Smooth polygons speed - GeForce weakness ?

I try using :
glEnable(GL_POLYGON_SMOOTH);
glHint(GL_POLYGON_SMOOTH_HINT,GL_NICEST);
to have some nice antialiased polygons…
Bad surprise !
On the ATI RadeON, there is no speed decrease using this feature.
But the NVidia GeForce 256 becomes as slow as software rendering !
Both cards are tested with the latest official Win2K drivers.
I was wondering if this is a GeForce 1 weakness and if the GeForce 2 corrects this … can someone test it ?

Well, I believe that in the field that is target for the GeForce2 (games), polygon smooth is hardly used, so we can say it supports it, but through non-optimized paths. Does the ATi display polygons correctly smoothed, or it simply ignores smoothing?

The ATI display the polygons perfectly smoothed, the features is really working … on the GeForce too, but much much slower !
The reason i want to use this feature is that is looks much better than FSAA (this is REAL subpixel drawing !)
By the way, same thing happens with antialiased lines, they don’t slow down the ATI but the GeForce dies on them …

[This message has been edited by paddy (edited 01-11-2001).]

Well, I have had weird performance in my geforce2 card when in wireframe (even without AA, about 1/10 the speed of a 100k model rendered solid), when all other cards I have tested perform faster in wireframe than in solid.
My conclusion is that, as a gaming card, the GeForce2 is not optimized for some scenarios that are not widelly used in games. I agree polygon smooth can give your far better quality, but you need to sort polys, so you end with a static scene (BSP ans stuff) or a slow scene (per frame sorting).
I would like to have an ATi radeon, it looks like a very good GeForce2 competitor (maybe a tad slower in some situations, but with an impresive feature set). How stable/bugless do you find the radeon drivers?

I’m quite happy with my RadeON.
I use the Win2k drivers version 3073.
What first impressed me on the RadeON is the 2D image quality. This is not a legend ! Crisp and clear image, fanstastic colors, you will think you have a brand new monitor ! Only matrox may compete in this domain.
On the OpenGL side, drivers are fast and stable, both in pro apps (LightWave, Maya) and in games.
And (that’s my fun) … 3D textures are really cool
I’m still waiting for some DX8 compliant drivers, but it seems the GeForce DX8 drivers are kinda beta too (7.xx series). The actual drivers have partial DX8 support anyway, even if all the features are not enabled yet.
Beside this, the video playback capabilities are impressive (mpeg, avi and dvd).

Great! Well, I’m happy with my GeForce2 also (a bit unhappier since I found this wireframe issue, but well…). 3D textures are king.
No current 3D card does DX8 pixel shaders, but I believe nvidia does vertex shaders (posible in sw inside the driver), even in opengl (beta 7 drivers).

> Well, I’m happy with my GeForce2 also (a bit unhappier since I found this wireframe issue, but well…).

Well, if you really need fast wireframe - you may find it interesting : GeForce/GeForce 2 -> Quadro/Quadro2

When I first inserted my GeForce DDR in my computer, I found out that all my OpenGL programs were running at the speed of a snail…

I soon discovered that the problem was that I enabled GL_POLYGON_SMOOTH.

After contacting nVidia, they confirmed that this feature should be avoided (not that it is buggy, it is simply slow !). They also told me that the other video cards (e.g. TNT) did not suffer from this problem because they just ignored polygon smoothing !

I am very surprised that the Radeon does not show any performance loss when using polygon smoothing. Are you sure it does what it is supposed to ??? Can you post screenshots to compare the two cards ???

I am pretty sure some people from nVidia won’t miss this one!

Regards.

Eric

Originally posted by Serge K:
>[i] Well, if you really need fast wireframe - you may find it interesting : GeForce/GeForce 2 -> Quadro/Quadro2

Well, I am sure they won’t miss this one either !

Regards.

Eric

P.S.: as far as I know, this modification works… but I would definitely not try it on a £200 card…

About the GeForce/GeForce 2 -> Quadro/Quadro2 stuff, yes, I know, but I dont really want to doit because:
1: Its scary to take your $300 card and start soldering/unsoldering stuff.
2: I want my card for gamming, and I’m sure that quadro drivers might have issues with some games (maybe not, who knows).
It’s a bit disapointing to see that both cards being so similar have so different prices.
So how does the radeon do in 3d apps? does it really shine in wireframe/high poly count? then I would buy a radeon instead of a quadro or geforce for 3D modeling and content development.

nVidia has a FAQ on their developer page that points out that using GL_POLYGON_SMOOTH slows it down a lot. It also helps a lot for finding out why your application that is supposed to do about a million triangles per second with full detail is only giving you about 50,000 per second.

j

coco:
It’s a bit disapointing to see that both cards being so similar have so different prices.

Nothing new.
Quadro1/2 are kind of cash cow for NVIDIA.
Same for M$ WinNT/Win2000 server - if you want it for >2 CPU - you have to pay a lot for the same OS just compiled with couple of different #define-s.
Same for Intel: Xeon always was way to costly.
Now situation even funnier : the only difference between new P3 Xeon with 256K cache and regular P3 is the slot/package type.
The same chip for a slot2 is cost - guess what? - twice more. But if you have this hi-end brandname workstation - you’re bound to buy Xeon for upgrade.
It’s called Market…

Ok, so the difference between a quadro2 and a geforce2 is only the way Id resistors are soldered and the clock speed? is there really something inside the chip that make it twice as expensive as the geforce2?

You seem to be under the mistaken assumption
that the cost of goods (silicon, metal,
plastics) or even development (hardware and
drivers) has a direct impact on the price
charged for a product.

A company will charge enough so that it
maximizes the formula:

Profit = (Price - Cost) * UnitsSold

Typically, they’ll sell more if the price is
lower, but they’ll lose margin. However, if
there is a way that they can sell “some”
units at a higher price, and then “more”
units at a lower price, they will.

Sure, there are real differences between
Quadro and Geforce cards, but if it’s cheaper
to produce one chip, and the take away some
features for the cheaper card, than to
produce two different chips, then that’s what
the company will do.

In retailing, this is why there are coupons
and sales; people who are stingy wait until
the sale and clip coupons at home; people who
don’t think the savings are worth the bother
pay more for, essentially, the same products.

bgl:
I know that the cost of the chips is not directly proportional to the price of the cards that use them. Let me redo my question:
Is the additional cost of a quadro2 over the geforce2 justified, or is just that the same product can have different names and prices depending on the target market only? Should this be legal? I’m not an expert in laws or economics, but my common sense tells me no to both questions.
The only justification for the quadro2 to be less cheap than a geforce2 ultra is driver developent and an increase or 10% or so in clock speed (yet it costs 2x the price of a geforce2 ultra). As Serge K pointed out, this video cards are only an example of what some companies are doing too.
I dont want to get my hands on fire with this, but I think users are the only affected with such practices. Selling a product with some features locked doesn’t cost less to produce than selling the product unlocked in the first place, as simplistic as it sounds. Its cheating the consumers.

[This message has been edited by coco (edited 01-13-2001).]

++++++++++++++++++++++++
They also told me that the other video cards (e.g. TNT) did not suffer from this problem because they just ignored polygon smoothing !
++++++++++++++++++++++++

I don’t believe that TNT ignored polygon.I have tried it,and when GL_POLYGON_SMOOTH is on the display is distinctly perfect In quality(not quantity !).slow…slow…very slow…So we’d better void use it in realtime display app.

but in order to get a piece of static image,it is the sharp point!

Originally posted by coco:
bgl:
I know that the cost of the chips is not directly proportional to the price of the cards that use them. Let me redo my question:
Is the additional cost of a quadro2 over the geforce2 justified, or is just that the same product can have different names and prices depending on the target market only? Should this be legal? I’m not an expert in laws or economics, but my common sense tells me no to both questions.
]

They can set any price they want, what’s interesting is what people are prepared to pay. Companies doing CGI and CAD/CAM are prepared to pay more (the Quadro is dirt cheap compared to other professional cards) for these features so Nvidia charge more.

From wich professional cards are you talking? I didn’t see any 3d card impressing the Geforce 2 Quadro like dirt.

Originally posted by Michael Steinberg:
From wich professional cards are you talking? I didn’t see any 3d card impressing the Geforce 2 Quadro like dirt.

all pices courtesy of PC Magazine Aug 7, 2000

Elsa Gloria 2 64MB (ie: Geforce 2 Quadro): $800
3Dlabs Oxygen GVX210 64MB: $1300
3Dlabs Wildcat 4110 64MB: $2000

Thats pretty dirt cheap in my book. And since that time, I think the Quadros have dropped in price a lot more than the other professional cards have (but dont quote me on that)

Originally posted by Suvcon:
I don’t believe that TNT ignored polygon.I have tried it,and when GL_POLYGON_SMOOTH is on the display is distinctly perfect In quality(not quantity !).slow…slow…very slow…

My memory is not as good as it was ! They were probably talking about other chips then !

Sorry for that…

Regards.

Eric

P.S.: as far as I remember, my Diamond Viper Ultra 770 (TNT2 Ultra) did not show any difference with or without polygon smoothing (neither in speed nor in quality!)… Perhaps that was a driver issue…