rec.autos.simulators

I ordered my current computer with a Voodoo 3000D (Dell OEM) since I knew it
would work better with GPL than the nVidia cards available at the time. But
I've always wondered what I'm giving up in Direct 3D performance in other
games such as F1 2000. Well, I just got the chance to borrow a brand new
Creative Labs Geforce2 GTS to see how it compared.

My system: Dell P3 550 (100Mhz FSB), 128MB, Win98, DirectX 7a.
Both video cards with the latest drivers I could get, not overclocked.
Frame rates measured with Fraps 1.2

I tested with F1 2000 at 1024x768, with graphics set at 'high', not 'max',
and all special features like sparks etc turned on. I used the Nurburgring
track since it has a lot of detail near the start/finish. Checked the frame
rate both at the start/finish v.s. most other less detailed areas, and with
2-3 opponent cars visible v.s running by myself.

Voodoo3 with other cars in start/finish straight: 18-21 fps
Voodoo3 with other cars in other areas: 21-23 fps
Voodoo3 by myself in start/finish straight: 21-24 fps
Voodoo3 by myself in other areas: 27-34 fps

Geforce2 GTS with other cars in start/finish straight: 16-18 fps
Geforce2 GTS with other cars in other areas: 21-24 fps
Geforce2 GTS by myself in start/finish straight: 21-23 fps
Geforce2 GTS by myself in other areas: 27-38 fps

(When I refer to other opponent cars on the start/finish straight, I mean
during laps of a race, not right at the beginning. At the very start of a
race, both cards dropped to about 12 fps.)

For starters, I'm amazed that the Voodoo3 kept up as well as it did,
considering it's a generation older than the Geforce2. In fact, the Geforce2
seemed to be slower in the high detail start/finish area, but then bounce up
higher in the less detailed areas. Maybe it is more processor-dependent,
i.e. when the game can generate simpler scenes more quickly, the Geforce2
can blast them onto the screen faster. But it's strange that it's slower in
the more complex section. Especially since the current issue of PC Gamer has
a comparison test between a reference Geforce2 board and a beta Voodoo 5500,
and the Geforce easily beats the 5500 in their sample games. (They tested
the cards in an Athlon 1000Mhz system.) Is this the result of FSAA?

The Geforce2 came with a second set of drivers from nVidia as well as the
Creative Labs ones. I didn't find any noticable performance difference but
they provide a very detailed tweaker for setting all sorts of parameters.
The above tests were done with all options in their default state. By
default, the FSAA slider is set about 1/3 of the way from the left
(minimum). I didn't find any documentation but I'm guessing this is 1X FSAA
and you could move it up to get 2X and 4X quality. I went back and looked at
the game, trying to see if it looked 'smoother'. To be honest, I didn't see
any great difference. There were still noticable jaggies. Maybe the image of
the car rotating when you select it was cleaner, but in the game it looked
much the same. Maybe the higher FSAA rates would improve the quality, but
the PC Gamer tests show a large performance hit.

So I think the Voodoo3 is well suited to my class of system with the current
state of games. The Geforce2 may provide better speed once more games start
using hardware T&L, or when I upgrade to a faster system, but for now the
Voodoo3 seems just as good and a fair bit cheaper.

        ...Klinn

I have a V3-2000, so I'm not a Voodoo naysayer.  However, I think you might
want to check it in some of the other games you play before unequivocally
saying "the current state of games".  It has already been reported here that
F12K seemed to be better on voodoo cards.  Perhaps many flight and racing sims
really are CPU-limited until, as you say, hardware T&L, or some other
whiz-bang features of these cards, is supported.  My own tests with OC'ing my
V3-2000 indicates that this is probably the case with F12K.  But if you still
have that GeForce2, I'd be interested if you can test some others games
besides F12K vs your V3.

One conclusion that I 've been coming to recently is that graphics
performance is very closely related to the game.  As you've probably
noticed, almost all of the performance tests are done with Quake3 demo.
Quake3 runs an excellent opengl engine that is very scalable with your
graphics hardware.  I just saw some benchmarks with the Geoforce2 and Unreal
tournament and the TNT2 cards did almost as well as the Geoforce2.  The
culprit appears to be the game engine or more likely Direct3D.  Maybe this
will change with DirectX 8, maybe not.  Something is rotten and nobody wants
to talk.  Unless your a Quake3 head don't junk your graphics cards just yet.
Try to get some benchmarks with the games you play.

Jcharm

Very key.
I'm sure as fellow simheads, you too are tired of the endless string of Q3
tests used on these cards, and a whole bunch of other First Person Shooter  
tests, if you are like me, you don't play.  Anandtech admits that he doesn't
use flight/driving sims as they don't show much difference between these cards
since they are more CPU-limited.  He has a point as he wants to test graphics
card limits, not CPU limits.  However, the bottom line for the sim community
is: what graphics card is best (whether that is pure fastest, prettiest, best
value, or combo thereof) on sims.  Not which gives 120 fps, or lets you go to
1600x1280, in Quake 3.  If, and that's still an "if", there is no or just a
slight difference in the vast majority of what you play (and in some cases,
and older board works better than the new $400 whiz-bang special), why switch?

Interesting results.  Please, do a similar comparison with GPL!!!!

Marc.

Can you tell me where I can get Fraps 1.2?

Many thanks in advance.

FYI, check out my post on V3/GF2 GTS and GPL

Keith

...<heh>... you got me - I didn't phrase that last paragraph very well.
Extrapolating from some rudimentary tests in one game suddenly cover *all*
games is kind of a stretch!

Unfortunately I could only borrow the card for a short time, so I won't be
able to test other games, or fiddle with the 'tweaker' to check the impact
of that FSAA slider.

        ...Klinn

Unfortunately I could only borrow the card for a short time, so I can't do
any other tests. Sorry. To be honest, I didn't even think of trying GPL,
given the problems people have with non-Glide cards and the hacks they have
to use to make it run acceptably.

        ...Klinn

www.fraps.com