Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!uunet!mcsun!unido!gmdzi!strobl From: strobl@gmdzi.gmd.de (Wolfgang Strobl) Newsgroups: comp.windows.ms.programmer Subject: 256-color drivers are NOT always slower than 16-color drivers Message-ID: <3537@gmdzi.gmd.de> Date: 1 Nov 90 14:16:08 GMT Organization: GMD, Sankt Augustin, F. R. Germany Lines: 80 As far as I know, the cheap SVGA cards run the 16 color modes with 8 bit wide video ram access, even if they are 16 bit cards and sit in a 16 bit slot. For the 256 color modes, this changes, these modes allow real 16 bit access. Because of this, I don't see why there should be much speed difference between 16 color and 256 color drivers of identical resolution. I did a little bit of research to check this out. First I compared the raw speed of the two modes 640*480*16 and 640*480*256 using John Bridges VIDSPEED utility. This confirmed that the memory access in the 256 color mode is twice as fast than in the 16 color mode (on my card, of course). Then I used Charles Petzold's DispInfo benchmark (ver 1.2, 1986) to compare the speed of two 640*480 drivers, one 16 color, the other with 256 colors. The results are quite interesting and confirm Bert Tyler's observation. For simple operations, the 256 color driver seems to be a little bit faster than the 16 color driver! I have no theory why there is such a big difference in the speed of the "rectangles" and "fill" operation. Both are different from the three other operations in that in the actual benchmark they paint a big part of the window ("reactangles") or the whole window ("fill"), while "BitBlt" and "scroll" move or paint small rectangles of fixed size, and the "lines" touch few pixels, in comparison with the other operations. Wolfgang Strobl #include PS: does anybody know whether there is a newer version of Petzolds benchmark - the version I have is a bit old. ---------------------------------------------------------------- The results. Environment Noname 386/25 cache, 8MB Optima 1024A Plus (Tseng ET3000 based 16 bit card) ---------------------------------------------------------------- Result of checking the raw output speed in both the 16 color and the 256 color mode. char/sec ratio 16 colors 652 1.0 256 colors 1248 1.9 ---------------------------------------------------------------- Result of Petzolds benchmark, test window enlarged to fullscreen, Windows running in enhanced mode: 1. standard Windows 3.0 640*480*16 vga driver 2. Orchid Prodesigner 640*480*256 from win30prod.zip (cica) lines BitBlt scroll rectangles fill 16 colors 3.2 8.0 8.6 14.8 68.0 256 colors 3.5 5.9 9.6 34.0 229.0 (all times in ms per operation) lines: drawing random lines of random length into the window BitBlt: painting a small rectangle of fixed size (50 pixel?) to random locations on the screen scroll: moving the same small rectangle one pixel down or to the right rectangles: painting rectangles of random sizes and random colors fill: filling the whole window with one color. ----------------------------------------------------------------