LOSTCIRCUITS
King of the Hill
|
Advice Beginners BIOS Guide CPUs Links Mainboards Memory Network Storage Video/Sound Cards Contact Forum SiteMap Sponsors WebNews Home |
. | . |
Prices: Mainboards ABIT ASUS Chaintech Shuttle Soyo Tyan CPU Intel P4 2.4C-800 P4 2.6C-800 P4 2.8C-800 P4 3.0-800 P4 3.2-800 AMD AthlonXP XP 1700+ XP 2000+ XP 2400+ XP 2500+ XP 2700+ XP 3000+ XP 3200+ Athlon64 Athlon64 3200+ Athlon64 FX-51 Opteron Opteron 240 Opteron 242 Opteron 244 Opteron 246 Memory Corsair Crucial Kingston Mushkin OCZ |
LOSTCIRCUITS |
|
| nVidia Quadro FX3000 King of the Hill | |
| (Review by MS, October 26, 2003) |
| PNY QuadroFX 3000 at: |
Phoenix
The pyrotechnics effects benchmark was run under A|W Maya 4.5 personal learning edition with the number of sprites set to the maximum value of 55 and with line antialiasing enabled at a screen resolution of 1280 x 1024. Phoenix strongly depends on the amount of frame buffer available for the sprites which is where the FireGL falls a bit short because of its more rigid LFB organization.
Phoenix: runtime in seconds, lower is better. The FX3000, as not expected otherwise, steals the show.
Springmark
Springmark is a huge model of a spring with 1 million vertices per loop that occupies 36 MB of frame buffer memory. Each additional coil will increase the number of vertices and the required amount of memory in a proportional fashion so that at 10 coils, the number of vertices equals 10 million and the required frame buffer memory is 360 MB. Needless to say that the only card that would feature this amount of memory is the 3DLabs Wildcat, every other card needs to access the system memory for DMA. For practical purposes, Springmark does not appear to be too relevant, even though arguably there are models or designs that will occupy that amount of memory, they will hardly be pressed to run in real time rendering mode but then, give it another 2 years and it may happen nonetheless.
Frames per second. At the smallest model size, the FireGL X1 (red) is the undisputed leader but falls off as soon as the model size exceeds the geometry part of the frame buffer. The FX3000 holds (blue) the top position up to 5 loops but has to yield to the Wildcat VP990 (yellow) as soon as the 256 MB of frame buffer are used up. Keep in mind that nVidia uses a flexible allocation of geometry vs. texture memory space in the on-board frame buffer, meaning that the memory can be dynamically allocated on an "as needed" basis to either texture or geometry. The Quadro4 980XGL (green) holds a place of its own.
It is a little bit easier to see what happens by looking at vertices per second instead of fps. Again, the FireGL X1 (red) is the undisputed leader with the smallest coil but falls back at the next higher model size. The FX3000 (blue) hovers at some 55 Million vertices/sec but drops down as soon as the main memory is accessed. On a side note, the 20 loop spring would not run on the Quadro FX3000 unless we increased the amount of system memory to 2 GB AND loaded the Unigraphics specialty driver settings (which slowed down everything else). The Quadro4 980XGL had no such problems.
Next Page: => Conclusions =>
If you enjoyed reading this article and found it useful, please consider making a small donation to LostCircuits.