Can you test this benchmark ?

Celeron300A o/c to 450MHz | 128Mb |TNT(viper550) | win98 | 1152*864*16 |

19-22fps and 2.4-2.7M Tri/s

alistair

Edited by - alistair b on July 23, 2000 8:20:32 AM

PIII 450/TNT2 32MB
18-20fps
2.4-2.6 M tri/sec

Zack

PIII-500 / ELSA Gloria L/MX (3Dlabs GLINT MX)
1.5M Tri/s, 12fps

PIII-800 / ELSA Gloria II (NVIDIA Quadro)
10.8M Tri/s, 89fps

If I run at 640x480 in 16 bit color (taking the default window size it gives me), I get almost 6.8 M Tri/s. At 1152x864 (again with default window size) I get 4.68 M Tri/s and 39 fps. My results also were different (read lower) when I was running in 32bit color, as well as when I turned options on and off.

I''m running with an Athlon 700 and a g-force.

What I''m getting at is that the benchmark is going to run at different speeds for more reasons than just processor/ video card combo''s.



Mark Fassett
Laughing Dragon Entertainment
http://www.laughing-dragon.com

PIII-500 | 64MB | TNT2 M64 32MB (Creative) | win98 | 16 Bit color

- default window size
- FPS: 13.2
- 1.59M Tri/s

Celeron 366 @ 550
GEforce SDR

80 fps

9.47M tri/sec

p3 600, 128MB, geforce 256

9.6M Tri/sec @ 78fps

Come on, we need more results

Anyway, here are a few observations:

1. Zeotron''s system (P3 450 + TNT2 ) is around 2.5 Mtri/s, which is almost twice faster than Richardve''s system (P3 500 + TNT2 ) which is around 1.59 Mtri/s.
Did you do the test in the same conditions ? I''m just guessing that Richardve used 32 bits colors and Zeotron 16 bits ? Else i can hardly explain a such difference
To compare i think we''d all need to make it in 16 bit colors in the default window (640x480).

2. So far the best result is Serge, with a PIII 800 + Else Gloria II.. 10.8 Mtri/s.. impressive The worst result is..(*arg*) my computer with 0.11 Mtri/s. Shame on me

3. LaughingD.. i don''t understand for your results. As i understandd it, the results shouldn''t be dependent on the resolution / framerate. Personnally, i get constant results whatever the resolution/framerate is. Though it''s true that the color depth is important. Probably the RAM too.

Y.

thats cause in this test youre not fillrate limited but geometry limited which is what i think its meant to test

ok, i''ve been changing settings around, and i''ve found that changing core speed or memory speed has no effect on my TNT, neither does changing resolution, or colour depth.

the thing which does make a difference is the processor/fsb speed
i get linear scaling with increases in fsb:

300/66MHz__:__16.5fps
450/100MHz_:__24.8fps
464/103MHz_:__25.5fps

the reason your p200/voodoo2 system is so slow is that you''re using software rendering (i think ''cause a p200 isn''t that much slower than a celeron300A)
i get 1.55fps & 0.18MTri/sec on a p225/75MHz with Matrox Mill2

alistair

of course, changing the core/mem speed on a geforce should change the results cause you''ve got hardware T&L (which is what they''re trying to show off!!!)

I used 32bit color.
My desktop resolution is 1152x864.
I have the detonator 2 drivers from NVIDIA.

PIII 450/128 MB RAM/TNT2 32MB
18-20fps
2.4-2.6 M tri/sec

Zack

Hem no, it''s fullscreen hardware-accelerated. In no way the software renderer could throw 110,000 polygons per second on a P200 But it''s an old computer with only 32 Mb of RAM EDO. I''m not really surprized.

Y.

...if you check the who's rendering thing in the 'about' menu then it should say microsoft generic 1.1.0 and list 2 or 3 extensions...this is the software renderer (i'm pretty sure it is anyway)
- the millenium2 doesn't support hardware opengl acceleration under win98 & i get better results than you. i get about 2fps on my TNT if i switch back to Standard Pci adapter VGA (hence disabling hardware opengl and going back to microsoft generic 1.1.0)

alistair

but this really doesn't matter, what matters is that anyone with a geforce gets quite good results, and with everyone else it just depends on cpu - which is what nvidia want

Edited by - alistair b on July 25, 2000 10:30:17 AM

Hell no, i don''t even wonder It''s hardware accelerated. I''m not a "newbie" to OpenGL (though not an expert either), so i''ve already used and compared software/hardware drivers. Anything in software using more than a few tenth of polygons per frame takes up to 10 seconds to render, sometimes even more. Btw, my extensions seem to be correct (i even got multitexturing), the pixel format is accelerated, and the driver is up-to-date (3dfx beta 2.1 driver, the most recent available). I can even play to Quake III in OpenGL (runs around 10/15 fps).
Other tests have proven that i can render around 2000 polygons per frame per second, with a speed of around 60 fps (which gives a total of around 120,000 polygons/sec).

Y.

ok sorry!...the results just surprised me