Hi: I've been running the df_dok benchmark program in my VIA ProSavage PN133 (8d01) based system, with and without the --system parameter. Here are the comparative results:
Benchmarking with 256x256 in 16bit mode... (16bit) TEST VIDEO SYSTEM UNITS Anti-aliased Text 30.789 140.166 KChars/sec Anti-aliased Text (blend) 14.561 26.623 KChars/sec Fill Rectangle 317.671 115.587 MPixel/sec Fill Rectangle (blend) 2.553 10.711 MPixel/sec Fill Rectangles [10] 317.621 117.764 MPixel/sec Fill Rectangles [10] (blend) 2.589 10.733 MPixel/sec Fill Triangles 162.808 104.857 MPixel/sec Fill Triangles (blend) 2.537 10.122 MPixel/sec Draw Rectangle 40.219 12.071 KRects/sec Draw Rectangle (blend) 0.836 1.752 KRects/sec Draw Lines [10] 145.951 39.640 KLines/sec Draw Lines [10] (blend) 3.738 7.135 KLines/sec Fill Spans 309.893 104.683 MPixel/sec Fill Spans (blend) 2.563 10.526 MPixel/sec Blit 161.259 99.036 MPixel/sec Blit colorkeyed 89.716 74.376 MPixel/sec Blit destination colorkeyed 3.598 101.288 MPixel/sec Blit with format conversion 13.351 19.181 MPixel/sec Blit with colorizing 9.413 12.192 MPixel/sec Blit from 32bit (blend) 2.254 6.744 MPixel/sec Blit from 32bit (blend) with colorizing 2.183 5.758 MPixel/sec Stretch Blit 32.475 80.754 MPixel/sec Stretch Blit colorkeyed 18.786 57.171 MPixel/sec As you can see, while there is a performance boost in most non-blend tests, results shows that blend ant text tests runs between 2 and 4 times slower when using video memory instead of system memory! Is this a limitation of the graphics card? Or does the savage driver need to be improved? Thanks in advance. _______________________________________________ directfb-users mailing list [email protected] http://mail.directfb.org/cgi-bin/mailman/listinfo/directfb-users
