home *** CD-ROM | disk | FTP | other *** search
- Documentation for VgaSpeed VGA Graphics Benchmark Program
-
- By: J. Rockford Cogar
- 119 Oklahoma Avenue
- Oak Ridge, TN 37830
-
- Use: VgaSpeed is FreeWare. This program may be used any way you wish,
- as long as you do not charge any money for the program itself.
-
- I. Introduction.
-
- The program VgaSpeed is a VGA Graphics Benchmark Program. What VgaSpeed
- measures is how fast a given computer system can generate VGA graphics. There
- are several important hardware characteristics that determine how fast a
- computer can do graphics operations:
-
- 1. CPU Speed
- 2. Expansion Bus Speed
- 3. Number of Video Wait States
- 4. Speed of Internal VGA Card Operations
-
- An example of a fast graphics machine would include the following: a 33 mhz
- 80486 CPU, a 10 mhz bus speed, very few video wait states and a 'fast' VGA
- card.
-
- Because graphics operations on VGA cards are mostly 8 bit I/O operations, 16
- bit bus connectors on VGA cards do not appear to greatly improve the speed of
- graphics.
-
- The results of VgaSpeed are given as relative to the graphics performance of
- an IBM PS/2 Model 60 VGA system as a VS (VgaSpeed) number.
-
- A VS of 2.00, would show that the current system has the ability to do VGA
- graphics twice as fast as an IBM PS/2 Model 60 VGA. The fastest 80486 system
- that has been measured gave a VS of about 2.70. Typical 80386SX VGA systems
- yield VS numbers of about 1.40. It is around VS numbers equal to 1.40 that
- Microsoft Windows 3.0 starts to look fairly good.
-
-
- II. The Tests.
-
- VgaSpeed is made up of four different graphics tests: Write Mode Zero Dot
- Test, Write Mode Two Dot Test, Bit Map Move Test and Bit Map Text Test.
-
- The graphics code in VgaSpeed was written in assembly language from
- documentation in the IBM EGA and VGA Technical Reference Guides. Every effort
- was made to make the code as fast as possible. After all, the goal is too
- find out how fast a given system can do VGA graphics, not how fast the the
- system can run a bunch of code through the CPU.
-
- VgaSpeed was written using Borland Turbo Assembler 2.0 and Borland Turbo
- 'C++' 1.0.
-
- By default, all tests are executed while the BIOS video mode was '16'. This
- is the same as EGA 640 pixels horizontal by 350 pixels vertical 16 color
- mode. This allows VgaSpeed to be run on EGA video systems.
-
- When VgaSpeed is run in VGA 640 x 480, 16 color mode (by using the 'v'
- command line switch) the VS number will be almost exactly the same as a test
- of EGA 640 x 360, 16 color mode. The graphics testing takes about 40
- seconds on a fast 80486 system to 140 seconds on an 8mhz AT system.
-
- 1. Write Mode Zero Dot Test.
-
- This test draws 817,452 dots (pixels) on the graphics screen using
- hardware write mode zero. Hardware write mode zero is the write mode used
- by most graphics programs. It appears that Microsoft Windows 3.0 uses
- write mode for all in all EGA and VGA type driver programs.
-
- 2. Write Mode Two Dot Test.
-
- This test draws 1,634,904 dots (pixels) on the graphics screen using
- hardware write mode two. In general, the process of writing a pixel to the
- graphics screen in write mode two is twice as fast as one written using
- write mode zero. Hardware write mode two is the mode used by most very
- high performance graphics programs. For example, all of the EGA/VGA
- graphics based data acquistion software from Tennelec/Nucleus Inc. (Oak
- Ridge TN), use hardware write mode two. Some of the faster commercial CAD
- packages also appear to use hardware write mode two.
-
- 3. Bit Map Move Test.
-
- This test does 504 bitmap block moves. The size of the bitmap is 30,102
- pixels (a 173 by 174 rectangle). The pattern leftover from the write mode
- two dot test is copied from video RAM to system RAM, then from system RAM
- to a different location in video RAM. This process is repeated 252 times.
- The hardware write mode here is 'zero'. It is likely that most all
- programs that use bit map moves to/from video RAM to/from system RAM use
- hardware write mode zero to do so.
-
- Because this operation involves memory block moves (using the: REP MOVSW
- operation), the number of memory wait states of the source and target
- memory can be a major factor in the speed of this test. In addition,
- computers with expansion bus speeds greater than the standard 8 mhz, do
- well on this test.
-
- 4. Bit Map Text Test.
-
- This test shows how fast text can be displayed on a graphics mode screen.
- The Bit Map Text Test writes a 39 character string in an 8 pixel tall font
- to the graphics screen 2048 times. In addition, a 40 character string is
- written to the screen in an 14 pixel tall font 2048 times. Both fonts are
- 8 pixels wide. All this means that 14,286,848 pixels get written on the
- screen in this test! The hardware write mode used in this here is 'two'.
-
-
- II. System Requirements.
-
- 1. A color VGA or EGA video system with at least 128KB of video RAM.
- 2. DOS version 2.0 or greater.
- 3. 64KB of free RAM.
-
-