VOICE Home Page: http://www.os2voice.org
[Newsletter Index]
[Previous Page] [Next Page]
[Features Index]

March 2001
editor@os2voice.org

The AOpen PA256MX display card and Scitech Display Doctor for OS/2

Article by Gilbert Lefebvre ©March 2001

AOpen PA256 MX: http://aoaen.aopen.com/products/vga/pa256mx.htm
SciTech Display Doctor Professional for OS/2: http://www.scitechsoft.com/sdd_os2.html

I consider myself as one of the lesser known OS/2 users although I have been using it since version 1.3 but really got into it at version 2. I am also one of the co founders of the Montreal OS/2 user group and present president. Ever since I have been involved in personal computers, I have been interested in hardware: hard disks at first but nearly all the rest of the system as time went by. I spend a good two hours every day fulfilling my hardware curiosity.

In September of 1999, I upgraded most of my system with a new full tower case with 300W power supply to house a new Abit BP6 twin Celeron motherboard, new 19" ViewSonic G790 monitor and new Lexmark Z51 printer. My main hard drive was less than a year old as well as my CD-ROM reader.

There was one weak component to this system and it was my graphic card which was a Matrox Mystique 220. I felt ready to wait until something special appeared. Choosing the right video card for OS/2 is not the easiest task as one needs a good driver. With the exception of Matrox, there are not many manufacturers that still support OS/2. The other aspect of the equation is that most hardware sites are nearly exclusively Windows oriented. How the card will fare under OS/2 is something that often remains a well kept secret.

In early Fall of 2000, I read good reviews of NVIDIA Geforce 2 MX chipsets as providing the best bang for the buck. I wrote Scitech's Kendall Bennett about possible OS/2 future support and he answered back that it was to be in the next beta of Scitech's Display Doctor(SDD). I asked in comp.os.os2.video about Nvidia's driver satisfaction for OS/2 but never got an answer. When I searched Deja News for info on NVIDIA, I did not find much of a lead. I kept reading reviews upon reviews on implementations of this chipset from different manufacturers such as Asus, MSI, Guillemot, Hercules, Creative Labs etc. All reviewers agreed that it was the best card one could get for one on a budget. Let's say I don't believe that getting a second mortgage on the house is a good idea to purchase computer hardware and also believe the cheapest is probably not the best buy one can make. I ended up with an AOpen PA256MX.

There are three flavors of this card: no tv output, tv output, tv and digital output. My choice was on the no tv output. For full specifications on this card: http://aoaen.aopen.com/products/vga/pa256mx.htm

Installation.

Before doing anything, I installed the latest beta version of SDD which went fine with Warp4 fp12 but would not install on my WSeB without applying FP2 to it. The SDD driver can be considered to be a generic driver and changing graphic cards does not require one to revert back to VGA. I simply removed my Matrox PCI card and installed the new 32 MB AGP card in its slot and rebooted in both OS/2 Warp4 and WSeB with no problem. There are drivers, in the box on a CD, for all flavors of Windows as well as for Linux and instructions to get the driver for BeOS.

The situation was different for my NT4 partition and, to make a long story short, requires that you assign an IRQ to VGA in the BIOS.

Differences and comparisons.

Since my screen is 19", my favorite resolution is 1280 by 1024 pixels. With my Matrox, the color depth was 65,566 colors with a 65 Hz refresh rate. I had done quite a few tests with Trevor Hemsley's Benchmark at these setting and proceeded to do the same with the new card. Here are my Matrox results with Sysbench 0.9.4d result file created Wed Feb 16 01:04:24 2000:

Machine Data

Brand/Board
Chipset
CPU
Coprocessor
Processors
RAM
OS/2 Version
CSD Level
Fix Level
Revision #
Priority
Maxwait
Timeslice
Protectonly
Motherboard - Abit BP6
Intel - 440BX
Intel Celeron 550xMHz with MMX stepping 6.6.5
Yes
1
112.00 MB
20.40
XRCM010_
XRCM010_
8.000
Dynamic
3
(32,32)
NO

Graphics card - Matrox Graphics, Inc. - MGA 1064SG "Mystique", 4MB, PCI

DIVE tests
Vid Mem
Video Bus Bandwidth
DIVE fun
M->S DD 1.00:1
Dive-marks
4MB
73.409 Megabytes/second
251.519 fps normalised to 640x480x256
250.912 fps normalised to 640x480x256
93.801
Video tests - Resolution 1280x1024x16 bits/pixel
Vid Mem
BitBlt S->S Copy
BitBlt M->S Copy
Filled Rect.
Pattern Fill
Vert. Lines
Horiz. Lines
Diag. Lines
Text Render
PM-marks
4 MB
68.344 Million pixels/second
40.985 Million pixels/second
1698.032 Million pixels/second
847.573 Million pixels/second
3.752 Million pixels/second
399.917 Million pixels/second
7.549 Million pixels/second
84.683 M Million pixels/second
372.820

Here are the results with the Aopen MX256 using Sysbench 0.9.4f result file created Fri Jan 12 03:48:45 2001

Machine Data

Brand/Board
Chipset
CPU
Coprocessor
Processors
RAM
OS/2 Version
CSD Level
Fix Level
Revision #
Priority
Maxwait
Timeslice
Protectonly
Motherboard - Abit BP6
Intel Corporation - 82443BX 440BX PCI-Host Bridge (AGP Enabled)
Intel CeleronA 366@550MHz (stepping 665) with MMX
Yes
1
256.00 MB
20.40
XRCM010_
XRCM010_
8.000
Dynamic
3
(32,32)
NO

Graphics card - NVidia - GeForce 2 MX, 32MB, AGP

DIVE tests
Vid Mem
Video Bus Bandwidth
DIVE fun
M->S DD 1.00:1
Dive-marks
32MBs
229.213 Megabytes/second
760.380 fps normalised to 640x480x256
781.615 fps normalised to 640x480x256
290.716
Video tests - Resolution 1280x1024x16 bits/pixel
Vid Mem
BitBlt S->S Copy
BitBlt M->S Copy
Filled Rect.
Pattern Fill
Vert. Lines
Horiz. Lines
Diag. Lines
Text Render
PM-marks
32 MB
568.925 Million pixels/second
81.553 Million pixels/second
2286.626 Million pixels/second
2265.326 Million pixels/second
39.687 Million pixels/second
449.450 Million pixels/second
29.825 Million pixels/second
111.034 Million pixels/second
683.581
In short:

Matrox: 372.820 PM-Graphics-marks et 93.801 DIVE-marks

Aopen: 683.581 PM-Graphics-marks et 290.716 DIVE-marks

The difference can be felt. As to the quality of the screen output, Matrox has the reputation of offering the best but could not notice a difference in this regard with the new card.

The Matrox is a 4 MB card of 1997 vintage while the Aopen has 32 MB and is a 2000 vintage. The new card offers additional capabilities such as true color depth and higher refresh rates. I then proceeded to set the card for 32 bits per pixel and 85 Hz refresh rate and proceeded to run the benchmark using Sysbench 0.9.4f, result file created Fri Jan 12 03:59:19 2001:

Machine Data

Brand/Board
Chipset
CPU
Coprocessor
Processors
RAM
OS/2 Version
CSD Level
Fix Level
Revision #
Priority
Maxwait
Timeslice
Protectonly
Motherboard - Abit BP6
Intel Corporation - 82443BX 440BX PCI-Host Bridge (AGP Enabled)
Intel CeleronA 366@550MHz (stepping 665) with MMX
Yes
1
256.00 MB
20.40
XRCM010_
XRCM010_
8.000
Dynamic
3
(32,32)
NO

Graphics card - NVidia - GeForce 2 MX, 32MB, AGP

DIVE tests
Vid Mem
Video Bus Bandwidth
DIVE fun
M->S DD 1.00:1
Dive-marks
32MBs
201.876 Megabytes/second
686.493 fps normalised to 640x480x256
683.273 fps normalised to 640x480x256
256.024
Video tests - Resolution 1280x1024x32 bits/pixel
Vid Mem
BitBlt S->S Copy
BitBlt M->S Copy
Filled Rect.
Pattern Fill
Vert. Lines
Horiz. Lines
Diag. Lines
Text Render
PM-marks
32 MB
238.114 Million pixels/second
50.590 Million pixels/second
967.740 Million pixels/second
963.436 Million pixels/second
20.060 Million pixels/second
448.753 Million pixels/second
25.123 Million pixels/second
112.565 Million pixels/second
328.605

The speed penalty is quite high for running in 32M colors. Is this setting worth it? I would say it is not worth it for the desktop and web browsing. For those occasions where you do some photo retouching, you can always set it at true color but do not see the need for general use that the home user is likely to do considering the speed penalty. Of course, the speed is quite acceptable but not as snappy.

How would it do on a smaller monitors such as a 15" running at 800 by 600? I just put up a new system with an AMD Duron running at 750 Mhz at true color on Warp4 fp14:

Graphics card - NVidia - GeForce 2 MX, 32MB, AGP

DIVE tests
Vid Mem
Video Bus Bandwidth
DIVE fun
M->S DD 1.00:1
Dive-marks
32MBs
63.128 Megabytes/second
215.546 fps normalised to 640x480x256
215.498 fps normalised to 640x480x256
80.548
Video tests - Resolution 800x600x32 bits/pixel
Vid Mem
BitBlt S->S Copy
BitBlt M->S Copy
Filled Rect.
Pattern Fill
Vert. Lines
Horiz. Lines
Diag. Lines
Text Render
PM-marks
32 MB
285.698 Million pixels/second
16.445 Million pixels/second
1144.086 Million pixels/second
1140.916 Million pixels/second
35.041 Million pixels/second
335.446 Million pixels/second
9.732 Million pixels/second
146.559 Million pixels/second
365.189

I can't explain the low DIVE results.

My WSeB takes advantage of both of my processors but had disappointing benchmark results showing to be somewhat slower than Warp4. WSeB SMP benchmark tests:

At 16 bits per pixel: 643.813 PM-Graphics-marks 206.522 DIVE-marks.

At 32 bits per pixel: 290.459 PM-Graphics-marks 187.477 DIVE-marks.

I wrote SDD about this and was told it was a known problem that would be eventually fixed.

One can wonder if a faster cpu would have much of an effect on these benchmarks. I ran the following on the AMD Duron 750 MHz at 16 bits per pixel at 800 by 600:

Graphics card - NVidia - GeForce 2 MX, 32MB, AGP

DIVE tests
Vid Mem
Video Bus Bandwidth
DIVE fun
M->S DD 1.00:1
Dive-marks
32MBs
63.212 Megabytes/second
215.629 fps normalised to 640x480x256
215.655 fps normalised to 640x480x256
80.611
Video tests - Resolution800x600x16 bits/pixel
Vid Mem
BitBlt S->S Copy
BitBlt M->S Copy
Filled Rect.
Pattern Fill
Vert. Lines
Horiz. Lines
Diag. Lines
Text Render
PM-marks
32 MB
595.747 Million pixels/second
32.477 Million pixels/second
2405.134 Million pixels/second
2394.510 Million pixels/second
54.230 Million pixels/second
372.011 Million pixels/second
11.315 Million pixels/second
147.687 Million pixels/second
705.671

I then overclocked the 750 Mhz to 918 Mhz (9*102 Mhz) with the same screen size and color depth:

Graphics card - NVidia - GeForce 2 MX, 32MB, AGP

DIVE tests
Vid Mem
Video Bus Bandwidth
DIVE fun
M->S DD 1.00:1
Dive-marks
32MB
64.430 Megabytes/second
219.841 fps normalised to 640x480x256
219.922 fps normalised to 640x480x256
82.194
Video tests - Resolution800x600x16 bits/pixel
Vid Mem
BitBlt S->S Copy
BitBlt M->S Copy
Filled Rect.
Pattern Fill
Vert. Lines
Horiz. Lines
Diag. Lines
Text Render
PM-marks
32 MB
595.747 Million pixels/second
33.131 Million pixels/second
2404.909 Million pixels/second
2394.510 Million pixels/second
54.230 Million pixels/second
443.195 Million pixels/second
11.650 Million pixels/second
160.607 Million pixels/second
714.775

There is not much of a difference considering the cpu speed increase.

Conclusion.

This is a fast card that is reasonably priced at around 120$US. It offers the advantage of having drivers for all the operating systems one is likely to use. It offers quite fast performance in 3D for those interested in gaming under the Windows environment. While a new IBM SDD driver has recently appeared, it does not yet support this chipset and one needs to get Scitech's SDD driver. AFAIK, the driver can only be purchased directly from Scitech's web site: http://www.scitechsoft.com/Merchant/merchant.mv?Screen=CTGY&Store_Code=scitech&Category_Code=Consumer and costs around 40$US. I am not an employee of any company mentioned here nor own stock in any of them. What I have written is only to relate my experience so others may have an idea of what to expect should they be tempted to follow my experience.

Features
editor@os2voice.org
[Previous Page ] [ Index] [Next Page ]
VOICE Home Page: http://www.os2voice.org