apps.ati.com/ATIcompare/

Both have 256 meg of memory.
Both have an 8 Gigapixel/sec fillrate
Both appear to draw about the same number of polys per sec

What I'd really like to see is an actual comparision of consumer cards AND FireGL cards. So far I have not been able to find a single one. If GL cards were noticibly faster fine, but so far I've seen no actual benchmarks to show this.


Your points are well taken. You're obviously not a kid. However if you are a professional who has worked in the games industry for as long you say you have then I am surprised that:-

a. You ask such a question. The Quadro/Consumer differences are well noted and have been much the same since Quadro came to the market. Not to mention the fact that a great many houses see Quadro's as being a little less than capable when compared with Wildcats and thier ilk

b. You don't understand the basic principle of market separation. Nvidia's R & D has to be absorbed somewhere. Their higher end product suits this need. Furthermore if you're a professional earning a professional's salary from using graphics hardware how can you even look at a consumer card.

c. That you engage in a debate with someone online.

Now I understand that I mistook you for a kid. Granted I got that wrong.

Now don't let me mistake you for a professional.

[greggman]

>a. You ask such a question. The Quadro/Consumer differences are well
> noted and have been much the same since Quadro came to the market.
> Not to mention the fact that a great many houses see Quadro's as being
> a little less than capable when compared with Wildcats and thier ilk

They are NOT well noted which is why I posted these questions. If you have some links that not only state the differences but back them up with benchmarks that compare them to consumer cards please point me too them. I couldn't find any.

> b. You don't understand the basic principle of market separation.
> Nvidia's R & D has to be absorbed somewhere. Their higher end
> product suits this need. Furthermore if you're a professional earning
> a professional's salary from using graphics hardware how can you
> even look at a consumer card.

That's an excuse SGI used to make, look where they are now. To run Alias Power Animator used to cost you $30,000 for the software and another $30,000 for the SGI to run it on. I know, my company bought 3 seats in 1995. Today's PCs run rings around those. The software is down to $1500 as is the PC. When PCs first matched SGIs for 3D performance the SGI snobs would say stuff like if you're a professional earning a professional's salary from 3D graphics hardware how can you even look at a consumer PC. Now-a-days if you walk into Digial Domain or PDI or any of the other effects houses you'll see rows of high end consumer PCs (Dells) not rows of SGIs.

Having been in the industry and having watched that process happen AND reading the specs of consumer cards vs FireGL card AND finding no actual evidence proving GL cards are still faster I raised the question. The doesn't mean I expect the answer to be the consumer cards match GL cards in performance but I want more than words, I want proof.

> c. That you engage in a debate with someone online.

I posted in this forum because Behemoth above had one of the few intellegent replies I've seen on the web. That lead me to hope that maybe someone here had more info.

[ragejg]

Chill out guys, or kids, or whatever. Jeebus.

[saturnotaku]

I posted in this forum because Behemoth above had one of the few intellegent replies I've seen on the web. That lead me to hope that maybe someone here had more info.

Did you happen to look at the date of the last post before your first one in this thread?

[greggman]

yes, i did look at the date of the post. If he was subscribed to this thread he would have gotten an email about it.

[Knot3D]

Greggman,
You work for Sony, in the Ico team !? Respect ! Ico's my fav game ever ! :ORDER:

And I totally agree with you on the Quadro vs Geforce stance. Certainly
for guys like me and the original thread starter this applies. I HAD a QuadroFX1000 and it was PERFECT for my skilllevel. However, a friend
of mine desperately wanted to buy it from me. The replacement card ?

A Gainward 6800GT Golden Sample card 256mb with dual DVI plus an additional 1 Gb of DDR500 ram which I could afford thanx to NOT buy a
card like the QuadroFX3000 (1500 Euro's in Europe).

I model and animate in Cinema4D and Maya, for my work and freelance work and the GT does all I need.

[novatech]

The workstation PC in the GFX department, went screwy early this year and the loan PC shipped with an FX5600 (geforce),The quality was far better than any of the team expected (especially in low poly animation), but ultimately the reduced detail and sharpness in the viewports made us demand our Quadro based system back in an extra hurry.

Bottom line is, if your card output is making you money, you don't want to compromise detail/clarity for a few hundred dollars saving.

if you need to play games, build a seprate PC for that :) or buy an xbox. (mag)

[Ssexyman8404]

Whoa, intense and informative debate. As a gaming and hardware fanatic, I would also like to see some charts and graphs comparing the two. Maybe someone with LOTS of money would actually use they're quadro/fireGL for games lol. Or maybe this thread is long dead.....

[Sworkhard]

This comparison has to do with outdated cards (Gforce 4 era) but have a look.
http://www.aceshardware.com/read.jsp?id=45000354