PDA

View Full Version : Nvidia GeForce FX vs. Quadro FX


adam|zeliasz
03-06-2003, 11:39 PM
Are you really going to see an incredible preformance increase when running a Quadro FX 2000 to a GeForce FX 5800 Ultra?

The GeForce 2's were based on the same board, and with a few jumper modifications you were able to harness the power of Nvidia's workstation card. Sure your going to be able to visualize with more lightsources, have nice antialiased lines, and maybe reduce the amount of visual "artifacts." But in the long run is it worth close to a $1000 difference?

What do you think?

elvis
03-06-2003, 11:58 PM
this has been covered to death. search through the forums for all discussions regarding "softquadro".

NVIDIA are fully aware of the driver hacking people are using to turn normal gaming cards into pro quadro cards. the geforce4 was the first card to introduce a different chip design between the geforce4 and quadro4 card that limited te effect of the softquadro features that could be enabled. in particular two major features for modellers: wireframe accelleration and dual-sided shading/lighting were removed form the gaming card in hardware. no amound of physical card hacking or driver hacking will enable these features on a geforce4, softquadro or otherwise.

the end story is this: if you are modelling very large scenes with high poly counts and large texture requiremnets, the $2000 is worth it when clients are knocking down your door for work. if you are a student, one-man business, or hobbiest, then it's probably not worth the cost.

i work for a large architecture and design office where every minute saved is dollars saved. spending an extra $2000 on a pro card could literally mean the difference between winning a $10 billion international design project competition and losing it.

for everyone out there reading this: if you can't see how a pro card is worth the extra cost, i can tell you straight up now you aren't even coming close to the full potential of your hardware. and quite possibly, the extra money spend on pro quadro and fireGL cards is a waste, as you can't ever harness their full potential. for anyone who's done any serious work, they'll know what i'm talking about.

on another topic: yes, i've promised to try and get some stickies up about questions like these that get asked every other week, but i've been flat out this week (major server upgrades here, as well as my birthday in there somewhere too - yay for overtime on your birthday). i promise as soon as time permits i'll organise some stickies and/or FAQs that will answer these reoccuring questions.

adam|zeliasz
03-07-2003, 12:02 AM
Right to the point. Thanks, this really helps me out.

loop29
03-10-2003, 05:43 PM
adam, for your proper information, it´s not hardware wireframe acceleration that misses on latest softquadroed boards (Geforce 4 Ti) , it´s HARDWARE ANTIALIASED LINES that´s ripped off the chip with 2 sided lightning. But still a valid point for not using it in professional work is that everyone will blame it on the softquadro if something fails on your PC. I´m working in an engineering office and we develop Laser weldment tools for the automotive Industry, 10 seats with Autodesk Inventor 6, 9 of them using Softquadro4 with Geforce 4 Ti and 1 seat with Quadro 4 XGL 750. I experienced nothing faster or better with the original Quadro 750 XGL compared to the softquadroed cards.
It depends on what applications you´re running, you can´t draw that conclusion that easy. At least you have to make the decision alone: nobody will give you support when using softquadro and you have problems with your graphics output. But from my experience Softquadro was allways a no brainer, it was always other software related problem that slowed me down in my work.

Here is a link from the coder of Softquadro about that wireframe acceleration statement from somebody who didn´t read the softqaudro 4 documentation properly.

http://www.guru3d.com/forum/showthread.php?s=&threadid=48639

regards

elvis
03-10-2003, 10:06 PM
Originally posted by loop29
I´m working in an engineering office and we develop Laser weldment tools for the automotive Industry, 10 seats with Autodesk Inventor 6, 9 of them using Softquadro4 with Geforce 4 Ti and 1 seat with Quadro 4 XGL 750. I experienced nothing faster or better with the original Quadro 750 XGL compared to the softquadroed cards.
It depends on what applications you´re running

loop29:

i'll graciously accept the facts above now that unwinder has confirmed it himself. earlier documentation i had read came from the guru3d guys that indicated the opposite. i certainly will take unwinder's word (or nawxtennis for that matter) over anyone elses.

of note: i work for an architectural firm that design sporting stadia. as you can imagine the size of the 2D and 3D drawings we create are massive. a relatively simple 3D model of a single stadium can easily hit 1 million polygons. add to it surrounding buildings, features, roads, etc and you can go up from there.

the biggest i've had the displeasure of seeing was a monstor 2.1 million poly scene that really hurt our systems. the problem with the nature of our work (as no doubt loop29 understands well in engineering) is that things must be modelled ACCURATELY. not doing so results on lots of problems. a guestimation of a few pixels in max can result in a building being 3 meters or more out of alignement, whcih is bad news.

anyways, back on topic: we tested a series of quadro cards and geforce cards on our existing dell dual processor workhorses to see how they performed. softquadro'ed geforce2 and geforce2mx were (as expected) pretty much on par with their real quadro bretheren.

we had the opportunity to test some geforce4ti and gf4mx cards also, but were greatly dissapointed in their performance. within specviewperf, specAPC and augi gauge benchmarks they performed fine, but hit them up with one of our large scenes and they choked. i don't know what was the cause, but this led me to believe again that the wireframe accelleration wasn't working due to the choppyness of the scenes compared to a real quadro4. at times even the softquadro2 cards outperformed them!

my guess now (after reading unwinder's post above) is there was some other factor that limited the softqadro4 cards when drawing much larger scenes in wireframe. again, or SPEC benchmarks and the like ran fine. it was only our VERY large scenes that killed it.

to cut a long story short: all of our guys now are decked out with dual xeon dell workstations running "real" quadro4 cards, and all is well. our translated ADT to Max/VIZ models are coming through no problems and wer are able to open and modeify them without too much hassle.

as to why the softquadro4 cards choked, i have no idea. perhaps a chat to unwinder might shed some light on the issue. the nature of the work we do here is very unique. there are only another 2-3 firms worldwide who do what we do. our CAD packages even have troubles handling the large scenes. we recently were given trial versions of REVIT and archiCAD, both of which couldn't even handle our drawings with the detail we need without crashing!

CgFX
03-11-2003, 06:30 AM
1:
Support and stability are inportant to many people and worth some money. That money difference is not very large for a card like the Quadro4 750 XGL and is worth being able to call A|W and get support, for example.

2:
Autodesk Inventor is a really really bad example as that application is not the best in 3D performance nor the design of the application (massive understatement). I am not sure you would see much performance difference running it on a 1990 laptop. :-)

3:
It is a bit premature to be talking about softquadro and GeForce FX/Quadro FX, don't you think? You might wait until they have at least commented on attempts to do it, let alone any success. I can only assume that, like Quadro4, there are real hardware differences between NV30 and NV30gl and I can only assume that has increased this time around. We have seen that Softquadro4 does not create a 100% Quadro4 this time. Something closer to 50%-75%. I would bet this time it would be even more difficult.

elvis
03-11-2003, 06:48 AM
1:
couldn't agree more

2:
couldn't agree more

3:
couldn't agree more

:applause:

derelict
03-11-2003, 08:18 AM
i will sum it up by saying one is cheap and the other is expansive. both actually does the same thing only that on some apps gf does it better.

loop29
03-11-2003, 12:11 PM
Now we´re talking business ;)

I understand your problems and agree that with no doubt that this is the strongest argument to go for the real deal. If it doesn´t fit your needs go for the original one.

@CgFX

1. I agree, it is worth invest the extra bucks for getting valuable support if you need it. On the other hand AW would give you support, no matter if you´re running GF or Quadro. How can one say that the problem is related to grafics card before solving it? If you say you´re using professional video adapter (quadro) they will give you support, I guarantee.

2. Yes, that seems to be the case in terms of grafics performance that there was not too much of a difference when comparing different generations of graphics hardware (especially between Quadro 2 Pro and Quadro 4 XGL 750 was no big margin), apllies to R6 the most I would say.
What do you mean with bad design of the application ?

3. Yes, I agree that nvidia will have something better on GFFX for protection of NVxxGL capabilities, but maybe you got more valid information on this than me.

regards

Pablo3d
03-12-2003, 06:16 AM
I've used softquadro on a evga gf4 ti4200 and win 2000 sp2 recognizes it as a quadro4 XGL 750. I don't really notice any difference using 3ds max 4.25. I downloaded and ran SPEC viewperf and the 3dsmax mean was 7.5 which is pretty low. I didn't run this benchmark prior to using softquadro so no comparison. Top end systems score above 19. My CPU is a 1.67GHz AMD XP and I have 512 MB 333MHz RAM.

I've been working with lots of contour lines imported from dwg files and have to normalize splines. I have ton's of vertices - maybe 400,000 - and the computer does a decent job of handling all that detail. One time I normalized a big group of splines and it took the computer 40 minutes to do it. My old computer (win98, Duron 900, gf2mx, 256MB ram) simple could not deal with so many splines and verts in a timely matter, i.e. CHOKE!

Do you guys ever look at xbitlabs.com benchmarks? The quadro FX looks like it burns rubber big time:airguitar

So in my humble opinion you can get softquadro to work with gf4 ti4200's but your not gonna end up with any major increase in performance.

CGTalk Moderation
01-14-2006, 02:00 PM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.