Steve:
I have interfaced numerous graphics LCDs using the concept I described. The scheme, using a Cygnal processor with good performance gets the display rewritten in about 5 milliseconds on a 128x64 display.

I have never experienced any flicker using this technique!!

The font writing into an XRAM buffer is far more efficient than trying to use low level memory reading/writing direct to the display controller. A process to perform read/modify/write to merge pixels into a display controller for fonts that cross the byte boundaries of the display controller memory are terribly inefficient. In my scheme the XRAM buffer image is not squirted to the display controller until the "application" has painted all of the current display image into the XRAM buffer. And then when that squirt takes place the memory pointers are set in the display controller a minimum number of times leading to virtually all of the transfer bandwidth to the display controller being raw data as opposed to control and addressing information.

Another factor to keep in mind is that some of the graphics display controllers support a serial interface mode that procludes the ability to communicate to the display in a read/write manner. I have had several of this type of interface wherein the write only capability in a serial manner actually required the use of the bitmapped memory buffer so that display updates of pixel chunks across byte boundaries could be supported.

Keep in mind that many currently popular graphics LCD modules do not have multi-page graphics memories on board. Particularly those using COG single chip technology. In my experience I have noted that displays with the Toshiba T6963 display controller do indeed offer multi page capabilities but these are rarely single chip solutions, rarely chip-on-glass (COG) technology, and tend to be the more expensive modules.

Now all that said I do agree that there are some applications that can make good use of a graphics LCD module where in a small chunk of XRAM memory for a bitmap buffer is not available. In these cases a display using a T6963C controller would offer a suitable solution, but I fully suspect that the additional cost of this display module would be more than choosing a microcontroller that has 1K, 2K, 4K or even 8K of on-board XRAM.

Michael Karas