PDA

View Full Version : GeForce3 Detonator Driver Tests


NickSpolec
05-31-03, 10:34 PM
Hey all,

These following tests were from various Detonator drivers for a vanilla GeForce3.


************
SYSTEM SPECS
************

Athlon XP 2500+ (1.83ghz @ 2.16ghz)
Shuttle AK37GT, Via KT400 (@180 FSB)
512MB OCZ PC2700 ULTRA Rev 3.2 (@DDR 360, CAS 2, 2,5,2, 1T)
VisionTek GeForce3 64MB (@255/540)
Sound Blaster Live Value
Western Digital 40GB (2mb buffer), ATA100 Hard drive.
LiteOn LTR-16102B (40x16x12x)
OS = Win98SE



It's important to note that the GAME tests were not being performed as a service to the community. These were personal tests for me to find the best driver for MY needs. That just so happens to include 2XAF, and generally a max resolution of 640x480.

Why only 640x480? Because, I usually play exclusively on a TV, and anyone that plays on a regular analog TV knows that anything other then 640x480 looks like sh*t.

I just thought I would share my results with everyone.




**********************
CARD and DRIVER SETTINGS
**********************

The GF3 was overclocked to 255 core, and 270 memory (540 DDR)

AGP Fastwrites was enabled.

AGP Sideband Addressing was enabled.

AGP 4X was enabled.


For all tests, No Anti-Aliasing was used.

For all tests, 2XAF was forced in all applications.

-Ansi-Booster patches from RivaTuner were applied to
both D3D and OpenGL parts of each driver set.

-D3D had the texture stages of 1, 2 and 3 being forced
to use a max Antisopy of bilinear, with trilinear being
disabled. However, the main texture stage, 0, was not
forced below the desired 2xAF, and Trilinear was not
neccessarily disabled. Weather any of these
optimizations are actually working, I can't tell.

-OpenGL was set to PERFORMANCE Optimization for
Antisopic Filtering

32Bit Color was used for all tests.

VSync was off for both D3D and OpenGL.

A LOD of 0 was set for both D3D and OpenGL.

OpenGL used DXT3 texture compression instead of DXT1.





First up is 3DMARK 2001SE.......

NickSpolec
05-31-03, 10:38 PM
3DMARK 2001SE


Detonator 30.83 - 8966

Game 1 - Car Chase - Low Detail 137.4 fps
Game 1 - Car Chase - High Detail 66.9 fps
Game 2 - Dragothic - Low Detail 126.5 fps
Game 2 - Dragothic - High Detail 72.0 fps
Game 3 - Lobby - Low Detail 140.5 fps
Game 3 - Lobby - High Detail 70.9 fps
Game 4 - Nature 36.3 fps
Fill Rate (Single-Texturing) 528.2 MTexels/s
Fill Rate (Multi-Texturing) 1031.8 MTexels/s
High Polygon Count (1 Light) 29.6 MTriangles/s
High Polygon Count (8 Lights) 6.5 MTriangles/s
Environment Bump Mapping 147.5 fps
DOT3 Bump Mapping 86.7 fps
Vertex Shader 84.0 fps
Pixel Shader 86.0 fps
Advanced Pixel Shader 62.5 fps
Point Sprites 18.8 MSprites/s


Detonator 41.09 - 9427

Game 1 - Car Chase - Low Detail 133.5 fps
Game 1 - Car Chase - High Detail 64.2 fps
Game 2 - Dragothic - Low Detail 144.8 fps
Game 2 - Dragothic - High Detail 83.8 fps
Game 3 - Lobby - Low Detail 140.1 fps
Game 3 - Lobby - High Detail 70.3 fps
Game 4 - Nature 43.9 fps
Fill Rate (Single-Texturing) 525.9 MTexels/s
Fill Rate (Multi-Texturing) 1026.8 MTexels/s
High Polygon Count (1 Light) 31.4 MTriangles/s
High Polygon Count (8 Lights) 6.5 MTriangles/s
Environment Bump Mapping 145.9 fps
DOT3 Bump Mapping 86.5 fps
Vertex Shader 50.5 fps
Pixel Shader 86.0 fps
Advanced Pixel Shader 64.4 fps
Point Sprites 18.9 MSprites/s


Detonator 43.00 - 9723

Game 1 - Car Chase - Low Detail 134.8 fps
Game 1 - Car Chase - High Detail 64.5 fps
Game 2 - Dragothic - Low Detail 155.8 fps
Game 2 - Dragothic - High Detail 84.4 fps
Game 3 - Lobby - Low Detail 142.3 fps
Game 3 - Lobby - High Detail 70.7 fps
Game 4 - Nature 50.2 fps
Fill Rate (Single-Texturing) 527.7 MTexels/s
Fill Rate (Multi-Texturing) 1026.7 MTexels/s
High Polygon Count (1 Light) 31.4 MTriangles/s
High Polygon Count (8 Lights) 6.5 MTriangles/s
Environment Bump Mapping 148.5 fps
DOT3 Bump Mapping 87.0 fps
Vertex Shader 51.0 fps
Pixel Shader 86.8 fps
Advanced Pixel Shader 64.5 fps
Point Sprites 18.9 MSprites/s

NickSpolec
05-31-03, 10:57 PM
Now, it's time for some GL EXCESS action....



30.83 - 6466
41.09 - 6596
43.00 - 6556


I have attached the original results.htm file so you can see a DETAILED result of the entire benchmark suite.

The attachment has been renamed results2.txt. YOU MUST RENAME THIS WITH A .HTM EXTENSION TO VIEW IT ONCE YOU DOWNLOAD IT

NickSpolec
05-31-03, 11:13 PM
SPLINTER CELL is next

Game Settings:

640x480
All Visual Settings at Medium (Shadow Res, Shadow Detail, Effects Detail)
Sound Enabled (High Audio Visualization Used)


NOTE: All game tests involved me actually playing the game. I used FRAPS to log the FPS. Due to this, it is not PERFECT like a synthetic benchmark. I have made many efforts to ensure it is as close as I can get it.. It should be subjective enough to give you an idea. I tried to get the amount of time spent on each test for each driver as close as possible. The amount of time is important, not the frames



3 Different Tests were done.


Test #1 - Training Part 1

This is the first part of the training mission, the obsticle course. The mission was played through, eleminating all hints and pop up messages. I then returned to the beginning, saved, and started again, with no interruptions, this time logging the FPS.Basic level stuff here


Test # 2 - Training Part 2

This is the second part of the training mission, the tactical coarse. The mission was played through, eleminating all hints and pop up messages. I then returned to the beginning, saved, and started again, with no interruptions, this time logging the FPS. This part is you basic in-door fare. Lots of lights and shadow casting, though.


Test # 3 - CIA Headquarters, Doughrety Carry

This takes place inside CIA Headquarters, from the time you knock out Mitchell Doughrety and the pick up point. Going from the time I knocked out Doughrety, I went to the end of the stage, taking out all the guards, then and went back to the "Smoke Room". I then started from that point and went to end of the level, without Doughrety, logging the FPS this time. This part features heavy rain, half in doors/half out doors, with some vast structures and lots of light/shadow casting.




GAME TEST #1

30.83
Frames: 6306 - Time: 124854ms - Avg: 50.506 - Min: 35 - Max: 113

41.09
Frames: 7675 - Time: 126747ms - Avg: 60.553 - Min: 36 - Max: 141

43.00
Frames: 7364 - Time: 121555ms - Avg: 60.581 - Min: 37 - Max: 137



GAME TEST #2

30.83
Frames: 7801 - Time: 122970ms - Avg: 63.438 - Min: 31 - Max: 115

41.09
Frames: 8595 - Time: 124659ms - Avg: 68.948 - Min: 34 - Max: 112

43.00
Frames: 8480 - Time: 123010ms - Avg: 68.937 - Min: 34 - Max: 113


GAME TEST #3

30.83
Frames: 4401 - Time: 92054ms - Avg: 47.808 - Min: 29 - Max: 76

41.09
Frames: 4916 - Time: 89678ms - Avg: 54.818 - Min: 31 - Max: 80

43.00
Frames: 4809 - Time: 86621ms - Avg: 55.517 - Min: 31 - Max: 80

NickSpolec
05-31-03, 11:22 PM
Finally, here is RALLISPORT

GAME SETTINGS

-640x480
-MAIN VISUAL Settings Medium High View Distance, Player Car Detail, Opponent Car Detail, Shadow Detail
-ALL EXTRA Visual Options Enabled (EXCEPT for 3D Trees)
-Sound Enabled (44.1khz)



NOTE: All game tests involved me actually playing the game. I used FRAPS to log the FPS. Due to this, it is not PERFECT like a synthetic benchmark. I have made many efforts to ensure it is as close as I can get it.. It should be subjective enough to give you an idea. I tried to get the amount of time spent on each test for each driver as close as possible. The amount of time is important, not the frames

NOTE: RALLISPORT REFUSED TO RUN WITH DETS 43.00





5 Different Game tests were used. Any spin outs, crashes, or trips off the track invalided that test, and it had to be started over. The 5 tests were...


Safari SS1 RALLY - Desert, some Tarmac. Lots of bumps, with a drive through a small village.

Pacific SS1 RALLY - Lots of trees, bushes, lots of big rocks. Open, expansive land and view.

Pacific SS3 RALLY - Rain, mud, fair sparse of trees, lots of open land.

Mediterrian SS3 RALLY - Lots of rugged cliffs and swirling turns. Big rocks and boulders, water.

Frost Ridge 1 Ice Circuit (5 Laps) - Ice, snow. Lots of trees, plenty of big structures.




*Safari SS1 RALLY*

30.83
Frames: 11544 - Time: 185390ms - Avg: 62.268 - Min: 44 - Max: 90

41.09
Frames: 12219 - Time: 182801ms - Avg: 66.843 - Min: 47 - Max: 91



*Pacific SS1 RALLY*

30.83
Frames: 8408 - Time: 133444ms - Avg: 63.007 - Min: 51 - Max: 116

41.09
Frames: 10108 - Time: 117485ms - Avg: 86.036 - Min: 56 - Max: 117



*Pacific SS3 RALLY*

30.83
Frames: 9350 - Time: 130953ms - Avg: 71.399 - Min: 54 - Max: 99

41.09
Frames: 9554 - Time: 130748ms - Avg: 73.071 - Min: 56 - Max: 99



*Mediterrian SS3 RALLY*

30.83
Frames: 7973 - Time: 132244ms - Avg: 60.290 - Min: 43 - Max: 81

41.09
Frames: 8152 - Time: 120553ms - Avg: 67.621 - Min: 50 - Max: 98



*Frost Ridge 1 Ice Circuit (5 Laps)*

30.83
Frames: 15184 - Time: 243172ms - Avg: 62.441 - Min: 44 - Max: 81

41.09
Frames: 15718 - Time: 218015ms - Avg: 72.095 - Min: 51 - Max: 118

NickSpolec
06-02-03, 07:15 PM
Oh, some advice for anyone that might play RalliSport on the PC with a GeForce 3 class GPU (even GF4 class)....

DO *NOT* ENABLE 3D TREES!

I have found, in some instances, that is decreases performance by up to 200% or more. General performance decrease is AT LEAST 100%.

This one stretch of straight-away course is very striking in this regard.. It's a very open part with TONS of trees around (I believe it's on Pacific SS1.. er.. One of a pacifics).

With 3D Trees enabled, driving on this stretch of straight-away gives an average of 31 FPS

With 3D Trees disabled, the average is about 93 FPS!