In the ARM architecture:
In the ARM architecture, the GPU does not carry its own memory, but is often reserved for the GPU as the memory in the system's Ram. The result is that the GPU uses the AHB Bus for video memory operations, and the LCD periodically switches from the frame
The cache refresh must also go through the AHB Bus. This will lead to the busy AHB Bus, resulting in performance bottlenecks.
The frequency of the AHB Bus is 133 MHz, the memory is used, the screen resolution is 800*600, and the 32-bit color is used.
First, the LCD obtains screen data from the SDRAM through DMA, which is highly efficient in storage.
The memory throughput is calculated first. The frequency of the SDRAM is 133 MHz, and the width of the Data Bus is 32 bits, the theoretical throughput of SDRAM is 133*4 = 512 Mb/s.
Data Volume of one LCD frame: 800*600*4 = 1,920,000.
The time required for the LCD to transmit a frame through the AHB Bus is 1.92 M/512 M = 3.75 Ms.
If the LCD refresh frequency is 60Hz, the time spent by the AHB Bus to refresh the screen on the LCD is 3.75*60 = 225 ms in one second. It can be said that 1/5 of the time is gone.
In the PC architecture:
In the PC architecture, the GPU itself has a display memory, which is the most important difference with the ARM architecture. In this way, the LCD can be directly connected to the GPU when refreshing the screen content, without the need for an internal bus. The LCD refresh frequency does not affect the internal bus.
The LCD refresh frequency is optimized for the ARM platform.
1. reduce the volume of data per frame: reduce the screen resolution and the number of digits per pixel.
2. Reduce the LCD refresh frequency. There is no need to achieve 60Hz for the LCD.
3,
If the LCD supports asynchronous local refresh, it can only refresh the changed content, which will reduce the amount of bus data transmission.
4. Use higher-speed memory, such as DDR, or increase the frequency of the AHB Bus.