Nvidia GeForce 2 Ultra Preview (Page 3 of 5)

Posted: August 14, 2000
Written by: Dan "Tweak Monkey" Kennedy

Test System

We would have had a very large range of systems to test the GeForce 2 Ultra with, but unfortunately, we were restricted by a deadline and the fact that we only had one GeForce 2 Ultra to use for benchmarking. For these reasons I used my personal machine to test the GeForce 2 GTS and GeForce 2 Ultra:

Celeron 566 CPU @ 850 MHz (100 MHz FSB)
128 MB PC133 RAM @ true CAS2
Abit BF6 motherboard (BX)
Sound Blaster Live! Retail
Adaptec 29160 UWSCSI card
Windows 98
Quake3 fresh install/3DMark 2000 v1.1 fresh install/Creature demo fresh install.
Quake3 Timedemo 1 / Demo 1 was used for testing.
6.16 Detonator 3 drivers were used in all testing unless otherwise noted. Vertical Sync (VSYNC) was disabled.

Benchmarks: GeForce 2 vs. GeForce 2 Ultra

Since the time I had to write this preview was extremely limited, we used 3DMark 2000 to test Direct3D, and we used Quake 3 Arena to test OpenGL. We also used Nvidia's Creature demo to test OpenGL w/ T&L.

Direct3D Performance: 3DMark 2000

Due to the lack of a better, more consistent Direct3D benchmark, 3DMark 2000 v1.1 was used. The settings were at 1024x768x16bpp with 32-bit textures enabled. If we had more time for testing we could've tried 1600x1200 or the same benchmarks with FSAA -- maybe we'll do this at a later time.

GeForce 2 GTS using 5.33 drivers: 5227 3DMarks
GeForce 2 GTS using 6.16 drivers: 5387 3DMarks

GeForce 2 Ultra w/ 5.33 drivers*: 5268 3DMarks
GeForce 2 Ultra w/ 6.16 drivers: 5642 3DMarks

* -- The GeForce 2 Ultra with 5.33 drivers (pre-GeForce 2 Ultra release) was for a later article comparing Detonator 3 and older Detonators.

As the numbers show, 3DMark 2000 saw a small improvement from the GeForce 2 Ultra. That isn't to say Direct3D won't show major improvements over the GeForce 2. It's just that with this benchmark, the difference is hard to see.

I also had the time to play with the GeForce 2 Ultra in Unreal Tournament and in Motocross Madness 2. I am happy to report that in both games, the performance increase was substantial if using the 6.16 Detonator 3 drivers. MCM2 was fully playable at 4x4 FSAA (full-scene anti-aliasing) 1024x768 -- which is very good considering this game is a hog on some systems.

OpenGL Performance (1): Nvidia Creature Demo

One of Nvidia's latest tech demos, Creature, is an underwater scene with very impressive visuals. The scene features several schools of fish in the background and of course, the creature. The demo runs at 1024x768 and has between 130,000 and 160,000 triangles in the scene at a time, or so, with plenty of eye candy.

Click to enlarge

The nice thing about this tech demo, other than the visuals, is the frame rate and polygon/triangle count, which helps to test T&L performance. Here's how the GeForce 2 GTS stacked against the GeForce 2 Ultra in this test:

As you can tell from the graphs, the GeForce 2 Ultra is quite a bit faster than the GeForce 2 GTS with this OpenGL/T&L benchmark. Not only that, but the GeForce 2 Ultra's extra memory and even higher speed allow it to have an increasing number of polygons with little performance hit; whereas the GeForce 2 GTS's performance was considerably lower with higher polygon counts. Perhaps the bigger issue with this sudden drop in performance around (140,000 triangles) is due to the limit on the GeForce 2 GTS's fillrate.

Next Page