Inno3D Home Page Inno3D Home Page

FAQ News Search Archive Forum Articles Tweaks Technology Files Prices SETI
Visit NVIDIA's home page.
Favorite Pics
Click to Enlarge
Articles/Reviews
OCZ Tech Titan 3
Absolute MORPHEUS
1.0GHz Pentium III
eVGA MX Shootout
nForce Preview
AMD AXIA CPU
VisionTek GeForce3
2001 Spring Lineup
GeForce3 Preview
eVGA TwinView Plus
VisionTek GF2 Ultra
OCZ GeForce2 Pro
Winfast GeForce2 MX
GeForce2 vs Quake3
Linksys Cable Router
GF2 FSAA Shootout
GeForce2 MX Preview
Benchmarking Guide
Quake 3 Tune-Up
Power DVD Review
Live! Experiences
Sponsors
Memory from Crucial.com


FastCounter by bCentral

 Visitors Are Online
Powered by Perlonline.com
Drivers/FAQ
NVIDIA
3D Chipset
Gamers Ammo
Reactor Critical
GeForce FAQ
Associates
Dayton's Misc.
G-Force X Sweden
Maximum Reboot
Media Xplosion
NVchips-fr
nV Italia
Riva Station
3D GPU
nV News Home Page

3D ExerciZer Benchmark

By: Mike Chambers - October 3, 1999



Introduction

3D ExerciZer is an OpenGL based graphics application designed for use on "Intel-based" workstations running Windows 95/98/NT. 3D ExerciZer is used to stress test the 3D graphics performance of a system and can be used to compare performance of different graphics cards, systems, and drivers.

With NVIDIA's GeForce 256 due to appear shortly, and a lack of readily available "consumer based" software to test the Geforce's Transform and Lighting capabilities, 3D ExerciZer may be a good candidate for comparison purposes (especially the processor and geometry stress tests).



Readme Details

Much of the information that is covered in the Testing section appears in the 3D ExerciZer readme. The readme covers such topics as:
  • Setting up your system for optimal performance
  • Details on how to use 3D ExerciZer
  • Special features found in 3D ExerciZer
  • Details on how the various tests work


The Tests

This is a list of the following benchmark tests that can be run using 3D ExerciZer:
  • Texture stress: 3D ExerciZer comes with 11 unique textures. By increasing the number of textures on the wheel, you load the texture buffer of the graphics subsystem progressively until you hit the physical limit (if there is one). (Note that although you may see the same textures being repeated, 3D ExerciZer is taking copies as if each is a new texture, so that the subsystem is genuinely being stressed.) When the subsystem texture buffer hits the limit, it is forced to start swapping textures to main memory, and the frame rate will reflect this point.


  • Lighting stress: The "Blob" appears first with a spotlight at the camera position. The lighting algorithm requires extensive use of any available geometry acceleration in order to compute path of light waves. Geometry accelerated subsystems will show higher frame rates.


  • Processor stress: The "Explosion" tracks the path of thousands of particles in space emerging from a core. This makes no use of texture memory, geometry acceleration or lighting capabilities, and is a pure processor-related test. The floating-point arithmetic capabilities of the processor will be reflected in the frame rates as you increase the number of particles in the picture.


  • Geometry stress: you begin the test with the figure "Ping", a man floating in space. In his metallic body you see a spherical reflection of a texture. As with all tests, you can change the texture using the "T"/"t" key. You can increase the instances of Ping using the + and - keys and observe the frame rate change.
Having read the following information, that there may be a couple of tests in 3D ExerciZer than will utilize the new features of the GeForce 256: the lighting capabilities in the lighting and processor stress tests and the number of polygons used in the geometry stress test.

During yesterday evenings chat session with NVIDIA's Nick Triantos and Dwight Diercks, I had the opportunity to ask if NVIDIA had tested the GeForce 256 with 3D ExeciZer.

While the response did not directly indicate if 3D ExeciZer would be able to take advantage of the GeForce's Transform and Lighting abilites, it may be a possibility - "We are focusing on running CAD apps as opposed to the Intergraph test (3D Exercizer). Benchmarks for Maya, AWADVS, and others look great!."



Test System

  • Intel Pentium II-450MHz
  • 128 MB Hitachi PC100 SDRAM
  • Abit BH6 440BX Motherboard
  • Diamond Viper V770 Ultra AGP - 175MHz/200MHz
  • Hitachi Superscan Supreme 752 19-inch monitor
  • Seagate Medalst Pro 6.4GB 7200RPM HDD
  • Addtronics 6890A full tower case with 300w PSU


  • Windows 98 with DirectX 6.1
  • NVIDIA Detonator Drivers - version 4.11.01.0208
  • Powerstrip - version 2.50.01


Benchmarking Results

The following bar charts indicate the benchmark results. The resolution and color depth used are listed above the chart.

Texture Test - 1280x1024 @ 32-Bit Color

Texture Test - 1280x1024 @ 32-Bit Color

What's intersting in the texture test is the drop to 1 frame per second after more than 32MB of textures are used. It appears that the use of system memory via the AGP bus is not being handled properly via an undetermined system component.

Lighting Test - 1280x1024 @ 32-Bit Color

Lighting Test - 1280x1024 @ 32-Bit Color

The light source test doesn't appear to cause a great deal of stress on the Viper V770 Ultra. The performance decreases as more light sources are added, but the changes are not too severe. It is quite possible that the benchmark is not complex enough to stress the geometry acceleration necessary to compute path of light waves.

Processor Test - 1280x1024 @ 32-Bit Color

Processor Test - 1280x1024 @ 32-Bit Color

There processor test, which relies heavily on the floating point capabilities of the CPU, causes a severe initial loss when moving from 1,000 to 1,400 particles. The subsequent decreases in performance after adding more particles are linear and do not result nearly as much of a loss as the initial test.  This observation was also mentioned in the readme.

Geometry Test - 1280x1024 @ 32-Bit Color

Geometry Test - 1280x1024 @ 32-Bit Color

There is definitely a huge corresponding decrease in the the frames per second, from 79 to 42, when the number of polygons were doubled from 4396 to 8792. Additional increases in the number of polygons did not affect performance as nearly as much.

This is more than likely due to the processors inability to maintain the high rate of instructions required by the graphics card to render the polygons. The GeForce 256 should significantly improve benchmark results in the geometry stress test.



Photo Gallery

The following screenshots were taken at a full screen resolution of 1152x864 (except for the Introdcution which was taken in a window).

Introduction

Full size is 67K

Texture Test

Full size is 67K

Lighting Test

Full size is 195K

Processor Test

Full size is 17K

Geometry Test

Full size is 147K

Credits

Full size is 62K




Final Words

Hopefully, the 3D ExerciZer benchmark will make use of the GeForce's Transform and Lighting capabilites. In the chat, Nick made a comment in regards to the types of OpenGL applications that would need to be retrofitted with T&L code: "Most OpenGL apps *WILL* have xform and lighting acceleration, but some only use the rasterization part of OpenGL. Those games will need some work before they'll be T&L acceleration."


Last Updated on October 3, 1999

All trademarks used are properties of their respective owners.