HQH's Resources: RageLtMan's ATI Driver Sets: Benchmarks With ATI Rage Mobility M1

Location: Editorials - Reviews > RageLtMan's ATI Driver Sets: Benchmarks With ATI Rage Mobility M1

Main Menu


	FastCounter by bCentral

Author: HQH
Released: July 12, 2001
Updated: November 11, 2002

Testing Components

Compaq Presario Notebook 17XL260

Audio	ESS Allegro (ES1988) & built in JBL Pro stereo speakers
CPU	Intel Pentium III (Coppermine) 500 MHz w/256 KB L2 cache
HDD	Ultra DMA 6 GB HDD
OS	Windows Millennium
RAM	128 MB of SDRAM PC-100
Video Adapter	ATI Rage Mobility M1 2x AGP 3D 8 MB of RAM

Driver sets tested	OpenGL ICDs tested
UD v1.11 UD v2 UD v3 beta FD v3 FD v4 RC1 FD v4 RC4 Dx8D (R1) Dx8D R2	1076 1077 1085 1091 1094 1096 1100 1103 1104

Note: I don't have these drivers up for download nor do I have the time to send them out on request. The closest drivers you can get are over at Big Z's website here. Click on the "Rage Pro Chipset" link.

What & How It's Tested

What is tested: ATI Rage Mobility M1's Direct3D and OpenGL performance over a wide variety of RageLtMan's ATI drivers. A little image quality comparison, but not by much.

How it is tested: For every driver used, I do an install over the previous driver loaded on the computer. I find no reason to do what people claim "clean installs." Clean installs has no performance effect whatsoever on these tests, as shown in the benchmarks results - they do change in the case of drivers changed. I have tested clean installs prior to the actual testing and found no discrepancies in the results that were produced.

During some testing, the fan of my laptop does turn on because of heat accumulating over long periods of testing. I find that running benchmarks at a later time and at a cooler temperature doesn't effect performance at all. Because I know that massive heat buildup internally, I have stripped my DVD-ROM drive, my network card and the battery. With these three components eliminated, heat buildup is reduced.

There is no overclocking done on any components. Everything is stock for stock.

Benchmarks that I've used are Half-Life v1.1.0.6, the timedemo blowup 1.1.0.4 and 3D Mark 2001. 3D Mark 2001 is a shareware benchmark utility made by Mad Onion. What's with the lack of benchmarking applications and tools? I don't need to use a lot of benchmark applications and tools just to determine Direct3D & OpenGL performances. That's why I only used two applications, which are sufficient enough to determine whether the drivers are doing poorly or excellent.

I ran Half-Life with the following settings:
Command line: hl.exe -console -heapsize 65535 -zone 1024 -gamegauge blowup1104

For Half-Life tweaks, I had the following:
bgmbuffer "0"
hisound "0"
loadas8bit "1"
s_buffersize "32768"
m_filter "1"
cl_bob "0"
cl_bobocycle "0"
cl_bobup "0"
r_drawviewmodel "0"
r_mmx "1"
scr_conspeed "10000"

Everything else was left at default values. Half-Life resolution: 640x480, 16 bit color.

3DMark2001 settings:
Display and CPU Settings:
- Display Settings:
- Resolution and color depth: 640x480, 16 bit color
- Antialiasing: None
- Frame Buffer: Double
- Rendering Options:
- Texture Format: 16 bit
- Z-Buffer Depth: 16 bit
- Rendering Pipeline: D3D Software T&L

Selected Tests:
- Game Tests: 6 of 7
- Game 1 - Car Chase
- Game 2 - Dragothic
- Game 3 - Lobby
- Theoretical Tests: 4 of 6
- Fill Rate
- High Polygon Count
- DX8 Feature Tests: 0 of 3
- Pro Version Tests: 0 of 4

Options:
- Disable title screens between tests

You should not expect the following tests to predict other application's Direct3D & OpenGL performances. Results provided herein are to be compared with the respectable benchmark applications used - Half-Life v1.1.0.6 and 3D Mark 2001.

Results

Half-Life OpenGL ICD Benchmarks

Click for bigger resolution picture - 223 KB

Half-Life OpenGL ICD Benchmarks Data Table:
The first row indicates OpenGL ICD version being tested, second row is average frame rate, third row is minimum frame rate, fourth row is maximum frame rate:

1076	1077	1085	1091	1094	1096	1100	1103	1104
24.84	24.65	23.41	23.43	23.61	23.53	23.54	23.53	23.53
9	9	9	9	8	8	8	8	7
53	54	49	50	49	49	48	49	51

Half-Life Driver Set OpenGL ICD Benchmarks

Click for bigger resolution picture - 124 KB

Half-Life Driver Set OpenGL ICD Benchmarks Data Table:
The first row indicates driver set being tested, second row is OpenGL ICD version, third row is average frame rate, fourth row is minimum frame rate, fifth row is maximum frame rate. NT means not tested. This is because some driver sets use the same OpenGL ICD, so performance will be nearly identical. To prove this, I ran Dx8D R2 w/Dx8 that had the same OpenGL ICD as UD v1.11. As you can see, different driver sets offer very little performance differences:

UD v1.11	UD v2	UD v3 beta	FD v3	FD v4 RC1	FD v4 RC4	Dx8D R1	Dx8D R2 w/Dx8	Dx8D R2 w/Dx7
1077	NT	1096	NT	NT	NT	NT	1077	NT
24.57	NT	23.37	NT	NT	NT	NT	24.62	NT
9	NT	4	NT	NT	NT	NT	9	NT
56	NT	50	NT	NT	NT	NT	56	NT

Half-Life Driver Set D3D Benchmarks

Click for bigger resolution picture - 228 KB

Half-Life Driver Set D3D Benchmarks Data Table:
The first row indicates driver set being tested, second row is the driver set's Direct3D HAL version, third row is average frame rate, fourth row is minimum frame rate, fifth row is maximum frame rate:

UD v1.11	UD v2	UD v3 beta	FD v3	FD v4 RC1	FD v4 RC4	Dx8D R1	Dx8D R2 w/Dx8	Dx8D R2 w/Dx7
DX6, 6017	DX6, 6017	DX6, 6017	DX7, 1021	DX8, 2010	DX7, 1027	DX8, 2010	DX8, 2010	DX6, 6017
20.15	20.35	20.31	18.89	18.4	18.68	18.44	18.4	20.31
6	6	7	6	3	6	6	6	7
42	40	40	40	40	41	39	39	41

3D Mark 2001 Benchmarks

Click for bigger resolution picture - 99 KB

3D Mark 2001 Benchmarks Data Table:
The first row is the driver set tested, second row is the driver set's Direct3D HAL version, third is how many 3D Marks it earned. Failed means that 3D Mark 2001's Game 1 High Details wouldn't run, it crashed, therefore can't compute final 3D Marks score. Also, the ATI Rage Mobility M1 couldn't run Game 4 benchmark, meaning all driver sets got 'Not supported by hardware':

UD v1.11	UD v2	UD v3 beta	FD v3	FD v4 RC1	FD v4 RC4	Dx8D R1	Dx8D R2 w/Dx8	Dx8D R2 w/Dx7
DX6, 6017	DX6, 6017	DX6, 6017	DX7, 1021	DX8, 2010	DX7, 1027	DX8, 2010	DX8, 2010	DX6, 6017
537	542	549	536	Failed	541	Failed	Failed	548

Conclusion

Here are my conclusions based on testing RageLtMan's ATI drivers on a Rage Mobility M1:

OpenGL ICDs

1076, 1077	Very fast, no problems.
1085	Slow downs, very choppiness movements, sound crackles. In other words, very bad.
1091, 1094 1096	Almost same as 1085 - still slow, no performance.
1100, 1103, 1104	Performance improves over 1085 & 109x, Half-Life crashes with 1103 & 1104, some choppiness movements but playable than 1085 & 109x, some sound crackling, although 1104 improves the sound problem.

So what is the fastest OpenGL ICD? 1076 or 1077. Since 1076 and 1077's performance are very close together, use 1077. Most likely, 1077 updates 1076's issues. What about its image quality? Well, I can't really determine from OpenGL ICD 1076 through 1104. There wasn't anything to look for. What I can tell you that from OpenGL ICD 1076 through 1104, they all give stellar image quality. Not a single one had significant problems that I could detect unlike Direct3D's image quality problems...

What about Direct3D? For the best results, go with the driver set Dx8D R2 with DirectX 7 choice (With Dx8D R2, you get to pick between using DirectX 7 or DirectX 8). One thing I noticed is that with Dx8D R2's choice between DirectX 7 or 8, it's actually DirectX 6's Direct3D HAL and DirectX 8's Direct3D HAL. The ATI Rage Mobility doesn't have DirectX 8 features, so don't think that DirectX 8 drivers will give you the benefit though.

For Direct3D's image quality, there were massive sporadic problems. Some driver sets rendered in high details benchmarks gave off discolorations, unable to render textures correctly, etc. Rest assured, driver set Dx8D R2 with either DirectX 7 or 8 choice will render everything as perfect as it can get. Another problem is that when I was benchmarking with D3D as hardware render choice in Half-Life, some driver sets blurred all the textures.

In the event that you ever play games with an ATI Rage Mobility, or perhaps any ATI Rage Pro chipset that uses these driver sets, always choose OpenGL if there's the option. OpenGL provides top performance and quality more than Direct3D can. ATI Rage Mobility and the Rage Pro works better with OpenGL than Direct3D. Although there might be a chance that it's in the reverse role - Direct3D outperforms OpenGL. So choose wisely from OpenGL and Direct3D if one way doesn't work the way it should.

Sub Menu

Testing Components
What & How It's Tested
Results
Conclusion