The following options are of particular interest for the SiS driver. Each of them must be specified in the `svga' driver section of the XF86Config file, within the Screen subsections of the depths to which they are applicable (you can enable options for all depths by specifying them in the Device section).
This option lets you to modify the memory clocking of your card. (only for 5597 and 6326) Modifying the memory timings can destroy the device, but usually the only ill effects of overclocking is to have some noise an drawing errors, but BE CAREFUL. Usually a little increment can improve the drawing speed, and allows also higher dotclocks. The server reports default memclock on starting messages, so take it as a base. Units are in MHZ.
This option lets you to modify the maximum allowed dotclock (only for 5597 and 6326). Without it, the server makes a conservative guess based on memory clock, speed and number of banks. If your monitor supports higher dotclocks and you know that your card can do it, give a try. If the speed is too high for your configuration (but not for your monitor), the effects can vary from some noise on screen to a black screen. Don't use speeds greater than 135 Mhz, (175 for 6326), even if your monitor supports the dot-clock.
By default the XAA (XFree86 Acceleration Architecture) is used. This option will disable the use of the XAA and will enable the old BitBlt acceleration operations. (see below).
On chips 86c202 and later, the default is to use the programmable clock for all clocks. It is possible to use the fixed clocks supported by the chip instead of using this option (manufacturer dependent).
The default is for using the hardware cursor.
By default linear addressing is used on all chips. However this might be broken in some implementations. It is possible to turn the linear addressing off with this option. Note that H/W acceleration and 16/24bpp are only supported with linear addressing.
This option will disable the use of all the BitBLT engine. It is useful for problems related to acceleration problems. In general this will result in a reduced performance.
It is useful for problems related to image writing, and possible stipple acceleration problems. In general this will result in a reduced performance.
5597/8 and 6326 have the option to extend the engine command queue on VRAM. With extended queue length, the driver only checks queue status on some color-expansion commands. This gives some performance improvement, but is possible to lose some commands, corrupting screen output. As the size of extended command queue is 16-32K, the probability is very low, but exists. The performance gain observed is around 8-10%. Currently, using this option with xaa_benchmark freezes the acceleration engine, causing weird image display.
This set a bit on some registers. Although documented, the utility of this option is unknown for me. I can't see any difference on stability or performance.
Enables 1 cycle memory access. Try it. Increased memory bandwidth reduces the possibility of glitches and noise on high resolution modes.
These options modify the arbitration thresholds on CRT FIFO. Fifo_aggressive gives more time to CPU for accessing the VRAM. Fifo_conservative reduces the possibility of noise caused when the CRT tries to read memory when it is used by CPU, but reduces performance. The default is between aggressive and moderate (more aggressive than moderate).